Product and Process Design Driving Innovation

Jan Harmsen, André B. de Haan, Pieter L. J.
Swinkels
Product and Process Design
Also of Interest
Chemical Product Technology.
Murzin, 2018
ISBN 978-3-11-047531-9, e-ISBN 978-3-11-047552-4
Chemical Process Synthesis.

Connecting Chemical with Systems Engineering Procedures
Bezerra, 2018
ISBN 978-3-11-046825-0, e-ISBN 978-3-11-046826-7
Process Synthesis and Process Intensification.

Methodological Approaches
Rong (Ed.), 2017
ISBN 978-3-11-046505-1, e-ISBN 978-3-11-046506-8
Process Engineering.
Addressing the Gap between Study and Chemical Industry
Kleiber, 2016
ISBN 978-3-11-031209-6, e-ISBN 978-3-11-031211-9
Jan Harmsen, André B. de Haan,
Pieter L. J. Swinkels
Product and
Process Design
|
Driving Innovation
Authors
ir. Jan Harmsen
Harmsen Consultancy B.V.
Hoofdweg Zuid 18
2912 ED Nieuwerkerk aan den IJssel
The Netherlands
prof. dr. ir. André B. de Haan

Delft University of Technology
Department of Chemical Engineering
Section Transport Phenomena
Van der Maasweg 9
2629 HZ Delft
The Netherlands
ir. Pieter L. J. Swinkels

Managing Director PDEng programmes
Faculty of Applied Sciences
Delft University of Technology
Departments of Chemical Engineering and Biotechnology
Van der Maasweg 9
2629 HZ Delft
The Netherlands
ISBN 978-3-11-046772-7
e-ISBN (PDF) 978-3-11-046774-1
e-ISBN (EPUB) 978-3-11-046775-8
Library of Congress Control Number: 2018934236
Bibliographic information published by the Deutsche Nationalbibliothek

The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie;
detailed bibliographic data are available on the Internet at http://dnb.dnb.de.
© 2018 Walter de Gruyter GmbH, Berlin/Boston

Cover image: Bart van Overbeeke Fotografie
Typesetting: le-tex publishing services GmbH, Leipzig
Printing and binding: CPI books GmbH, Leck
www.degruyter.com
Preface
We, the authors, have extensive experience in designing and developing innovative
products and processes, for a broad variety of companies. Some of these companies
are strongly product driven, while others are strongly process driven, and some by
both products and processes. The companies belong to various industries, such as oil
and gas, consumer products, chemicals, life sciences, food, and engineering, procure-
ment, and construction.
We also have extensive experience in teaching design methodology for products
and processes both in universities and at companies. Coming from these combined
industrial and academic experiences we felt a strong need to write a book covering all
relevant design methods and heuristics focused on innovative solutions to establish
breakthrough as well as adapted innovations. This book can be used as a textbook
for teaching in academia and industries but also for inspiring industrial product and
process engineers to create novel designs.
From our industrial experiences we have gleaned that industrial innovation can
be tremendously accelerated by incorporating design activities throughout the inno-
vation process to drive projects from early-stage idea generation to commercial imple-
mentation. Our academic experience has taught us how to unlock and develop the
design skills of students and enable them to generate novel concepts for products and
processes.
These experiences, complemented by recent insights into design and innovation
in the literature, have been used to create this book. During this process, a lot of dis-
cussion took place regarding the content, the connections between the chapters, and
the format and structure of the book. We felt privileged that such a large group, with
different knowledge and experiences but with the same conviction and motivation,
were in a position to cooperate with us to complete this book.
We would like to acknowledge the people who contributed to this book: Johan
Grievink for writing Chapter 9 and contributing to Chapters 4 and 8; Ido Pat-El, for
writing Chapter 11 on health and safety, and Section 5.7 on designing for human fac-
tors; Jos van Reisen, for writing Section 5.8, on designing for process energy efficiency;
Christhian Almeida Rivera, for contributing to Chapters 4 and 8 and proofreading
Chapter 9; Rob van der Lans for contributing to the characteristic times of Chapter 8;
Gabrie Meesters for contributing to Chapters 3 and 8; Mark Toonen, Innovation Man-
ager at Cosun, for providing information on innovation management in practice; Tom
Brooijmans of Cosun Biobased Products for reviewing Chapter 10; Gijsbert Korevaar,
Asst Prof. of energy, services and systems, TU Delft, for reviewing Chapter 12; Maarten
Verkerk, Professor at TU Eindhoven, for reviewing the use of modal aspects in Chap-
ter 3; Yvette van Herwijnen, of AMI consultancy, for providing input on Chapter 3 on
project management.
https://doi.org/10.1515/9783110467741-201
VI | Preface
We thank Giljam Bierman for many years of contribution in design projects and
design methods at the (post-MSc) PDEng level at TU Delft, and for contributing this
expertise in Chapter 4. We thank the following people for their inspiration and coop-
eration in product and process design at TU Delft over the last 25 years:
Hans Wesselingh, Peter Appel, Gert Frens, Johan Grievink, Cees Luteijn, Gil-
jam Bierman, Peter Jansens, Jacob Moulijn, Freek Kapteijn, Ernst Sudholter, Michiel
Kreutzer, Ruud van Ommen, Ger Koper, Margot Weijnen, Paulien Herder, Peter Bongers,
Cristhian Almeida Rivera, Gabrie Meesters, Rafiqul Gani, Kees van Overveld, Adrie
Huesman, Henk van den Berg, Jan Dijk, Jeu Lambrichts, Peter Daudey, Urjan Jacobs,
Henk Nugteren, Peter Hamersma, Jos van Reisen, Sorin Bildea, Tony Kiss, Karel As-
selbergs, Luuk van der Wielen, Adrie Straathof, Mark van Loosdrecht, Sef Heijnen,
Marcel Ottens, Maria Cuellar Soares, Geert Jan Witkamp, Henk Noorman, Peter Verhei-
jen, as well as the many, many MSc students in chemical engineering, PDEng trainees
in process & equipment design, bioprocess engineering, chemical product design,
and PhD students in process systems engineering over the last 25 years.
We thank the photographer Bart van Overbeeke, DuPho, Dutch Professional Pho-
tographers, for his photograph on the book cover and David Vervloet (PhD student),
and his supervisors Ruud van Ommen and Freek Kapteijn, for allowing their experi-
mental reactor setup to be photographed. We thank Marjo Dekker, Moments of Mem-
ory fotografie, for her photograph of Jan Harmsen and we thank Harm P.W.L. Swinkels
for his photograph of Pieter Swinkels.
Contents
Preface | V
Authors’ biographies | XIX
Part A: Innovation and industry
1 Goal, scope, and structure | 3

1.1 Goal | 3
1.2 Scope | 5
1.2.1 Products classification and description | 5
1.2.2 Description of process industries | 5
1.3 Book structure | 6
2 Systems relevant to design for innovation | 7

2.1 Introduction to all relevant system levels | 7
2.2 Earth system level | 8
2.3 Society system level | 9
2.4 Technosphere system level | 10
2.5 Value chain system level | 11
2.6 Industrial symbiosis system level | 12
2.7 Industrial complex system level | 13
2.8 Factory system level | 13
2.9 Process system level | 14
2.10 Product system level | 14
2.11 Process step system level | 14
2.12 Unit operation system level | 15
2.13 Main equipment system level | 15
2.14 Characteristic process subsystem level | 15
2.15 Microelement system level | 16
2.16 Elementary system level | 16
3 Managing innovation | 17
3.1 Overview | 17
3.1.1 Innovation terms | 17
3.1.2 Stage-gate approach | 17
3.1.2.1 Short description of stage characteristics | 18
3.2 Business focus and motives for innovation | 19
3.2.1 Business focus trends in product and process innovation | 19
VIII | Contents
3.2.2 Business motives for innovation | 20

3.2.2.1 Competition as innovation driver | 20
3.2.2.2 Learning curve as innovation driver | 21
3.2.2.3 Circular economy as innovation driver | 22
3.2.2.4 Sustainable development as innovation driver | 22
3.2.2.5 Corporate social responsibility as innovation driver | 23
3.2.2.6 World problems as driver for the product-process
industries | 24
3.3 Innovation classes and types | 25
3.3.1 Innovation classes | 25
3.3.2 Innovation by serendipity | 28
3.3.3 Social innovations | 29
3.4 Innovation partners | 29
3.4.1 In-house versus open innovation | 29
3.4.2 Innovation partners | 30
3.4.3 Universities | 30
3.4.4 Contract research organizations | 31
3.4.5 Startup innovators | 31
3.4.6 Technology providers | 32
3.4.7 Equipment providers | 32
3.4.8 Engineering procurement construction contractors | 32
3.4.9 Tolling manufacturers | 33
3.5 Portfolio innovation management | 33
3.5.1 Objectives of portfolio management | 33
3.5.2 Ambition matrix for innovation portfolio management | 34
3.5.2.1 Core cluster | 34
3.5.2.2 Adjacent cluster | 35
3.5.2.3 Transformational cluster | 36
3.5.2.4 Budget distribution over clusters | 38
3.5.3 Risk adjusted value for innovation management | 38
3.5.3.1 Risk adjusted value in development stage and beyond:
crossing the valley of death | 38
3.5.4 Innovation management guidelines for small enterprises | 39
3.5.4.1 Innovation guidelines for small enterprises | 39
3.5.4.2 Guidelines for breakthrough innovation companies | 39
3.5.5 Project failures and their causes | 40
3.5.5.1 Innovation failure statistics | 41
3.5.5.2 Causes of project failures | 41
3.5.5.3 Megaproject failures | 41
3.5.5.4 Project failures in detailed engineering stage | 42
3.6 Project management of innovation | 42
3.6.1 Objectives of innovation project management | 42
Contents | IX
3.6.2 Project management by stage-gate system | 43

3.6.2.1 Product and process innovation stages: general aspects | 43
3.6.2.2 End-of-life stage | 44
3.6.3 Project entries to stages by technology readiness level
method | 44
3.6.4 Team formation | 47
3.6.4.1 Team member behavior characteristics required over the stages
and teams | 47
3.6.4.2 Group design with Belbin team roles | 47
3.6.5 Intellectual property creation and protection | 48
3.7 Discovery stage | 48
3.7.1 Purpose discovery stage | 48
3.7.2 Managing projects in discovery stage | 49
3.7.3 Creativity methods | 49
3.7.4 Discovery stage-gate decision evaluation criteria | 50
3.7.5 Idea description | 50
3.7.6 Proof of principle | 50
3.7.7 Business case | 50
3.7.8 Creating value | 51
3.7.9 Strategic fit | 51
3.7.10 Necessity of the idea | 51
3.7.11 Idea development doable (with others) | 51
3.8 Concept stage | 51
3.8.1 Purpose of concept stage | 51
3.8.2 Customer value proposition in concept stage | 52
3.8.3 Technical product requirements in concept stage | 53
3.8.4 Process concept generation | 53
3.8.5 Proof of process concept | 53
3.8.6 Selection of best process alternative | 53
3.8.7 Process concept design | 54
3.8.8 Basic design data generation | 54
3.8.9 Concept stage gate evaluation | 54
3.9 Feasibility stage | 54
3.9.1 Concurrent product-process design and testing | 54
3.9.2 Microplant in feasibility stage | 55
3.9.3 Scale-up strategy and information | 56
3.9.4 Scale-up information equipment | 58
3.9.5 Business case feasibility stage | 58
3.10 Development stage | 58
3.10.1 Purpose development stage | 58
3.10.2 Product testing in development stage | 58
3.10.3 Process testing in development stage | 59
X | Contents
3.10.3.1 Miniplant purpose and design | 59

3.10.3.2 Pilot plant purposes and design | 59
3.10.3.3 Pilot plant as downscaled version commercial scale design | 60
3.10.4 Pilot plant engineering, procurement, and construction (EPC)
company choice | 62
3.10.4.1 Equipment scale-up effects determination | 62
3.10.4.2 Phenomena in equipment sensitive to scale-up and influencing
performance | 63
3.10.4.3 Mockup design and testing for hydrodynamic scale-up
effects | 63
3.10.4.4 Development stage gate evaluation | 64
3.11 Detailed design stage | 64
3.11.1 Detailed product design | 64
3.11.2 Detailed process engineering | 64
3.11.3 Choice of EPC contractor | 65
3.11.4 Demonstration plant | 66
3.12 Process startup and product launch | 67
3.12.1 Introduction | 67
3.12.2 Panel for product testing | 67
3.12.3 Product launch planning | 67
3.12.4 Matching marketing and manufacturing timing | 67
3.12.5 Information to supply chain and customers | 68
3.12.6 Process implementation | 68
3.12.7 Recognition of new commercial implementations requires special
preparation | 68
3.12.8 Case of Shell Moerdijk where new catalyst charge leads
to explosion | 69
3.12.9 First commercial scale process startup | 70
3.12.9.1 Startup preparation | 70
3.12.9.2 Startup manual | 70
4 Designing for innovation | 75

4.1 Introduction | 75
4.2 Design thinking | 75
4.2.1 Design for innovation theory | 75
4.2.1.1 Design and risks | 76
4.2.1.2 Design links with society and nature | 77
4.2.1.3 Further reading on design | 77
4.2.2 Design knowledge types | 77
4.2.3 Differences between design thinking and scientific
research | 78
4.3 Designing for innovative products and related processes | 78
Contents | XI
4.3.2 Product-process design method: Delft Design Map (DDM) | 78
4.3.3 Explaining the Delft Design Map for product-process
design | 79
4.3.3.1 Design (cycle) steps description: | 81
4.3.3.2 Executing design activities in all twelve design levels | 83
4.3.4 How to plan and execute design activities using the Delft Design
Map | 88
4.3.4.1 Practical benefits of working with the Delft Design Map | 90
4.3.5 Design planning with Delft Design Map for various innovation
classes | 90
4.3.5.1 Fingerprints of Delft Design Map for design task planning | 90
4.4 Planning for design and experimentation in innovation | 94
4.4.1 Decision sequence ranking method | 94
4.5 Embedding design by criteria and context setting | 95
4.5.1 Introduction to the purpose of criteria and context setting | 95
4.5.2 Comprehensive list of modal aspects for defining criteria from
modal aspects of reality | 95
4.5.3 SHEETS criteria list | 97
4.6 Role of modeling and simulation in concurrent design | 97
4.7 Exploiting experience in design (design heuristics) | 99
4.7.1 Strength of design heuristics | 99
4.7.2 Weaknesses in using design heuristics | 100
4.7.3 Tapping into experience | 100
4.8 Industrial example of design driving innovation | 101
Part B: Design generation
5 Special design approaches | 107

5.1 Introduction to special design approaches | 107
5.2 Design for biomimicry | 107
5.3 Design for industrial ecology | 108
5.4 Design for circular economy | 108
5.5 Design for cradle-to-cradle | 109
5.6 Design for industrial symbiosis | 109
5.7 Design for renewable energy sources | 111
5.7.1 Present bulk chemicals production from fossil fuel
resources | 111
5.7.2 Design for renewable energy based on product-process chain
methods | 111
5.8 Design for energy efficiency | 112
XII | Contents
5.9 Design for human factors | 119

5.10 Design for six sigma | 121
6 Scoping the design | 123

6.1 Introduction to scoping the design | 123
6.2 Defining design goal and name | 123
6.2.1 Purpose design goals and names | 123
6.2.1.1 Design goal setting | 123
6.3 Defining the design scope (system levels, boundaries, and
context) | 124
6.4 Defining constraints (specifications) | 125
6.4.1 Design constraints | 125
6.4.2 Product specifications | 126
6.4.3 Process specifications | 127
6.4.3.1 Identifying reference cases to improve design constraints | 127
6.5 Generating basic design data | 128
6.5.1 Ideation stage | 129
6.5.2 Concept stage | 130
6.5.3 Data generation feasibility stage | 131
6.5.4 Development stage | 134
7 Executing designs | 136

7.1 Introduction to executing designs | 136
7.2 Synthesizing preliminary solutions | 137
7.2.1 Synthesizing solutions in general using heuristics | 137
7.2.2 Synthesizing products | 138
7.2.2.1 Product synthesizing design using simple heuristics | 138
7.2.2.2 Synthesizing physically structured products in concept
stage | 138
7.2.2.3 From functional product specification to product structure | 139
7.2.3 Synthesizing processes | 142
7.2.3.1 Radically novel process synthesis | 142
7.2.3.2 Modified process synthesis | 143
7.2.3.3 Media choice | 143
7.2.3.4 Unit operation choice | 143
7.2.3.5 Heuristics for reaction plus separation synthesis | 144
7.2.3.6 Heuristics for reactive distillation synthesis | 144
7.2.3.7 Heuristics for separation synthesis | 144
7.2.3.8 Process separation synthesis from complex input | 145
7.2.3.9 Choice of batch processing versus continuous processing | 145
7.3 Analyze interim solutions | 147
7.3.1 Introduction to analysis | 147
Contents | XIII
7.3.2 Qualitative analysis of interim solutions | 148

7.3.3 Quantitative analysis of interim concept solutions | 149
7.3.3.1 Safety and health | 149
7.3.3.2 Environmental analysis | 149
7.3.3.3 Sustainable development analysis | 150
7.3.3.4 Economics rapid analysis in the concept stage | 150
7.3.3.5 Role of modeling and analysis | 150
7.4 Best selection from alternative concept solutions | 150
7.4.1 Selecting best process concept option | 152
7.4.2 Economic ranking of process concepts | 152
7.5 Finalizing a design | 154
7.5.2 Balancing designs | 155
7.5.3 Increasing robustness toward future uncertainties using
scenarios | 155
7.5.4 Intellectual property (IP) creation and protection | 156
7.6 Evaluating and reporting designs | 157
7.6.2 Reference case for evaluations | 157
7.6.3 Evaluating designs | 158
7.6.4 Reporting | 158
8 Product modeling and optimization | 160

8.1 Verbal, schematic, mathematical, and physical models | 160
8.2 Process design schematic and mathematical models useful for product
design | 162
8.3 Product design schematic models | 165
8.3.1 House of Quality model for consumer function and property
function | 165
8.3.2 Understanding the product application/use process: visualizing
systems splits | 167
8.4 Mathematical models for consumer and property functions | 170
8.4.1 Characteristic times and regime analysis | 170
8.4.1.1 Estimation skills – Fermi problems | 173
8.4.1.2 ‘Systemic’ time constants – eigenvalues | 173
8.4.2 Data driven nonlinear product modeling: artificial neural
networks | 174
8.4.3 Scientific models for product state and behavior | 174
8.4.3.1 Constituents | 174
8.4.3.2 Product structure | 175
8.4.3.3 Interfaces between dispersed phases | 177
8.4.4 Product structure matrix | 178
XIV | Contents
8.5 Relations between product and process modeling | 180

8.5.1 Causal flow of information on process – product modeling | 180
8.5.1.1 The product manufacturing ⇒ product structure and state
model | 181
8.5.1.2 A product structure and state ⇒ product properties model | 182
8.5.1.3 The product property ⇒ customer qualities (attributes)
model | 182
8.5.1.4 Inter-connectivity between product and process models | 182
8.5.1.5 Mathematical model-based product-process optimization | 183
8.5.2 Notes on first-principles models of formation of a structured
product | 185
8.6 Product models: overview | 188
8.7 Modeling for ‘Safety, Health, Environment, Economy, Technology,
Social (SHEETS)’ | 190
Part C: Design optimization
9 Process modeling and optimization | 197

9.1 Justification and objectives of process modeling | 197
9.2 Contributions to a concept stage with linear modeling | 200
9.2.1 Modeling for a concept stage | 200
9.2.2 Cases for concept stage modeling | 202
9.3 Nonlinear process model simulations:
contributions to a development stage | 206
9.3.1 Nonlinear modeling and simulation for a development
stage | 206
9.3.1.1 Process representation | 207
9.3.1.2 Product modeling | 208
9.3.1.3 Process equipment modeling | 208
9.3.1.4 Scope of a process model | 208
9.3.1.5 Process analysis scenarios | 210
9.3.1.6 Process performance evaluation metrics | 210
9.3.1.7 Sensitivity analyses | 211
9.3.1.8 Process design and synthesis cases with targets | 211
9.3.1.9 Additional specification of an external scenario for a
design | 212
9.3.1.10 Consistency in model formulation | 213
9.3.1.11 Process model computations | 214
9.3.1.12 Analysis of obtained model solutions | 215
9.3.1.13 Process performance evaluation metrics | 216
9.3.1.14 Using uncertainty information in a sensitivity analysis | 216
Contents | XV
9.3.2 Cases for development stage modeling | 217

9.4 Nonlinear process model optimization:
contributions to a feasibility stage | 219
9.4.1 Nonlinear modeling & optimization for a feasibility stage | 219
9.4.1.1 Overview of common elements in an optimization frame | 220
9.4.1.2 Mixed integer nonlinear programming format | 221
9.4.1.3 Flow of information in the computational process | 221
9.4.2 Cases for feasibility stage modeling | 224
9.5 Concluding remarks | 227
10 Evaluating economic performance | 231

10.2 Economic project evaluation | 231
10.2.1 Project cash flow | 231
10.2.2 Economic potential project evaluation method | 235
10.2.3 Simple project evaluation methods | 235
10.2.4 Present value project evaluation methods | 236
10.3 Manufacturing costs | 237
10.3.1 Direct costs | 237
10.3.2 Indirect and general costs | 240
10.3.3 Cost sheet | 241
10.4 Estimation of capital costs | 242
10.4.1 Capital cost components | 242
10.4.2 Evolution and purpose of capital cost estimates | 246
10.4.3 Order of magnitude estimates | 247
10.4.3.1 Existing plant data | 247
10.4.3.2 Step counting methods | 250
10.4.4 Study and preliminary estimates (factorial methods) | 252
10.4.4.1 Estimation of purchased equipment cost | 253
10.4.4.2 Overall installation (Lang) factors | 254
10.4.4.3 Individual factors | 255
11 Evaluating for safety and health | 259

11.2 Risk reduction during design | 259
11.3 Inherent safer design principle | 260
11.4 Identification studies | 261
11.5 FMEA | 270
11.6 Dow index methods | 272
11.7 Bowtie assessments | 272
11.8 Consequences assessment | 274
11.9 Toxic exposure | 286
XVI | Contents
11.10 Codes, standards and designing ALARP | 288

11.11 Risk analysis | 290
12 Evaluating for environmental, social and sustainable development

aspects | 298
12.1 Environmental evaluations | 298
12.1.1 Introduction to environmental evaluations | 298
12.1.2 Environmental evaluation: discovery stage | 298
12.1.3 Environmental evaluation: concept stage | 299
12.1.4 Environmental evaluation: feasibility and development
stages | 303
12.1.5 Environmental evaluation: detailed design stage | 304
12.2 Evaluating for social acceptance | 304
12.2.1 Importance of social acceptance | 304
12.2.2 Social acceptance aspects | 305
12.3 Sustainable development assessment | 307
12.3.1 Sustainable development history and consensus | 307
13 Communicating | 311
13.1 Communicating: project team and stakeholders | 311
13.2 Communicating using the Delft Design Map (DDM) | 311
13.3 Activity reports | 313
13.4 Meetings: agenda and minutes of meeting (MOM) | 314
13.5 Models | 315
13.6 Presentations | 315
13.6.1 Quality checks: FOOFI list for presentations | 316
13.7 Reporting in stage gate reviews | 316
13.7.1 Concept stage | 316
13.7.2 Feasibility and development stages | 320
13.7.3 Quality checks: FOOFI list for reports | 324
Part D: Education
14 Education | 329
14.1 (Bio)chemical design education: a long history | 329
14.2 Education programs | 329
14.2.1 BSc programs: TU Delft | 331
14.2.2 MSc programs: TU Delft | 332
14.2.3 PhD programs: TU Delft | 333
14.2.4 PDEng programs: TU Delft | 334
Contents | XVII
14.3 Design oriented courses at TU Delft’s BSc, MSc, PhD and PDEng
level | 338
14.3.1 BSc molecular science and technology (MST) and BSc life science
and technology (LST) | 338
14.3.1.1 Chemical product design (4052TLEOY, 6 ECTS) | 338
14.3.2 MSc chemical engineering | 339
14.3.2.1 Product and process design (CH3804, 5 ECTS) | 339
14.3.2.2 Conceptual design project (CH3843, 12 ECTS) | 339
14.3.3 Process and product design PDEng courses | 340
14.3.3.1 Advanced principles of product and process design (ST6064, 6
ECTS) | 340
14.3.3.2 Process simulation laboratory (ASPEN Plus® ) (ST6063A, 2
ECTS) | 340
14.3.3.3 Advanced process energy analysis and optimization (ST6101, 2
ECTS) | 341
14.3.3.4 Technology management, economical evaluation in the process
industry (ST6612, 6 ECTS) | 341
14.3.3.5 (Personal and) project management (ST6111, 2 ECTS) | 342
14.3.3.6 Sustainable design of processes, products and systems PDEng
course (ST6792, 4 ECTS) | 343
14.3.3.7 Group design project (GDP) (ST6802, 21 ECTS, ST6814, 17 ECTS, or
ST6815, 17 ECTS) | 344
14.3.3.8 Loss prevention in process design (ST6042, 5 ECTS) | 345
14.3.3.9 Individual design project (IDP) (ST6902, or ST6903, or ST6904, 60
ECTS) | 346
14.3.4 OSPT courses (PhD, PDEng and participants from
industry) | 348
14.3.4.1 Chemical product centric sustainable process design (PhD/PDEng
course) | 348
14.3.4.2 Other OSPT courses | 350
A3 Appendix to Chapter 3 | 352
A9 Appendix to Chapter 9:
Cases of process modeling for simulation and optimization | 361
A9.1 Linear model applications for concept stage | 361
A9.1.1 Nonlinear models and process simulations for development
stage | 389
A9.1.2 Nonlinear models and process optimization for feasibility
stage | 395
XVIII | Contents
A13 Appendix to Chapter 13: Communicating | 411

A13.1 Activity report, example | 411
A13.2 Stream specification (passing battery limit) | 412
A13.3 Concepts/criteria matrix | 412
A13.4 Pure components properties table | 413
A13.5 Agenda and minutes of meeting (MOM) templates | 413
A13.6 Frequently Occurring Opportunities For Improvement (FOOFI) list for
design project presentations | 415
A13.7 Equipment summary and specification sheets | 419
A13.8 FOOFI (Frequently Occurring Opportunities For Improvement) list
for design project reports | 423
Index | 429
Authors’ biographies
ir. Jan Harmsen
Harmsen Consultancy BV
Hoofdweg Zuid 18, 2912 ED,
Nieuwerkerk aan de IJssel, The Nether-
lands
Email: [email protected]
Jan Harmsen is currently a consultant for

sustainable process innovation. He pro-
vides advice to and teaches courses for
industry and academia. After graduating
in chemical technology from Twente Uni-
versity in 1977, he joined Shell. There he
held professional positions in process re-
search, notably on three-phase reactors
and biotechnology, in process develop-
ment, in reaction engineering, process
concept design, process implementation,
and finally in process intensification un-
til 2010. He became, part-time, Hoogew-
erff Professor of Sustainable Chemical Technology in 1997, first at Delft University of
Technology and later at Groningen University until 2013. He authored the following
books: Jan Harmsen, Industrial Process Scale-up: A practical innovation guide from
idea to commercial implementation (Elsevier, 2013); Gerald Jonker and Jan Harmsen,
Engineering for sustainability: A practical guide for sustainable design (Elsevier, 2012);
Jan Harmsen, Joseph Powell (eds.), Sustainable Development in the Process Industries:
Cases and Impact (John Wiley & Sons, 2010).
XX | Authors’ biographies
prof. dr. ir. André B. de Haan
Delft University of Technology (TU Delft)

Faculty of Applied Sciences,
Van der Maasweg 9, 2629 HZ Delft,
The Netherlands
Prof. André de Haan currently combines

a position as principal technologist at the
Cosun Innovation Centre with a position
as part-time professor at Delft University
of Technology. After finishing his PhD,
he held various positions at DSM (1991–
1999), held the chair of separation tech-
nology at the University of Twente (1999–
2006) and the chair in process systems
engineering at the Technical University
of Eindhoven (2006–2010). He worked for
Corbion–Purac as a corporate scientist in
process technology (2010–2016), initially in combination with a part-time position at
the Technical University of Eindhoven (2010–2013), which he exchanged for a post at
Delft University of Technology in 2014. Since 2002 he has been editor for the journal,
Separation and Purification Technology. He authored the books Industrial Separation
Processes and Process Technology, both published by De Gruyter.
Authors’ biographies | XXI
ir. Pieter L. J. Swinkels

Faculty of Applied Sciences,
The Netherlands
Photo by: H. (Harm) P. W. L. Swinkels
ir. Pieter Swinkels is director of the post-

MSc PDEng designer programs in chem-
ical and biochemical product and pro-
cess design at Delft University of Tech-
nology (TU Delft). After obtaining his
ir. (MSc) degree at Technical University
Eindhoven, with specialization in biopro-
cess engineering at TU Delft, he was em-
ployed by Unilever. He was a member of
the R&D team that invented novel deter-
gent powder formulations and manufac-
turing processes (1986–1991). Within Unilever’s specialty chemicals businesses (later
part of National Starch & Chemical Company and ICI) he worked in various positions
in manufacturing, product/process design & development, and as divisional process
development manager (1992–2001). In 2001, he moved to TU Delft as Asst. Professor of
product and process design and engineering, and became PDEng program director. He
develops chemical and biochemical product/process design methodologies, which he
teaches and applies in MSc and PDEng design projects in cooperation with companies
from his extensive industrial network.
XXII | Authors’ biographies
prof. em. ir. Johan Grievink

Faculty Applied Sciences,
The Netherlands
Johan Grievink is a chemical engineering

Professor Emeritus at Delft University of
Technology. He was educated as a chem-
ical engineer and applied mathemati-
cian at Twente University, after which
he joined Shell Research, Amsterdam in
1971. There, he contributed in various po-
sitions to the rapidly evolving field of
mathematical modeling and computing
for development, design, and operations
of chemical processes. In 1992 he trans-
ferred to TU Delft, to teach process model-
ing, design, and optimization of operations as well as to develop educational policies
for MSc and PDEng programs. The focus of his research was on processes for conver-
sion of energy carriers (e.g., natural gas-to-liquid fuels) and for industrial crystalliza-
tion. His research gradually broadened from process engineering to the interactions
between product and process engineering. Having retired in 2008, he continues to
perform some academic duties and to offer industrial consultancy in process model-
ing and optimization.
Authors’ biographies | XXIII
ir. Ido E. Pat-El, MSc.
Royal HaskoningDHV
Industry & Buildings
Contactweg 47, 1014 AN Amsterdam,
The Netherlands
ir. Ido Pat-El graduated from the Delft

University of Technology in 2007 with
a Masters in chemical engineering. Dur-
ing his college years he performed safety-
related projects for organizations such
as TNO Defence Security and Safety, the
Dutch State Supervision of Mines, and
the Center of Industrial Fire Safety of the
Rotterdam–Rijnmond Port District Fire
Services. He is currently employed as
a process safety specialist at the engi-
neering and consulting company, Royal
HaskoningDHV The Netherlands B.V. In
that role he performs health, safety and environmental assessments for the oil, gas
and chemical industry. He has been involved as Lead Technical Safety Engineer in the
design of various offshore oil and gas facilities. Ido has written papers about his field
of work and has given presentations on process safety related subjects in national and
international forums. He provides technical safety guest lectures at the Delft Univer-
sity of Technology and the Eindhoven University of Technology.
XXIV | Authors’ biographies
dr. ir. Jos L. B. van Reisen, PDEng
CB&I
Engineering & Construction
Prinses Beatrixlaan 35,
2595 AK The Hague, The Netherlands
Jos van Reisen holds the position of se-

nior process engineer at CB&I, engineer-
ing & construction group, located in The
Hague. As guest lecturer, he teaches en-
ergy efficient process design at Delft Uni-
versity of Technology since 1993. He has
a MSc (ir) degree and a professional doc-
torate in engineering (PDEng) in chemi-
cal engineering from Delft University of
Technology. He has also received a PhD
in chemical engineering from Delft Uni-
versity of Technology for his research on
a design method for the application of
multistream heat exchangers in the energy saving retrofit of chemical processes. He
worked for ABB Lummus Global from 1998 until its merger with CB&I and worked on
a wide range of projects from basic to detailed engineering and construction supervi-
sion for clients throughout the energy chain both up- and downstream.
|
Part A: Innovation and industry
1 Goal, scope, and structure
1.1 Goal
The world requires novel products and their processes to increase human health, to
reduce the burden on the environment, and to increase prosperity, particularly in
deprived areas, as defined in the United Nations Global Grand Challenges and sub-
scribed to in a declaration by the presidents of the World Congress Chemical Engineer-
ing (WCCE), the European Federation of Chemical Engineers, the European Society of
Bioengineering, the IACChE, and the APCChe, at the WCCE10 in Barcelona, during
October 1–5, 2017 [1]. These combined requirements call for the radical innovation of
products and their processes.
These needs are necessary for innovation but are not sufficient. The need for in-
novation has to also be identified by industrial companies, and then projects must be
started to generate the novel products and technologies and to successfully implement
them in the market.
Moreover, the innovation projects must be effective and efficient. Effective means
that a need is fulfilled. Efficiency means that the novel product and/or technology
require minimal resources and that the innovation project is carried out with minimal
effort, time, and cost. Even more importantly, the novel technology developed in the
company must be able to be competitive with existing technologies in the markets.
These innovation projects can only achieve all these goals by having trained and
experienced people skilled to execute innovation projects. Trained people are re-
cruited from universities and have completed the appropriate programs and courses.
The focus of this book is, therefore, two-fold. It is written as a textbook for design
courses and as a guideline for industrial innovators to generate better products and
processes.
To be precise, the authors had the following goals and readers in mind:
– To educate student engineers for their future careers in designing new products,
processes, and systems that address the needs of people, save the environment,
and create prosperity locally and globally
– To inspire educators with novel methods in educating students in the art of design
– To inspire industrial researchers to invent, design, and develop novel product and
processes
– To support start-up entrepreneurs in organizing their endeavors with the right mix
of people in terms of education, experience, and behavior, through cooperation
with companies and institutes in open innovation projects
– To assist managers of research and development departments in getting a good
mix of people with specific skills and providing them with innovation methods
that have been proven to work
https://doi.org/10.1515/9783110467741-001
4 | 1 Goal, scope, and structure
– To help companies organize and communicate an integrated approach to R&D,

design, manufacturing, and marketing of novel products and processes for suc-
cessful launches
– To inform organizations who want to enter the market of process-product indus-
tries how these industries are structured and how their innovation is organized
This last point may need some clarification. The authors have noticed that what is
going on in this set of industries is not known nor understood by people outside this
set of industries. Even for people working in these industries but not directly involved
in the actual manufacturing process, it is still hard to understand how they function.
There are several reasons for this lack of understanding:
(1) The first reason is that most people have never been in contact with these indus-
tries, because the manufacturing sites are at locations remote from cities and of-
fice buildings.
(2) A second reason is that these industries relate primary resources via long supply
chains of intermediate products to the final consumer. People only see that final
product, but not the intermediate products and their processes.
(3) A third reason is that these industries are very complex. They have many system
levels.
(4) A fourth reason is that there are many industry branches related to processes and
products. Often people within a certain branch are not aware of the existence of
most other branches.
For companies wanting to enter this set of industries, for companies wanting to change
their supply chain, and for people wanting to work in these industries, this book
should also be useful. The focus of the book is, therefore, on methods, guidelines,
and heuristics for concurrent novel products, novel process design, and the develop-
ment process that takes an idea to the market in which business case development
and experimental data generation are systematically integrated.
Because major and decisive effects on final products and processes occur in the
concept stages, the most emphasis is on those stages. However, the other stages to
commercial implementation are also treated.
The content of this book is partially derived from the tacit knowledge of the au-
thors, shaped by experiences in design and innovation. This tacit knowledge has to
the be transformed into coded knowledge using published structures on design steps
and innovation stages and based on findings and categories in design knowledge as
explained in Chapter 4.
The content of this book is validated as much as possible, where available sci-
entific literature is used to back up the knowledge of the authors. In many parts, the
information is made plausible by engineering reasoning. In several parts, the content
is based on one or a few cases. For the education parts, the validation is that students
using the theory were able to generate novel and practical designs.
1.2 Scope | 5
1.2 Scope
The scope of this book provides general knowledge on generating novel products and
their processes through a combination of designing and experimentation. The content
of the book is limited to generic knowledge and methods of design and innovation.
The book does not provide detailed descriptions of specific process technologies
or specific process steps such as unit operations. For that the reader is referred to other
books, notably to Process Technology by de Haan [2], and for descriptions of specific
processes, to the book by Moulijn [3].
1.2.1 Products classification and description
The class of products treated in this book is homogeneous down to the micron scale
and to molecular scale. Typical products belonging to this class are: transport fuels
(petrol, diesel, kerosene) chemicals, plastics, resins, medicines, healthcare products,
detergents, food products, feed products, and ceramics.
A product is called a ‘chemical product’ when it is formed by molecular scale
transformations by means of reactions and, possibly, by creating multiscale physical
structures, such as dispersed or stacked thermodynamic phases. All such products are
manufactured by the chemical process industry and related manufacturing sectors,
such as food, energy, and electronic materials.
A special subclass is a set of products that are homogeneous down to the micron
scale but become heterogeneous below this scale and have two or more phases. These
are called ‘structured products’. An example is mayonnaise, which is a stable disper-
sion of water in an oily phase. Another example is a heterogeneous catalyst. Designing
these types of products requires special knowledge as described in the section on the
design of structured products.
A special class of products are in fact microprocesses that operate on chemical en-
gineering principles (reactions, separations, fluid flow) and have information as the
primary output and not a chemical product or energy as such. The information con-
cerns features of the feed material(s) being processed. Examples of such microprocess
products are integrated labs-on-a-chip where the purpose is rapid screening of many
samples and finding the ones with best performance or generating medical diagnos-
tics from samples of body fluids.
The design of discrete products and devices such as cars do not belong to the core
of this book, but many design and innovation methods provided in this book are also
useful for designing such products and devices.
1.2.2 Description of process industries
Industries that produce homogeneous products are often called process industries.
There are many process industry branches. Here is a list to get an impression: oil
6 | 1 Goal, scope, and structure
and gas (petroleum), chemicals, plastics, rubber, pharmaceuticals, food, beverages,

ceramics, base metals, coal, textiles, coating and paint, paper and pulp, electricity
(power), and (drinking) water.
Most process companies are large and have their own research and development,
design department, manufacturing processes, marketing and sales and a head of-
fice with supporting departments such as legal and public affairs. Nowadays the re-
search and development is combined with the design department into a technology
center so that design works more closely with R&D and can be better integrated into
the whole innovation trajectory. Apart from these manufacturing companies there are
many other companies and institutes relevant to innovation in the field of process in-
dustries. Those are described in detail in Chapter 3.
1.3 Book structure
The structure of the book follows a systematic top down systems approach.
– Part A is about the what, why, and how of industrial innovation. It provides con-
text for any innovation and design by a description of all system levels in na-
ture, down to the smallest molecular scale. It also describes innovation structures,
methods, and collaboration options with other organizations and elaborates on
the innovation stage-gate approach with all major steps included. These stages are
referred to throughout other parts of the book. This section also contains manage-
ment guidelines for governing innovation portfolios and innovation projects.
– Part B is about the why and how of concept design for innovation. This section
focuses on generating novel ideas and concepts. It provides, in a central chapter,
a comprehensive method for concurrent product and process design for each in-
novation stage. This chapter will be particularly useful for experienced designers
and innovators to generate, plan, and manage complex radical innovations. This
section also contains other chapters that systematically guide students and less
experienced designers from design scoping through to design execution. These
chapters focus on discovery and concept stages.
– Part C is about how to optimize designs through modeling, evaluations, and re-
porting.
– Part D is about how to teach design for innovation. It describes established edu-
cation engineering programs and courses at Delft University of Technology.
References and further reading

[1] Negro C, Garcia-Ochoa F, Tanguy P, Ferreira G, Thibault J, Yamamoto S, Gani R. Barcelona Decla-
ration – 10th World Congress of Chemical Engineering 1–5 October 2017, Chemical Engineering
Research and Design, 129, 2018, A1–A2.
[2] De Haan AB. Process Technology An Introduction, Berlin: De Gruyter, 2015.
[3] Moulijn JA, Makkee M, van Diepen AE. Chemical Process Technology, 2nd edn, Hoboken: J. Wi-
ley, 2013.
2 Systems relevant to design for innovation
2.1 Introduction to all relevant system levels
Table 2.1 provides an overview of all system levels relevant to product-process inno-
vations, as derived from Allenby [1] and Hartmann [2]. The wording of each level has
been chosen such that it should be easy to understand for all involved in innovation.
Each system level is connected to the neighboring levels via mass and energy
streams. Changes inside a system level thereby affect all other system levels. Prod-
ucts and their manufacturing processes are, therefore, not islands, but are connected
to larger and smaller systems.
The purpose of this systems approach is to:
– Help innovators to get the big picture of how the object of innovation is connected
to other systems
– Assist designers to include higher and lower system levels in the design scope
– Provide managers a framework to communicate with other parties and stakehold-
ers involved or affected by the innovation. It shows for instance the connections
between industry, society, and nature. This should help to generate innovative
solutions that are attractive to society and have no negative impact on the envi-
ronment.
– Facilitate effective and efficient communications, in particular by using block flow
diagrams to represent the systems
Tab. 2.1: All system levels. Derived from Allenby [1] and Hartmann [2].
System Level Size (m) Description; example
Earth 40 × 106 World system of nature and society

Society 10 × 106 All human activities
Value chain 10 × 106 Cradle to grave chains
Industrial symbiosis network 40 × 103 Industry + societal
Industrial complex 10 × 103 Factories connected at industrial site
Factory 1 × 103 Processes at same location and owner
Process 400 Manufacturing set transforming feed to product
Process step 100 Subprocess to intermediate product
Unit operation 10 Generic rules defined process step; distillation
Main equipment 1 Mechanically defined artefact; pump
Characteristic subprocess 0.1 Defined exact geometric boundaries; tray
Microelement 10−3 Elementary element; catalyst particle
Elementary 10−9 Characteristic principle; chemical reaction
https://doi.org/10.1515/9783110467741-002
8 | 2 Systems relevant to design for innovation
2.2 Earth system level
The earth system level and its main system elements are schematically represented in
Figure 2.1. Mass and energy flows connect all these elements. The whole earth system
is approximately closed in terms of mass.
SUN Space
Atmosphere
biomass Fauna Society
Soil and Water
Fig. 2.1: Earth system: elements, energy and mass flows.
In terms of energy, it is in a dynamic near equilibrium. Sunlight energy, thermody-

namically a high-quality energy input equivalent to a temperature of 6000 K, enters
the system and is radiated back from the earth’s surface (black mass radiation temper-
ature) at 256 K. That radiation is absorbed by gases in the atmosphere, notably water
vapor, carbon dioxide, methane, and other components. These in turn radiate heat
back into space at a temperature of 288 K [3]. The difference between black body ra-
diation from the earth surface and radiation from the earth’s atmosphere of 32 K, is
called global warming.
The system element of biomass stores a tiny fraction of the solar radiation by the
photosynthesis of water and carbon dioxide into carbohydrates (and some other com-
ponents). The system element of fauna converts carbohydrates back to carbon dioxide
and water, while using the released high quality energy for muscle activities.
For the last ten thousand years, up to the industrial revolution in the nineteenth
century, this system level has been dynamically stable with no significant accumula-
tion of materials in any of the system elements [3].
This earth system level is an inspiration for engineers, as it shows that a nearly
steady state on a global scale is obtainable for mass transformations driven by solar
energy. This inspiration is an important part of design for industrial ecology and is
described in Chapter 5.
2.3 Society system level | 9
Presently, due to human activities, represented in Figure 2.2 by the system “soci-
ety,” some system elements are no longer in a dynamic steady state. A major distur-
bance is the burning of fossil carbon fuels, by which carbon dioxide is accumulated
in the atmosphere.
This accumulation increases the absorption of black body radiation, by which
global warming increases. This in turn means that more water evaporates from the
ocean, which condenses elsewhere, causing more wind and more rain and thereby
changing climate behavior in various parts of the earth. These changes are summa-
rized in the term “climate change” [3].
2.3 Society system level
Figure 2.2 shows the society system level and its system elements of culture, politics,
and the technosphere.
The technosphere contains all human activities of creating and using technolo-
gies. This creation and use of technologies all cause mass and energy flows to and
from the earth system.
Cultural and political system elements influence the technosphere via informa-
tion flows and legislation. Major global actors include the United Nations (UN), cli-
mate change conferences, nongovernmental organizations (NGOs), the World Busi-
ness Counsel for Sustainable Development (WBCSD), and universities world-wide.
Examples of these influences are the Paris Agreement on climate change, set up by
the United Nations, through which 195 country governments agreed to take measures
to reach zero net global warming gas emissions by 2050 [4] and the formulation of
Sustainable Development Goals (SDG) by the UN and the WBCSD [5].
These global agreements are turned into policies by governments and by indus-
tries that stimulate innovation to create technologies that are embedded in the earth
Energy
Earth Material
Sun
Atmosphere Society
Culture
Water Politics
Soil
Technosphere
Biosphere
Material
Energy
Fig. 2.2: Earth and society systems: their ele-

ments and connections.
system, and structure society in such a way that they keep these surrounding systems
intact, while creating prosperity.
On lower society levels, stakeholders also interact with and influence the techno-
sphere. Governments provide laws and budgets for research institutes and universities
to generate knowledge and novel technologies. Governments in general also provide
budgets for education, which in turn generate the next generation of skillful people
to develop society.
Chapter 3 provides more information on important parties in society and their
relations to innovation in the technosphere.
2.4 Technosphere system level
The technosphere level is a subsystem of society and is the collective activities of

mankind creating and using technology. The technosphere is connected by material
and energy flow to the earth system.
As an example, take the carbon cycle between the technosphere and nature. Large
amounts of carbon are being extracted from the soil in the form of fossil resources,
coal, crude oil, and natural gas, which are burned and put into the atmosphere in the
form of carbon dioxide. This is a major cause of carbon dioxide concentration increase
in the atmosphere, causing additional global warming.
Through innovation, increasingly renewable sources of energy are being devel-
oped, notably water, wind, solar, and biomass, in such a way that in the end the tech-
nosphere will be connected to nature via mass and energy in a steady state flow, hence,
Biomass Recycle
Farming Base production
End Producer Recycle
Biosphere
Reuse
Consumer User
Collector decomposer
Fig. 2.3: Circular Economy mass flows of technosphere and biosphere. Derived from Ellen MacArthur
Foundation.
2.5 Value chain system level | 11
with no accumulation of carbon in one subsection and depletion in a different subsec-

tion of nature. This is shown in Figure 2.3.
All major stakeholders involved in innovation in the technosphere are described
in Chapter 3.
2.5 Value chain system level
The value chain system level is about the economic value increase of material flows
from primary natural resources transformed into consumer products via industrial
upgrading steps. Value chains are a subset of society and are connected to the tech-
nosphere via mass and energy flows. Value chains are also called supply chains, by
companies further down the value chain.
A major current value chain for the chemical industry is shown in Figure 2.4. It
starts from fossil sources of crude oil or natural gas and ends in consumer products.
A growing supply chain starts from biomass such as sugar beet, corn, or sugar
cane, to glucose and then to food, plastics (polylactic acid), and also to pharmaceuti-
cal products, as indicated in Figure 2.4.
This growing supply chain from biomass shows that in innovation not only fossil
resource based value chains have to be considered but also these renewable value
chains. Through breakthrough innovation, more products are likely to be produced
from renewable resources at prices competing with products from fossil sources.
Active
Raw Basic Building Basic Consumer
ingre-
Materials Feedstock Blocks Products Products
dients
Oil Nafta Ethylene Polyethylene Plastic Bags

Natural Propylene Acrylonitrile Polypropylene Car Parts
Gas Ammonia Caprolactam Nylon 6 Carpet Fiber
Air Toluene Benzaldehyde Phenylglycine Cephalexine
Butane Maleic Glyoxylic Acid Aspartame Candarel
Anhydride Z-ASP
Beet Glucose Penicillin 6-APA Amoxillin
Commodity Chemicals
Fine Chemicals
Specialty Chemicals
More complexity in molecular architecture: higher value
Fig. 2.4: Chemical industry value chain.

Chapter 4 shows how in product and process design driving innovation, novel value
chains are taken into account.
2.6 Industrial symbiosis system level
The industrial symbiosis system level is about connections between process factories
and other society elements such as cities. District heating by industrial waste heat is
an example of industrial symbiosis.
It is also the system level where different engineering disciplines meet. Sustain-
able housing design for example, means that architects and process engineers must
interact to get an integrally optimized housing system where washing, waste reuse,
heating and lighting and the use of sunlight as energy source are to be combined.
Industrial symbiosis is, therefore, a level in which innovation can flourish. A good
example showing the power of industrial symbiosis thinking in saving cost and re-
ducing the environmental impact is the municipal waste water conversion to boiler
steam water for Dow Chemical in Terneuzen, the Netherlands.
For decades, Dow Chemical took salty surface water from the Westerschelde and
converted that into boiler feed water. The most energy intensive and costly process
step was the desalting using ion-exchange resin beds. The central Dow organization
embraced industrial symbiosis as a useful method. A methodology was developed and
local factories were invited to come up with industrial symbiosis solutions.
Dow Chemical Terneuzen picked up the idea by looking around for options and
identifying the waste water stream of city Terneuzen as an opportunity, since this wa-
ter contains far less salt than the surface water. Through cooperation between the city
of Terneuzen, the Zeeuws-Vlaanderen water board, Elides, and Dow Chemical, the re-
gional infrastructure was adapted by making a 6-km long pipeline from the municipal
waste water treatment plant to the Evides pumping station. The Reverse Osmosis (RO)
plant for treating brackish estuary water was partly changed to biologically-treated
waste water. The implemented solution meant the reuse of 2.5 million km3 of munici-
pal waste water; rather than treating this as an effluent, the waste water replaced the
surface water, causing 65% in energy savings, and a reduction of chemical intake for
cleaning the RO membranes. In 2007 the novel solution obtained the EU Responsible
Care Award [6]. This new industrial symbiosis system is illustrated in Figure 2.5.
This industrial symbiosis case demonstrates the power of thinking in more than
one system level. Only by going to a higher system level than the industrial complex
could the solution be found. It is also an example of social innovation. Chapter 3 also
describes various other types of innovation. Chapter 5 shows a design approach based
on industrial symbiosis.
2.8 Factory system level | 13
Water
Desalting
treatment
City of Terneuzen Water company Evides Dow Chemical
Fig. 2.5: Turning municipal waste water into industrial boiler feed water: industrial symbiosis by the
city of Terneuzen and Dow Chemical – winner of the European Responsible Care Award [6].
2.7 Industrial complex system level
An industrial complex contains many factories connected by networks of many dif-

ferent mass and energy flows. Often certain shared utilities such as steam generation
and waste water treatment are common. In Germany, these complexes are often called
Verbund.
Designing a process to be incorporated in an existing industrial complex, is called
designing for a brown field, while designing a process as a standalone with no com-
mon utilities is called designing for a green field. Designing for a brown field situation
has many advantages as the major facilities such as utilities and infrastructures for all
kinds of energy such as electricity, steam, base chemicals, and waste water treatment
are available.
Designing a process for a so-called green field situation, so for a location with
nothing there, means that the infrastructure must also be designed. This means that
a concept for the whole complex has to be generated. This on the other hand also pro-
vides many opportunities for designing a very efficient and environmentally friendly
complex, where industrial symbiosis level design thinking is applied.
2.8 Factory system level
A factory consists of one or more manufacturing processes of the same owner at the
same location. It also includes storage facilities for feeds and products. In most cases
it is a subsystem of an industrial complex. It converts input materials into higher value
output products. Other outputs such as byproducts and energy may be sent to other
parts of the industrial complex.
In designing a factory, it is always beneficial to take the higher system levels into
account. Often facilities can be shared and byproducts used in other factories of the
industrial complex.
2.9 Process system level
A process is part of a factory. It is a set of process steps to transform input materials

into a market product. Some companies call this a (chemical) plant. Often the process
(or plant) is named after the product it manufactures.
Process design is one of the core topics dealt with in this book; notably in Chap-
ter 4–13.
2.10 Product system level
The term ‘product’ means in general something that is manufactured and sold to cus-
tomers. If it is sold to industrial customers, it is often called an intermediate product.
If it is sold to consumers, it is called a consumer product. This book focuses on ho-
mogeneous products (see Chapter 1). These are for instance intermediate chemicals,
such as solvents and olefins, and consumer products such as food products, deter-
gents, and plastics. Chapter 1 provides examples of product types from various indus-
try branches.
The majority of these products have a special function in consumer use. Most of
these products react chemically or physically when used. Examples are: food prod-
ucts that react inside humans and animals to other components that are used to grow
and maintain a healthy body, as well as detergents that interact with dirt and clothing
during washing, such that the dirt is separated from the clothing.
Because of these special functional purposes novel product design is a complex
act, as the behavior of the client with the novel product and the product behavior
have to be considered. The product design is even more complex, as design of the
manufacturing process belonging to the product also has to be taken into account, and
the higher system levels, such as the supply chain, have to be considered. Chapter 4
provides an elaborate method for doing so.
2.11 Process step system level
The process step system level needs some explanation. It is a subsystem of a process.
It is a cluster of process unit operations that perform a desired action. Examples are
feedstock pretreatment, chemical conversion, separation, product purification, and
product packaging.
This system level is very useful when designing a large novel process or an indus-
trial complex with several large processes. It keeps the representation of the whole
design simple with a limited number of block flow diagrams, in which each block is
a process step. Connection changes between process steps, for instance, are easily
made. Communicating the whole design is then also easy.
2.14 Characteristic process subsystem level | 15
2.12 Unit operation system level
A unit operation is a subprocess step characterized by the same physical and chemical
rules as for any application. Thus, the same unit operation can be applied to many
different applications using the same rules. This system level was defined by Arthur
D. Little in 1916 [7]. Its definition resulted in enormously more efficient process designs
because it allowed for a) educating students about unit operations applicable to any
process design, b) improving the generic rules through research, and c) improving
actual process designs by applying rules (prior to this unit operation invention, only
field-specific technologies existed such as sugar technology, coal technology, crude
oil technology, etc. Knowledge generated was only applicable for a particular field).
Process design based on unit operations still dominates the design method of
chemical processes. New unit operations such as membrane separation appear and
find their way into many applications. Design based on function identification and
function integration is a novel process design method that facilitates breakthrough
process designs. Both ways of process design are treated in this book, in particular in
Chapter 4–10.
2.13 Main equipment system level
Main equipment denotes all individual major apparatuses required for the processes.
A unit operation distillation for instance will consist of the following main pieces of
equipment: a column with internals, a top condenser with a heat exchanger, a top
vessel to separate vapor from liquid and to act as a holding tank, a bottom boiler heat
exchanger, and a pump to feed liquid reflux back to the column. Disciplines involved
are mechanical engineering, electrical engineering, and materials engineering.
2.14 Characteristic process subsystem level
A characteristic process subsystem is a subsection of a unit operation or main equip-

ment. It is characterized by a few well described physical and/or chemical phenomena
and clear boundary conditions. A characteristic process subsystem, with well-defined
geometric boundaries, is a tray in a distillation column, or a static mixer element in
a static mixer. In general it is a very useful entity in design, because it allows for the
dedicated modeling of the subunit and connects it to all subunits. This type of mod-
eling is then often standardized as modules. The modules of the process subsystems
are then connected by simple energy and mass flows. Computer modeling packages
for these process subsystems and their connections that form integral models are for
sale.
2.15 Microelement system level
A microelement, such as a catalyst particle or a droplet in a liquid extraction unit, is

governed by boundary conditions (such as interphase equilibrium), basic heat trans-
port, and mass transport rules. The innovation on this level can involve catalysis, mi-
crobiology, interface chemistry, particle technology, and chemical engineering reac-
tion engineering.
The conditions at the interface are part of microelement modeling. In this way
the microelement can be studied separately from the hydrodynamics of the fluids sur-
rounding the microelement. It can, therefore, be studied prior to the selection of the
process equipment using dedicated laboratory equipment with controlled uniform
conditions outside the microelement.
2.16 Elementary system level
The elementary system level is uniform over its scale and governed by basic physical
and chemical properties and chemical reactions. All these properties and reactions
are independent of the equipment type and scale, and hence can be studied sepa-
rately from scale dependent phenomena, by experts in chemistry, kinetics, and phys-
ical properties.
[1] Allenby B. Industrial Ecology: Policy Framework and Implementation, Saddle River New Jersey:
Prentice Hall Inc., 1999.
[2] Hartmann K. Analysis and Synthesis of Chemical Process Systems, Hoboken: Elsevier, 2013.
[3] Boeker E, an Grondelle R. Environmental physics, 2nd edn, Chicester: J. Wiley, 1995.
[4] UN. Paris Agreement, United Nations Treaty Collection, 8 July 2016, Sourced from: https:
//treaties.un.org/pages/ViewDetails.aspx?src=TREATY&mtdsg_no=XXVII-7-d&chapter=27&
clang=_en. Last access date: 10 January 2018.
[5] UN. Transforming our world: the 2030 Agenda for Sustainable Development, Resolution adopted
by the General Assembly on 25 September 2015, accessed on 5 May 2017. Sourced from: http:
//www.un.org/sustainabledevelopment/sustainable-development-goals/. Last access date: 10
January, 2018.
[6] Wu Q. Industrial Ecosystem principles in industrial symbiosis: by-product synergy, in Harmsen J,
Powel J (eds), Sustainable Development in the process Industries, Cases and Impact, Hoboken:
J. Wiley, 2010.
[7] Little AD, et al. Chemistry in history, Chemical Heritage Foundation, Retrieved 13 November
2013. Sourced from https://en.wikipedia.org/w/index.php?title=Unit_operation&oldid=
709344969. Last access date: 15 January 2018.
[8] De Haan AB. Process Technology: An Introduction, Berlin: Walter de Gruyter GmbH, 2015.
[9] Moulijn J, et al. Chemical Process Technology, 2nd edn, Hoboken: J. Wiley, 2013.
3 Managing innovation
3.1 Overview
3.1.1 Innovation terms
The word innovation in this book is used in three ways. First, it means the success-
ful introduction of a new product or process into the market. Second, it means having
management methods and systems in place to always have a consistent portfolio of in-
novation projects in progress. This is called innovation portfolio management. Third,
it means the management of individual innovation projects into successful market in-
troductions. This is called innovation project management. The overall meaning of
this chapter is to show that innovation pays off and is even more beneficial to compa-
nies when managed by proven portfolio and project methods.
The historic business trends in product and process innovations leading to con-
current product-process innovation, including the supply chain and the product mar-
ket, are shown in Section 3.2. This section also provides business motives for innova-
tion.
Various types of innovation are described in Section 3.3. Innovation partners of
industrial innovation are described in Section 3.4. Portfolio management of innova-
tion inside enterprises is described in Section 3.5. The remaining sections all deal with
project management in the innovation stages. An overview of stages and key activities
is shown in the next section.
This chapter treats the management of innovation in a limited way. Far more in-
formation on innovation management is found in dedicated textbooks such as those
written by Goffin [1] and Cooper [2]
3.1.2 Stage-gate approach
Innovation project management, as described in this book, is based on a stage-gate

approach. Table 3.1 shows the names of the stages and some stage characteristics. The
same stages may have different names in companies, but the activities during each
stage will be nearly the same and the sequence of stages will be similar. Gates are
placed in between the stages. At these gates, the decision is taken to stop or to continue
the project.
This stage-gate approach ensures that with minimum effort in the early stages,
only the most critical aspects are investigated so that a project can be stopped timely.
A full rationale of the stage-gate method is provided in Section 3.6. Sections 3.7–3.12
provide descriptions of the stages.
https://doi.org/10.1515/9783110467741-003
18 | 3 Managing innovation
Tab. 3.1: Overview of Innovation Stages.
Innovation Stages
Discovery Concept Feasibility Develop- Detailed Implementation
ment engineering
Purpose Idea Screening Concept Validation Engineering, Product launch,

genera- feasibil- procurement, process
tion ity construction commissioning,
and start up
Invested % 0.01 0.1 2 10 94 100
Stage-gate Strategy Business Business Business Top mgmt.
decision fit case case case
Duration 0.3–1 1–2 1 1–5 1–2 0.1–3
(years)
Design 1 10 50 90 100
scope %
Experimental Proof of Proof of Product Pilot – Start up
principle concept Prototype plant
3.1.2.1 Short description of stage characteristics

The discovery stage comprises the generation of ideas. In some companies this is also
called the ideation stage. Very little design is carried out. The novel product and the
process will be sketched. The investment in a project will be a very small percentage
of the overall investment. The main critical question to pass the gate is: Does the idea
fit the business strategy? Successful ideas that pass the gate to the concept stage will
be followed up by design and experimentation to such a level that a crude estimate of
the commercial scale investment cost can be made. In the feasibility stage a commer-
cial scale design will be made by process engineers, in which the scope of the design
will be 50% of the total scope. This type of design is called front-end loading 1 (FEL-1)
in America and Asia and may be called “50% concept design” or “50% concept engi-
neering” or, in Europe, “50% base engineering.”
The investment estimate will be based on the design inside the scope and on an
estimate of the elements outside the design scope using correlations. The design and
cost of a downscaled pilot plant will also be part of the feasibility stage. If the business
case is attractive enough, a decision to go into the development stage will be taken.
In the development stage new product testing and pilot plant testing will be car-
ried out, followed by a commercial scale design based on the product testing and the
pilot plant testing. The design scope will now cover 90% of all items of the commercial
scale plant. The investment cost estimate has thereby an uncertainty of maximally 10
percent, but is likely to be only a few percent. This is called front-end loading 3 (FEL-3).
For very large investments a design with a scope of 80% definition will be made, as
this requires less design cost to be made and is still accurate enough to take a decision
to continue or not). This is called front-end loading 2 (FEL-2).
If the decision is made in favor of the commercial scale, an Engineering Procure-
ment Construction (EPC) contractor will be asked to make a detailed engineering de-
sign, and execute the procurement of equipment and construction of the process.
When the construction is finished, the new plant will be commissioned and will be
launched in start up. The new product will then be launched into the market.
The duration of each stage strongly depends on the project innovation class (see
Section 3.3). Breakthrough innovations of Class 6 can take up to 19 years. Class 0 may
take 1 year. The figures provided are indications only.
3.2 Business focus and motives for innovation
3.2.1 Business focus trends in product and process innovation
The trends in business focus on chemical product and process innovation of the last
100 years are summarized in Table 3.2. The information is from Arora [3].
In the period between 1920–1950, the innovation focus was on novel products
such as plastics. Process design became more systematic by the introduction of the
concept of unit operations. Between 1950–1980, the focus shifted to far larger pro-
cesses to improve efficiency and reduce investment cost per ton of product.
In the next two periods of 1950–1980 and 1980–2015, global competition and con-
cerns about the environment (acid rain, ozone layer depletion, and global climate
change) induced process innovation with process intensification as the major lead-
ing principle. Computer modeling greatly facilitated this innovation [4]. This is also
the reason that Chapter 8 and 9 on product and process modeling are included in this
book.
Tab. 3.2: History of product and process innovations.
Period: 1920–1950 1950–1980 1980–2015

Business Driver Novel products Global scale-up Global competition
Global environment
Design Unit operations Unit operations Function integration
Methods Monographs Phenomena models Integrated models
Simulations
Scale-up Empirical Pilot plants Pilot plant validation
Source Business Driver: Ashish Arora, Implications for energy innovation from
the chemical industry, NBER, 2010.
In all past design methods, product design and process design were carried out sepa-
rately. In some companies, their research and development were also carried out sep-
arately. The product was developed under the assumption that the process would con-
sist of available standard unit operations and that no interaction between product de-
velopment and process development was needed. The textbook Product and Process
Design Principles by Seider, shows how these designs can be made [5].
Recently a simultaneous product and process design method based on simulation
and optimization has been proposed. However, the method requires a complete math-
ematical description of all physical and chemical phenomena of the product (and of
potential processes), so can only be used in limited cases [6].
Almeida and others developed a so-called anticipating reciprocal approach to
product and process design, which facilitates optimum combined design of product
and process. His method is the base for our book as found in Chapter 4, 6, 7, 8, and 9.
The advantages are briefly [6]:
(1) It incorporates a larger solution space than sequential approaches, by which more
optimum combined solutions of product and process design can be found.
(2) It allows for qualitative and quantitative design approaches.
(3) It has been successfully applied in industry.
3.2.2 Business motives for innovation
Companies in general differ in their motivation to innovate. The sections hereafter

describe most encountered innovation drivers.
3.2.2.1 Competition as innovation driver

Arthur D. Little carried out a global innovation benchmark study of many companies,
covering both processes and consumer products. The bench study revealed that in-
novation is regarded by all companies as the major pillar for staying in business. For
most companies, innovation is regarded as the major pillar for sustaining competitive
advantage and the long term increase of company value [7].
Top innovators actively try to open up new business areas and product develop-
ments are significantly important to them. The following statistics of these top inno-
vators indicate their importance [7].
Top innovators:
– Realize up to 2 × higher sales from new products/services
– Have up to 2 × higher earnings from new products/services
– Have half the time for the breaking even of ratio of earning to innovation cost
What happens if companies lag behind in innovation? That is indicated by the statis-
tical results of a large worldwide benchmark study on energy efficiency of chemical
Tab. 3.3: Energy efficiency ratio of processes in various world sectors [9].
World sector Average Energy Efficiency Ratio %

Western Europe 176
North America 207
Rest of the world 224
Asia 230
processes by PDC. They gathered data of different processes in commercial operation

for the same product, and did so for 100 different chemical products. The results are
shown in Table 3.3 [8]. For each product, the most efficient process is set at 100% en-
ergy efficiency. Each energy consumption per ton of product (for the same product) is
calculated and divided by the figure for the most efficient process and expressed as a
percentage called the Energy Efficiency Ratio (EER). The average EER per world sector
has also been calculated and is shown in Table 3.3. It shows that the average EER is a
factor 2 times higher than the best processes. It also means that a lot of processes are
even less energy efficient. These far less efficient processes thus lag in innovation and
will struggle to survive in a competitive world.
3.2.2.2 Learning curve as innovation driver

Learning curves (also called experience curves, or Boston learning curves) are plots
of the market price of a product (vertical axis) plotted against all cumulative sales of
that product (horizontal axis) and taking for both axes the logarithmic value, so taking
a log-log plot. In most cases a down sloping straight line is obtained and a slope is
determined. These learning curves thus quantify the rate at which the market price
decreases with increasing cumulative sales [9].
These experience curves often hold for many decades and become a self-fulfilling
prophecy as companies keep innovating to reduce the cost, knowing that the compe-
tition also does so. A good example of such a learning curve is the case of the pho-
tovoltaic solar cell, as provided by Swanson. It has been determined from 1979 until
2005 and shows a cost reduction of 20% per doubling of installed capacity over that
period [10].
The beginning of a learning curve is when commercial scale production starts and
the product enters the market. The time prior to that beginning can be very long. The
basic principle of a solar cell has been known since the 18th century. The breakthrough
technology for commercial production arrived in the seventies with the silicon photo-
voltaic technology of both the product and its required manufacturing process. That
technology breakthrough resulted in a breakthrough innovation. Commercial scale
implementation was then enormously successful with a rapid increase of installed
solar cells. This success in turn caused a strong drive to reduce the manufacturing
cost. This cost reduction follows a learning experience curve of 20% cost reduction
for every doubling of installed capacity, price expressed in ($/w) from 1979 until the
present, called Swanson’s law [10].
The main implication of innovation for products for which learning curves are
available is that the innovation must aim at the moving future cost target. A book
about learning curves has been written by Jaber [9].
3.2.2.3 Circular economy as innovation driver

Contributing towards a circular economy is for many companies an innovation driver.
It means that products and processes are invented that fit the circular economy con-
cept. Chapter 5 contains a section on the definition of a circular economy.
The World Business Counsel for Sustainable Development (WBCSD) provides the
following five innovative business models for this purpose [11]:
– Circular supplies: meaning that all inputs are biobased or fully recyclable and that
only renewable energy is used for driving the processes.
– Resource recovery: meaning that useful resources are recovered from used mate-
rials, byproducts, and waste.
– Product life extension: meaning that product lifecycles are extended by repairing,
upgrading, and reselling, as well as through innovative product redesign.
– Sharing platform: meaning that product users are connected to one another and
are encouraged to share its use and access or have ownership to increase product
use.
– Product as a service: Meaning moving away from product ownership and offering
customers paid access to products, allowing companies to retain the benefits of
circular resource productivity or ownership to increase product use
The WBCSD also provides three disruptive technologies facilitating circular econ-
omy [11]:
– DIGITAL TECHNOLOGIES such as Internet of Things (IoT), big data, blockchain,
and RFID help companies track resources and monitor utilization and waste ca-
pacity.
– PHYSICAL TECHNOLOGIES such as 3D printing, robotics, energy storage and har-
vesting, modular design technology, and nanotechnology help companies reduce
production and material costs and reduce environmental impact.
– BIOLOGICAL TECHNOLOGIES such as bio-energy, biobased materials, biocatal-
ysis, hydroponics and aeroponics help companies move away from fossil based
energy sources.
3.2.2.4 Sustainable development as innovation driver

The UN general assembly of September 25th, 2015, defined 17 specific measurable tar-
gets, named the Sustainable Development Goals (SDGs), to be achieved by 2030. These
are listed in Chapter 13.
Chakravorti mentions three reasons for adopting the sustainable development

goals of the UN for business [12]. “First, the global goals campaign represents a sig-
nificant new opportunity for companies that view emerging and frontier markets as
their source of long-term growth. According to estimates from McKinsey, consumers
in these markets could be worth $30 trillion by 2025, a significant step up from the
2010 value of $12 trillion. Since 2011, as emerging markets have suffered from slower
growth and fresh social unrest, that $30 trillion prize seems more distant. Acting on
the global goals could help address several of these obstacles that are giving rise to
“trapped value” in the emerging markets.
Second, with the public declarations by many companies to help with the goals,
there is likely to be competitive pressure. Some companies could get a jumpstart in
their industry in organizing partnerships and even positioning themselves as leaders
in sustainable development using the goals as a branding anchor. Being slow to act
could lead to the risk of being left out of these relationships. It then could form a com-
petitive disadvantage from a brand equity perspective.
Third, the goals cannot be realized without business participation. The price tag
for accomplishing these global goals is estimated to be up to $3 trillion a year, for
15 years. For most governments, financing the global goals campaigns will be a stretch;
governments have already committed in the past for similar targets.”
There are special links between the business driver for sustainable development
and value chains. For the environmental impact of products, life cycle assessment is
now common practice. The value chain is a subpart of the life cycle chain, so inevitably
other companies of the value chain must be involved to reduce the environmental im-
pact. For renewables such as feedstock even completely new value chains must be
established.
For the social dimension of sustainable development, partners along the value
chain have to also be involved. This is the case when the early part of the value chain is
in developing countries. A major part of the value should be created in those countries
and negative impacts, such as child labor, must be avoided.
3.2.2.5 Corporate social responsibility as innovation driver

In the business realm, sustainable development is often called Corporate Social Re-
sponsibility (CSR). It is founded on the United Nations description of sustainable de-
velopment with environmental, social, and governance (ESG) dimensions. One motive
for firms to endorse a CSR driven strategy is to obtain capital at a lower cost.
In a large and recent empirical research on firms’ abilities to access finance in
financial markets, Cheng proves that companies with some better CSR performance
face indeed lower capital constraints.
Cheng investigated data from 2,439 public listed firms during the period 2002 to
2008. As a measure for CSR performance he used a panel dataset with environmen-
tal, social, and governance (ESG) performance scores obtained from Thomson Reuters
Tab. 3.4: Corporate social responsibility ASSET4 performance information structure [14].
Economic Environmental Social Corporate governance

Client loyalty Resource reduction Employment quality Board structure
Performance Emission reduction Health & safety Compensation policy
Shareholder loyalty Product innovation Training & development Board functions
ASSET4; a Swiss-based company that specializes in providing objective, relevant, au-

ditable, and systematic ESG information and investment analysis tools to professional
investors, who built their portfolios by integrating ESG (nonfinancial) data into their
traditional investment analysis. It is estimated that investors representing more than
€2.5 trillion assets under management use the ASSET4 data, including prominent in-
vestment houses such as BlackRock [13].
The CSR measurements of Thomson Reuters ASSET4 involve 220 criteria grouped
under four pillars each with main categories as shown in Table 3.4 [14].
Corporate Social Responsibility (CSR) is a process for companies to integrate so-
cial, governance, environmental, and supply chain sustainability into operations
and corporate strategy. This subject is beyond the scope of the book. However, assess-
ing a novel technology on contributions to sustainable development is a necessary
step in incorporating the technology in the CSR strategy of the company.
3.2.2.6 World problems as driver for the product-process industries

Specifically, for the product-process industries the results of the Delft skyline debates
on world problems and innovation technology area are of interest. In the Delft Sky-
line Debates, 75 scientists and engineers from several disciplines, but the majority
Tab. 3.5: World scale problems and innovation areas [15].
Subject Aspects
World scale problems:
Shortage Water, food, energy materials
Pollution Air, water, soil
Human Health Diseases
Innovation areas Food, health, living, transport
Constraints No waste, renewables, solar based
Technology Areas:
Solar cells and electricity storage Small scale, local
Fuel cells + manufacturing Small scale, local
Molecular controlled reactions Personalized medicines, tissues, organs
Membranes Safe affordable water
Process Intensified biomass conversions Small scale, local
from chemistry and chemical engineering, from all over Europe, discussed world scale
problems and innovation areas for solutions [15]. The results are summarized in Ta-
ble 3.5.
The purpose of this table is to stimulate thinking on solving these world scale
problems. Focusing on innovation in these areas for both products and process inno-
vation is likely to provide extra business in the future and will also contribute to the
Sustainable Development Goals.
3.3 Innovation classes and types
3.3.1 Innovation classes
There are many classifications for innovations along similar definitions. Verloop [16]
provides a classification originally made by Gaynor with the terms, breakthrough,
new, and incremental. McKinsey includes (lack of) knowledge of the market as an ad-
ditional classification element [17]. We add as an additional parameter, a new value
chain up front for the product (and process), as a new feed and a new supplier have ad-
ditional risks. Combining all these elements led us to the innovation class definitions
of Table 3.6.
The data of the last three columns of Table 3.6 are obtained from McKinsey [13].
They analyzed 118 chemical company innovation results. Their definition of time to
market is from the formal project initiation, which is likely to be the same as the start
of the development stage, to the time where the cumulative product sales equal the
R&D cost. Their success rate is defined as the portion of projects in an innovation class
Tab. 3.6: Innovation classification for combinations of markets, supply chains, products, and pro-
cesses.
Innovation Market Product Supply Process Time to Success Extra

class chain market % Margin
years %
6 Breakthrough New New New New 8–19 15–20 0–60
radically new
5 Radically new Existing New New New
4 New Existing Modified Modified New 6–15 30–40 0–10
4 New Existing Existing New New
3 New Existing Existing Existing New
3 New Existing Modified Existing Modified
2 Novel Existing Existing Existing Modified 2–5 40–50 0–5
1 Conventional Existing Existing Existing Increased
capacity
0 Conventional Existing Existing Existing Existing
that created positive returns on investment, expressed as net present value, using the
cost of capital with no risk adjustment.
Here are short descriptions of each innovation class.
Innovation class 6
‘Breakthrough radically new’ involves a new market, a new product, a new supply
chain, and a new process. The new market may be nonexistent or new to the company.
For homogeneous new products, a market nearly always exists, so an even higher
innovation class was not defined. With the new product, a radically new product is
meant, which also involves a new supply chain, as new ingredients are needed, and
of course the process is new.
Examples are: Lab-on-a-chip products for medical personal diagnostics and large
household sized batteries for balancing local solar energy production (i.e., Tesla ‘Pow-
erwall’).
Innovation class 5
‘Radically new’ involves an existing market known to the company, but the product,
supply chain, and process is new. So apart from existing market knowledge about the
product, it has the same degree of novelty because of the combination of the new prod-
uct, new supply chain, and new process. The time to market, success rate, and extra
margin will be in between Classes 6 and 4.
Examples are:
– Hydrogen as transportation fuel (next to diesel and gasoline)
– Fischer–Tropsch produced gasoline and diesel, with superior fuel efficiency, and
no sulfur components emissions
– Biobased soft drink bottles with improved barrier properties (PEF)
– Water based paints (solvent-free)
Innovation class 4
‘New’ involves a modified product, a modified supply chain, and a new process. A
modified product means a product that has nearly the same application with a change
in relative amounts of ingredients or one or two new ingredients. In product innova-
tion, this sometimes is called a platform derivative. This class also includes an existing
product for a new supply chain and a new process.
Examples are:
– Bio-ethanol containing gasoline (bio-ethanol process and supply chain is new)
– Biobased polyethylene (monomer sourcing part modified)
– Low-VOC water based paints (improved postreaction or VOC removal processing)
– Ice cream with improved properties (caloric value and sensory properties; flavor
delivery and mouth feel; improved product formulation and structure through
new processing)
Innovation class 3
involves a radically new process for an existing product.
A radically novel process means that novel principles for mass movement, reac-
tion, separation, and heat exchange are applied. These principles are also combined
in novel ways in novel equipment.
A past example is the Fluid Catalytic Cracking Process that converted refinery frac-
tions into gasoline. The process, developed during World War II, is entirely based on
gas-solid fluidization in catalytic cracking and in coke combustion, while the heat ex-
change is carried out by the solids flow. It allowed very large production capacities at
very low investment and processing cost.
At that time fluidization was not practiced in the process industries and for im-
portant elements of the process such as the standpipe between the combustor and
the catalytic cracker riser, precise knowledge of the flow phenomena was not known.
This standpipe has two functions: a) fast, reliable catalyst transport from the com-
bustor to the cracker by gravity and b) a seal for gasoline vapor to prevent flow to the
combustor. Due to the enormous need for gasoline, the oil company (Exxon) took the
risk and constructed the commercial process with enormous commercial scale suc-
cess. The process is still in operation in refineries [18].
A more recent example is reactive distillation. It combines reaction and distilla-
tion in a single column. Eastman Chemical wanted to replace a conventional design
with 11 unit operations with a single column combining extractive stripping and ab-
sorption with reaction. Design knowledge for these combined functions was minimal
and it required several pilot plants before the basic principles were fully understood.
The commercial scale process was an enormous success, with reduced investment and
operation cost by a factor of 5, while increasing the production safety, health, environ-
ment, and reliability [19].
This type of innovation requires an enormous design effort, not only in creativity,
but also in developing new design methods, modeling, and simulation tools.
Innovation class 2
Involves a modified process as only novel aspect. It contains no new processing func-
tions in relation to the product manufacturing, but only fringe novelties, such as a heat
integration with additional heat exchangers; improved process control; optimized op-
erating conditions; alternative utility sources; and retrofitting a heat exchanger net-
work. It requires on average 2–5 years for commercialization [17]. The average success
rate of only 40–50% may come as surprise. Section 3.5.5 provides the reasons for this
low success rate.
Innovation class 1
Involves a process with a higher capacity than previous commercial scale processes
already in operation. Examples are capacity increase by expansion or incremental de-
bottlenecking, or a new plant with the same technology with existing processes for
the same product.
A word of warning is needed here. Often a process is considered conventional
when the applied unit operations are conventional, and the project is placed in this
category. However, if those unit operations have not been applied for the particular
product or streams, that process is new and should be treated as new. It still may need
a pilot plant to remove major uncertainties and reduce the risks to an acceptable level.
Section 3.10 describes this risk of not considering a process as new.
Innovation class 0
Involves a carbon copy of an existing process in commercial operation. It may come
as a surprise that projects in this class still sometimes fail. Section 3.5.5 of this book
reports the reasons for the large percentage of failure even if the project involves no
apparent novelty.
Overall remarks
Not all possible combinations are covered in these classifications. The reader, how-
ever, can easily place his project in an innovation class, by interpolation using the
degree of novelty for the following parameters: market, product, supply chain, and
process.
3.3.2 Innovation by serendipity
The New Oxford Dictionary of English defines serendipity as the occurrence and devel-
opment of events by chance in a satisfactory or beneficial way, understanding chance
as any event that takes place in the absence of any obvious project (randomly or acci-
dentally) that is not relevant to any present need, or in which the cause is unknown.
Innovations presented as examples of serendipity have an important characteris-
tic: they were made by individuals able to “see bridges where others saw holes” and
connect events creatively, based on the perception of a significant link. Walpore [20]
defined the term as: “Serendipity is finding by accident and by sharpness of mind
something not looked for.” So, it combines an accident and cleverness.
A recent example of serendipity is the discovery by my former Shell colleague
Einte Drent, of a novel polyketone polymer. He was trying to synthesize solvents from
synthesis gas. He saw in one of his experiments that white flocks had been formed.
Rather than discarding the result as a failure, he investigated the chemical composi-
tion of the flock and found that it was a polyketone. This novel product was further
developed and appeared to be an engineering polymer with several properties supe-

rior to nylon [21].
Companies who want to be innovative foster a climate and have employees by
which serendipity is stimulated. The results of serendipity innovations will, in the end,
fall in the categories of the previous section.
3.3.3 Social innovations
In some innovations, no new technology is involved and yet they can be called innova-
tions. Some can be called social innovations. These are novel combinations of stake-
holders, such as described in Chapter 2 for system level industrial symbiosis. The case
is cooperation between the city of Vermeulen, the waterboard, and Dow Chemical to
take municipal waste water and use that as feedstock for boiler feed water production.
These social innovations require a lot of communication and discussions, but in-
volve low innovation project costs and low investment costs. The benefits are often
large economically, ecologically, and socially.
3.4 Innovation partners
3.4.1 In-house versus open innovation
Some companies perform all innovation inside their organization. This is called in-
house innovation. Others perform innovation with partners. This is called open inno-
vation. The advantages and disadvantages are summarized in Table 3.7
Tab. 3.7: Advantages and disadvantages of in-house and open innovation.
Parameter In-house innovation Open innovation

Speed Rapid Slow; takes extra contracts and organization
IP Intellectual property easily protected IP protection needs negotiation and attention
Quality Can be limited by narrow view Can be high due to outside expert knowledge
Cost High Low by sharing cost between partners
When Strong in-house knowledge base Breakthrough concept knowledge is outside
When more than one company is involved in an innovation project then that innova-
tion project is called an open innovation project. If several projects with a common
theme are clustered and developed with the same partners then this may be called an
innovation platform. The advantages of these latter types of collaborations are that
the best partners for each knowledge element can be selected and the costs are shared.
The disadvantages are that:
(a) Intellectual property contracts should be made and signed. Often these take a con-
siderable time, before the actual project can start and
(b) Organizing and leading the project is more complex. Each partner has his or her
own agenda and interest. To define a common goal and stay focused on this goal
takes effort and requires a project leader experienced in open innovation.
3.4.2 Innovation partners
Potential innovation partners are depicted in Figure 3.1 together with main informa-
tion flows.
Each potential partner type of an open innovation project is briefly described in
the subsequent sections.
UNIVERSITIES Non-Government Organisations
Contract Research Institutes Knowledge Institutes
Startup innovator
Government
Technology providers
Manufacturer
Equipment Companies
Clients
Engineering Procurement Contractor
Fig. 3.1: Innovation stakeholders and connections.
3.4.3 Universities
Universities produce educated students and research results. University education

programs focused on homogeneous product and process education and research are
first of all technical universities and polytechnics. They have in most cases a chemical
engineering department and/or a biotechnology department and/or a food technol-
ogy department. General universities also have in some cases chemical engineering,
biotechnology, and food processing departments. Mining departments also may have
a section on ore processing.
University education programs are in general for BSc and MSc students. Some uni-
versities also have programs for PDEng students. For PhD research students, univer-
sity departments often also offer education programs. Chapter 14 illustrates the con-
tent options for designing each of these education programs.
University research characteristics include a focus on the new elements of a tech-
nology and not on a complete integral solution, and on describing and publishing
experimental results. The research is in general carried out by PhD students with no
technical experience. The research papers and the PhD thesis in most cases have a
Technology Readiness Level (TRL) of 0 (TRL is defined in Section 3.6). Most academic
research stays at this bottom level.
Because of the enormous number of university research groups worldwide and
the publications in scientific journals, readable by all, a large number of inventions
from universities nevertheless move up in technology readiness levels and become
successful innovations.
Academic research also produces design methods and modeling methods. These
methods facilitate the innovation velocity and facilitate optimized solutions. Mod-
eling methods for reactive distillation and extractive distillation, for instance, have
shortened the development time and have resulted in more optimal designs.
3.4.4 Contract research organizations
Contract Research Organizations (CROs), sometimes called knowledge institutes,

carry out research for a customer. Often these customers are small and medium enter-
prises with a small research and development department, or no such department.
CROs also carry out research and development on their own budget, and then later
when the results are positive and protected by patents and other means, they sell the
results to customers, often via a license agreement.
CROs in general pick up ideas and research results from universities, and help
Small Medium Enterprises (SMEs) to innovate. In this way, they play an important role
in the general innovation level of a country. Because of this role often governments
subsidize these organizations and often a government customer, such as a military,
logistics, and safety department, is a customer of a CRO.
3.4.5 Startup innovators
Startup innovators are companies that start from nothing with an invention. The
leader is a person who strongly believes that their product or technology will change
the world and goes for it. Often the product is ill-defined and the innovator lacks
marketing skills. By trying to sell it she learns the hard way that she must modify her
ways to make a product successful and how to position it in a market.
A spectacular example of a chemical product-process startup is Synthon. It was

started in 1991 by a Dutch chemist Jacques Lemans. In 1993, the first generic product,
dobutamine, a sympathomimetic drug used in the treatment of heart failure and car-
diogenic shock, was launched, and quickly became a commercial success. Synthon is
now a global company with over 1600 employees, producing generic drugs at afford-
able prices [22].
3.4.6 Technology providers
Technology providers provide unique knowledge about a specific technology to their

customers, so that the customer, an end user, or an Engineering-Procurement-Con-
tractor, can complete the design of a product or process with that knowledge. Often
technology providers also have an engineering department so that they can deliver the
complete engineering package to purchase and construct the technology. Their unique
knowledge is often applied to a specific design problem and then tested in their pilot
plant facilities.
3.4.7 Equipment providers
Equipment companies manufacture and sell equipment. The design knowledge about
the equipment is retained for the larger part inside the company. They also have spe-
cialized manufacturing facilities for their equipment. Some equipment companies are
very innovative with new equipment for existing and new problems.
3.4.8 Engineering procurement construction contractors
Engineering procurement construction (EPC) contractors perform in general nearly

all activities for staged detailed engineering. They take the front-end loading design
of the product manufacturer as input. With that and the manufacturer requirements,
they make a detailed design, purchase the equipment and materials, and construct
the process.
Their work force consists of engineers of all disciplines and of a variety of experi-
ence levels. They are often specialized in process design for one industry branch, such
as petrochemicals, food, pharmaceuticals, or electricity production.
For a product manufacturer, it is of paramount importance to select an EPC firm
that has experience in the industry branch. For EPC activities, the motto is: the devil
is in the details. If for instance an EPC is chosen for a food production plant, but has
never had such experience, then it is likely that the process will not clean properly for
decontamination, the most important criterion for a food production facility.
3.4.9 Tolling manufacturers
Tolling manufacturers have processes that can be hired for producing an amount of
product. Tolling manufacturers for instance have several small-scale processes suit-
able for reaction, distillation, and crystallization. Often the process operation range is
wide and the construction material is such that many different media can be applied
without causing strong corrosion. The toller’s knowledge about what can be processed
varies enormously. Some tollers leave the process operation to some extent to the oper-
ators of the hirer. Some tollers have enormous knowledge and experience and provide
not only the process equipment but also their knowledge on how to best operate the
process for the new product.
The advantages of involving a toll manufacturer are that: the product is quickly
produced, no pilot plant has to be designed and constructed, and knowledge of the
toller is obtained for the operation mode.
3.5 Portfolio innovation management
3.5.1 Objectives of portfolio management
Portfolio management is about managing all innovation projects to meet strategic

business objectives. This well-stated definition is obtained from the Project Manage-
ment Institute (PMI) as shown by Bakker [23]. The objective of portfolio management
is to determine the optimal mix and sequence of projects in their innovation stages. It
is, therefore, a dynamic decision-making activity, as some projects will stop at a stage
gate, while others proceed, needing more resources.
There are four major rules for portfolio management [23]:
Rule 1: All projects of a portfolio represent planned investments by the company.
Rule 2: These projects are all aligned with the company’s strategic objectives.
Rule 3: A large portfolio will be divided in clusters of projects of similar characteristics.
Rule 4: All projects are quantifiable so that they can be measured, ranked, and prior-
itized.
Ad Rule 1: Often companies also have small projects that are not part of the portfo-
lio. These loose projects can be damaging to portfolio management, as these are not
discussed in the portfolio management meetings, so are not properly managed. An
investigation by Blichfeldt and Eskerod in 2008 of 128 companies showed that these
small projects together often tie up considerable resources originally allocated to the
portfolio projects [23]. In this way, they damage projects inside the managed portfolio.
This problem can be avoided by having a cluster, called discovery or significant
change, where all small embryonic projects are placed and an annual budget for that
cluster is allocated. In this way, these projects no longer consume budget in an un-
controlled way. Management of such an ideation cluster can be further improved by

having experienced researchers part-time allocated for that management, facilitating
ideas popping up during the year by advice, and providing a budget (within the overall
budget for that cluster).
Ad Rule 2: This rule can only be applied if the company has a clear innovation
strategy.
Ad Rule 3: This clustering on similar characteristics is not easy. Several options are
available, along business division lines, or along technology type lines, but these all
have the danger of being locked in thinking, meaning that novel ideas cannot be con-
sidered. It is better to first have a high-level innovation matrix clustering as described
in the next section. Further clusters underneath each main heading of the innovation
ambition matrix can then be formed. Section 3.5.2 proposes a very practical main clus-
tering, from which further subclustering can be done.
Ad Rule 4: Metrics applicable to the whole portfolio and to individual innovation
projects for ranking and prioritizing are very important. Wrong metrics will cause
wrong outcomes. Metrics for companies can be developed or enriched using the
generic criteria list of Chapter 4.
3.5.2 Ambition matrix for innovation portfolio management
For large established companies, the innovation ambition matrix described by Nagji
can be very useful. It groups all innovation projects of a company into three clusters:
core, adjacent, and transformational [24]. There are similar clusters described in liter-
ature, based on the terms in the innovation classes of Section 3.3, but the advantage
of Nagji’s clustering over others is that it has generic names and clear definitions ap-
plicable to all product and process innovation oriented companies.
3.5.2.1 Core cluster

In this cluster, existing products for existing customers and existing processes are fur-
ther developed. The development of these minor changes of products will often be
carried out in departments supporting existing products and processes, often close to
the manufacturing facilities.
The projects of this cluster fall in innovation Class 3; see Section 3.3. Their time
to market is short. The success chances of introducing these modified products are
high, and the expected internal rate of return sufficient. Management is simple, as it
uses in-house innovation, by experienced employees with known methods for product
development.
A ranking of potential projects and the best selected with a total budget that meets
the set budget can be done according to the following criteria. First, for Safety, Health,
and Environment, clear questions regarding acceptability are to be answered by yes
or no. A no means that project will not be part of the ranking procedure. Ranking of
the remaining potential projects are mainly through Economics (forecasted additional
sales, profit, and return on investment), Technical feasibility and Social (market) ac-
ceptance. The capitalized criteria words in this paragraph form the acronym SHEETS,
described in Chapter 4 in detail.
3.5.2.2 Adjacent cluster

In this cluster, new businesses for adjacent customers and adjacent markets are de-
veloped. Here it is important to recognize that new customers and new markets create
additional risks. These risks can be mitigated by generating new knowledge about
those customers and markets. Often adjustments in products must be made to meet
the requirements of those new customers. This in turn means adjustments to the man-
ufacturing processes.
Projects of this cluster fall into innovation Class 4; see Section 3.3. The time to
market is much longer than for the core cluster. For chemical products, the expected
internal rate of return is 13–18%.
Management of this cluster is critical to success. It is very important for these
projects to go through all the innovation stages starting at the ideation stage, so that
alternative options are generated and evaluated. If the ideation stage is neglected and
a project is jumpstarted in a stage further up in the innovation funnel, then the best
idea may not be pursued. When the new product enters the market, competitors may
already have found a better way of designing and producing it. The ideation stage can,
however, be quickly executed and most of the emphasis and budget will be allotted to
the follow-up stages.
Ranking of projects and selecting the best (in such a way that the total budget
stays within the total set budget for this cluster) is more difficult. Ranking criteria may
be fixed by the company and experienced managers can use their judgment to perform
this ranking. Table 3.8 is used inside a product-process innovation company for this
purpose.
Tab. 3.8: Scoring table for economic attractiveness of novel products.
Criterion Score 4 Score 3 Score 2 Score 1
Market size Very large Large Moderate Limited

Market growth Very strong > 10% Strong > 5% Light 2% None
Sales potential Very large Large Moderate Small
after 10 years
Profitability > 40% > 25% > 10% <10%
after 10 years gross margin gross margin gross margin gr. margin
Tab. 3.9: Scoring table for potential innovation projects using risks criteria.

Entry barrier for us None Overcome < 1 y Overcome < 2 y Overcome > 2 y
Entry barrier for Overcome > 2 y Overcome < 2 y Overcome < 1 y None
competitors
Scale-up financial Easy, can be Moderate; cannot be Hard; cannot Unclear, risks
risks outsourced outsourced be outsourced unknown
Our competitive- Very strong Strong Light No advantage
ness advantage seen
Own IP position Strong for Limited for short time Unlikely Unclear
long time
For each score, the company has to first quantify what they mean by the terms. This
will strongly depend on the company and their branch. Some companies consider 500
million euros very large, others consider 500 billion euros very large.
The scoring is carried out by experienced marketing employees. The total for all
criteria can then be added up, so that each potential project gets a simple economic
dimensionless value.
Attractiveness of opportunity including risks with the following criteria and their
scoring is suggested with Table 3.9. It is used inside a company with innovation in
product-process combinations.
The entry barrier is an important criterion about being legally allowed to produce
and sell the product. For instance, for a novel food product the entrance barrier is food
approval by a government body.
‘Scale-up financial risk’ means the risk involved in investing in production pro-
cesses (for the new product). It has to do with the novelty of the process. If the process
consists of classic unit operations, can be completely defined such that an EPC con-
tractor can design it, and a pilot plant validation is not needed (see Section 3.11), then
a score of 4 is obtained.
The scoring is carried out by experienced managers and the individual score re-
sults as well as the total result are used for evaluating projects for the concept stage
gate to feasibility and for the next stage gate.
3.5.2.3 Transformational cluster

In this cluster, breakthrough products are generated for new (or even not yet existing)
markets targeting new customer needs. New manufacturing processes for these novel
products are also generated.
Projects of this cluster fall into innovation Classes 5 and 6. They have high risks
and high rewards. The risks are reduced by following the stage-gate process.
For this cluster, it is important that sufficient numbers of new ideas are generated
and are nurtured to reach sufficient definition so that the best options can be selected
for follow-up. Because new competences need to be acquired for these novel products
for novel markets, it is also important to form new basic research groups. These can
generate basic design data and knowledge so that the novel products can be embed-
ded in the company. Chapter 6 describes in detail for each innovation stage how basic
design data are generated.
Roadmapping to identify and link basic design data generation, skill develop-
ment, supporting technologies, manufacturing technologies, and the products is very
useful for this cluster. Goffin provides a generic picture of such roadmapping [1]. Chap-
ter 4 on concurrent product-process design will help to plan the design and research
effort, with the Delft Design Map (DDM) and the fingerprint tables for innovation
Classes 5 and 6.
This transformational cluster may also be subdivided into technology platforms
to serve specific (new) business divisions. DSM for instance has three technology plat-
forms: [25]:
– DSM Biomedical: Innovative materials that deliver more advanced clinical pro-
cedures and improved patient outcomes
– DSM Biobased Products & Services: Advanced enzymes and yeast platforms:
enabling advanced bio-energy and biobased chemicals
– DSM Advanced Surfaces: Smart coatings and surface technologies to boost per-
formance in the solar industry
For potential projects in the transformational cluster the strategic fit is the key crite-
rion. Table 3.10 is used in an innovative product-process company and can be of help
in evaluating these ideas and potential projects, notably in the discovery and concept
stages.
Tab. 3.10: Strategic fit score for breakthrough ideas and projects.

Input Existing to us Existing, new to us New, proven far away Emerging
Process Existing to us Existing, new to us Proven elsewhere Emerging
Technology
Product Variants of Existing, new to us New, variant of existing Radically new
existing to us
Market Existing to us Existing, new to us Immature, high entrance None, unknown
barrier, niche product
Clarification of some terms: ‘to us’ means to the company involved. ‘Proven far away’ means proven
in a world continent, where the company has no manufacturing presence.
3.5.2.4 Budget distribution over clusters

The best distribution of innovation budget over the three clusters will depend on the
business strategy of the company. Ngai provides a rule of thumb for the distribution
of budget over the three clusters [24].
Core: 10%,
Adjacent: 20%
Transformational: 70%
The total corporate budget for innovation comprises for most companies the stages
discovery, concept, and feasibility.
The budget for individual projects in the stage development, detailed design, and
startup is decided by the relevant business section or by a corporate decision.
3.5.3 Risk adjusted value for innovation management
Risk adjusted value for innovation management is still not a well-developed field. Aca-
demic research based on industrial practice is sparse. There is a variety of methods
used by companies, but the most used method is judgment by management [23]. Here
are some plausible guidelines for the innovation stages.
For the early stages discovery, concept and feasibility, the total spending is limited
and the only risk is that all money spent is lost, because the projects do not all pass
the stage gate to development. If this total spending can be carried by the company,
the risk is very low. The total budget spent on these early innovation stages can be
related to the typical spending of the industry in the particular branch. Most compa-
nies chose to go for the average value of a branch. If they should catch up with their
competitors they may spend more. If they are ahead of the competition then they may
keep the spending the same as in previous years, in particular when the overall return
of investment on innovation is high.
The temptation to cut innovation portfolio budgets for the early stages is always
there when the going is tough. The immediate effect on business performance is posi-
tive, as cost is reduced. In the longer run, however, market share is lost to competitors,
which then appears to be nearly impossible to get back as innovations take time. The
author experienced this when in the eighties, research for a particular polymer prod-
uct was stopped. Within 10 years market share was lost and could never be recovered.
3.5.3.1 Risk adjusted value in development stage and beyond:

crossing the valley of death
For the development stages, often major investments on prototyping tests and pilot
plant tests are needed. In general, the maximum budget for the whole development
stage of the portfolio is set by top management. The individual projects are first eval-
uated using risks and expected added value. Table 3.10 will be useful in this respect.
Then all projects for the development stage can be presented in the same way and
ranked.
Special attention is paid to the so-called ‘valley of death’. When the projected cu-
mulative profit versus time of an innovation project is made, it will first show an ever
deeper negative value until the lower minimum is reached and then losses will de-
crease as sales start to generate money and finally a net profit will be made. The neg-
ative part of this projected curve is called the valley of death.
The projected profit curve can only be made with some meaning for the first time
at the end of the feasibility stage, when the cost of the commercial scale process, the
prototype, and the pilot plant are known. The deepest point of the profit curve, so
the maximum projected loss, is the most critical part. If that projected loss is larger
than what the company can bear, or is willing to bear, then the project will stop, or be
modified such that the valley of death is less deep.
The deepest point of the profit curve should also be compared to the long run
profit when more than one manufacturing plant will be built and the total sales be-
come large. Chapter 10 and Figure 10.1 provide a detailed description of the projected
profit curve and the valley of death. Further reading on this topic is provided by
Markham [26]. Cooper provides further reading material into innovation portfolio
management [27–29].
3.5.4 Innovation management guidelines for small enterprises
3.5.4.1 Innovation guidelines for small enterprises

The main success factors for innovative small companies and institutes gathered from
numerous literature sources have been summarized by Conquest [30] and are found
in Table 3.11.
3.5.4.2 Guidelines for breakthrough innovation companies

If the innovation concerns breakthrough technology that can alter the company com-
pletely then the intuition of top management plays a very important role. Verganti,
who worked inside very innovative successful companies applying design driven in-
novation found no formal ranking method, but instead found that the only criterion
was: do the CEOs find the idea sufficiently exciting and sustainable. If the CEO asked
for a financial analysis of the value of the design, then that meant that the project
would be stopped [31].
It is also worth considering whether the CEO’s intuition has been built by years of
experience in successful breakthrough innovations. If that is the case then follow-up
is warranted. If, however, the CEO has no experience with successful breakthrough
innovations then failure is likely to happen. Kahneman, the Nobel prize winner de-
Tab. 3.11: Management success factors for innovation in small enterprises [30].
Aspect Description
Leadership
Visionary Top management has innovation vision and shares it
Long term commitment Improvements are judged by contribution to long term
Risk acceptance Top accepts that not all innovations are successful
Strategy
Positioning innovation Improvements and renewal are an important pillar of the strategy.
Signals from environment Our company can pick up signals, transform the strategy accordingly,
and make clear what the impact needs to be in the present business
process structure.
Cooperation in chain Our company shares actively strategic information and business
drivers with chain partners to innovate services, products, and
processes.
Steering
Tasks and responsibilities Innovation tasks and responsibilities are explicitly delegated.
Innovation protection Innovation initiatives are given room to grow.
Rewards Steering and rewards for innovation are explicitly embedded in
steering and reward structure.
Culture
Felt need to innovate Employees’ hearts and minds know the purpose of needing to
innovate.
Learning climate There is an open communication learning climate in which mistakes
and problems are shared.
Tolerance Novel ideas are always appreciated, even when they are contrary to
common wisdom.
Knowledge and competence
Personnel diversity Our company has a large variety of people regarding knowledge and
competencies.
Knowledge sharing It is common to share knowledge of procedures and processes.
Innovation ownership People responsible for innovation have the knowledge and
competence to create a collective atmosphere to sell innovation.
scribes in his book that intuition that has been formed by experience in a regular en-
vironment can on average be trusted [32]. A breakthrough innovative company may
form such a regular environment.
3.5.5 Project failures and their causes
This section deals with project failure rates and their causes. As the causes are also
related to portfolio management the subject is treated here.
3.5.5.1 Innovation failure statistics

We treat here specifically process innovation failures, so failures where only the pro-
cess is new (innovation Class 3) and the product is conventional. The practice of the
last 30 years of introducing novel processes is that more than 50% of those projects are
disasters. Disaster is defined as having more than 30% cost growth beyond the budget
and more than 38% schedule slippage. These statistics have been obtained by Inde-
pendent Project Analysis from over 12,000 projects in oil and gas, petrochemicals, fine
chemicals, and pharmaceutical sectors [23]. This means that the risk of failure is large.
3.5.5.2 Causes of project failures

Here is a list of causes of project failures in the stages up front of detailed engineering
gathered by IPA and published in Table 1.2 of [23].
– Lack of adequate analysis of potential solutions
– Schedule is agenda driven rather than data driven
– Optimistic forecasting
– Overly aggressive appraisal strategy
– Lack of front-end loading (feasibility and development stage deliverables)
– High degree of complexity/innovation
– Unrealistic plan
– Unrealistic budget
– Lack of understanding of stakeholders’ aspirations
In addition to this list an experienced innovation manager of a product-process inno-

vation company observed in his career following additions causes:
– Lack of strategic fit
– Lack of freedom to operate
– Lack of understanding of customer needs and profit
All causes can be avoided by sound project management.
3.5.5.3 Megaproject failures

In addition to causes reported by Bakker for all types of projects, Merrow reports
specifically failures of mega-industrial process projects [33]. He found seven major
causes of failures derived from the data of mega-industrial process projects in the
IPA database. These megaprojects contain also more novel technologies than the
other projects in the IPA database, so the conclusions are very interesting regarding
innovation. Here are the major causes of megaprocess failures.
(1) Too little money spent in innovation stages, often due to greedy lead sponsor of
budget
(2) Too ambitious schedule, not taking into account good industrial practices and
time required
(3) Not a good deal is arranged between the budget provider and the project executor
early in the project
(4) Too little effort spent on the early innovation stages up front of detailed engineer-
ing
(5) Reduce the established budget for the innovation project by 20% by the CEO (lead
sponsor)
(6) Transfer the risk to the Engineering Procurement Construction Contractor by
lump-sum contract
(7) Beating up (threatening to fire) project managers for an overrun-on cost
The first four causes are mainly related to the early innovation stages up front of de-
tailed engineering, where novel technologies are developed.
The last three causes are mainly related to the detailed engineering and imple-
mentation stages.
This again leads to the conclusion that sound project management in each inno-
vation stage is needed to increase the success rate of innovation projects.
3.5.5.4 Project failures in detailed engineering stage

Here is the list of reasons for project failure from IPA for this stage as reported by [23]:
– Lack of clearly defined scope
– Incomplete design and errors
– Wrong contracting strategy
– Wrong contractors
– Lack of adequate resources (skills and numbers)
– Personnel changes
– Personality clashes
– Inadequate risk management
– Lack of adequate change control
3.6 Project management of innovation
3.6.1 Objectives of innovation project management
Project management is about individual projects. It is needed to ensure that ideas re-
sult in a successful novel product and process market implementation. A project in
general is a temporary organization with a defined beginning and end brought to life
to deliver a unique scope within given boundary conditions (schedule, cost, and qual-
ity). Management of these projects is about controlling the execution of the project so
that it reaches its goal and stays within its constraints.
An innovation project is an endeavor in which human, financial, and material

resources are organized in a novel way to undertake a unique scope of work of given
specification, within constraints of cost and time, to achieve beneficial change defined
by quantitative and qualitative objectives [23].
In the early innovation stages of ideation and concept when the number of peo-
ple and disciplines involved is limited to a few, this management is limited to ensuring
alignment to business objectives and to ensure exploring and analyzing many poten-
tial solutions. In these early stages of discovery and concepts, the project manager will
only spend a fraction of his time on this management. Arranging meetings between
project team members to monitor progress and to ensure that the project is focused on
the goal and stays within its constraints is the main task.
In the next stages feasibility and development when more people, more disci-
plines, and more stakeholders are involved, this management becomes a full-time job.
The factors critical to success for this project management are:
– Making sure what the deliverables for each stage gate need to be
– Making sure that sufficient time and effort is spent to produce these deliverables
and that no shortcuts are taken to jump over the stage gates, without the deliver-
ables
3.6.2 Project management by stage-gate system
3.6.2.1 Product and process innovation stages: general aspects

A major method to innovate at the lowest cost and time has been the introduction of
the stages and gates approach. The basic reasoning is that in the early part of inno-
vation, major risks of failure and major risk on nonoptimum concepts, are reduced
quickly and at low cost, by exploring the whole field of options, selecting the best,
and then performing proof of principle experiments for the best option.
At each gate between two stages a decision is taken to continue, or abandon the
project. Some companies also include other options, such as putting the project on
hold, or recycling the project, i.e., sending the project team back to provide more or
better information, or give a conditional go to continue, depending on specific future
events happening [29].
The purpose of the stages and gates method is to avoid spending large sums of
money on wrong ideas for a long time and spending those large sums of money on
good ideas and only in the last stages when the risks have already become smaller,
because more reliable information is available.
Cooper initiated a stage-gate method for product innovation in 1990. Many prod-
uct innovation companies quickly adopted his method. His method has also been val-
idated in the end by investigating many companies, including health care, chemicals,
and polymer companies. Companies that had a formal stages and gates method in
place and those that also had committed gatekeepers did far better than companies
with no formal stages and gates method [28]. A reference book on the stages and gates
method is by Cooper [27].
This stages and gates approach to innovation was applied by many process fo-
cused companies, in particular oil and gas companies even before the product inno-
vation oriented stages of Cooper. They used, however, different names for the stages.
Bakker provides an overview of different names for the same or similar stages [23].
In this book, we provide a uniform nomenclature for all product and process fo-
cused industry branches for each stage’s major activities, as shown in Table 3.12. It is
partly based on the product innovation stage definitions of Cooper and partly based on
process focused companies, and Engineering Procurement Construction (EPC) com-
panies [23]. We condensed these descriptions into a single stage-gate description to fit
the concurrent innovation of products and processes.
Table 3.12 is structured from top to bottom on general personnel management,
business, design and experimentation. This table will be helpful in the overview of
steps from idea to implementation, and the mutual understanding of employees in
the R&D, engineering, and business departments in their use of terms, and in under-
standing their differences in attitude.
Table 3.13 indicates for each stage which disciplines are involved to what extent.
In combination with Table 3.12 and Belbin team roles as described in Appendix A3,
teams can be composed for each stage.
This stage-gate approach to industrial innovation ensures that unfeasible ideas
are quickly discarded with minimal effort and that good ideas are pursued. When com-
bining the stage-gate approach with the systematic design approach of this book then
good ideas are further enriched so that both the company and society greatly benefit
from the novel products and processes.
3.6.2.2 End-of-life stage

We do not show in the overview the end-of-life stage for product and process, as for
that stage no generic design methods are available. In the design of products and pro-
cesses in the earlier stages considerations for end-of-life will have to be considered, so
that product and process recycling are made easy. This foresight aspect of design for
end-of-life is treated in the design chapters of this book and in Life Cycle Assessment
part of Chapter 12.
3.6.3 Project entries to stages by technology readiness level method
Technology transfer from a university, institute, or technology provider to a company,

such as an end-user company is not easy. The readiness level for commercial scale
implementation should be established, so that is clear what still needs to be done.
Tab. 3.12: Key items stages and gates for product-process design driven innovation.
Innovation Stages
Items Discovery Concept Feasibility Develop- Detailed Implementa-
ment engineering tion
General management
Purpose Idea gen- Screening Feasibility Validation Engineering Launch
eration concept Procurement Startup
Construction
Required Optimism Optimism Realism Realism Pessimism Pessimism
attitude
Project Nurture Research Project Project Project Project +
management part-time part-time part-time part-time full-time Operation
Open Low Academic Knowledge T. provider EPC –
innovation Institute toll manuf.
Funding Game Annual Annual Specific Specific Business
changer R&D R&D Project Project
budget budget
Business
Risk of failure % 100 100 50 30 10 10
Invested % 0.01 0.1 2 10 94 100
Capex estimate 200 50 30, 10 0
uncertainty %
Market items Demand Size Size deter- Launch
opportu- estimate mination
nity
Business What Full range Best Everything in What did we
question starting? explored? Selected? place? learn?
Business case Initial Confirmed
Value, Risks
Stage-gate Strategy Business Business Business Top mgmt.
decision fit case case case
Duration (years) 1–2 1–5 1–2 0.5–1 1–2 0.2–0.5
Design
Product design Sketch Concept Prototype Complete Detailed Description
Process design Sketch Concept Unit Op- Main Detailed Startup plan
erations Equipment
Main Sketch Concept Basis data Design Plan, Post
deliverable concept Economic Design, books startup
potential Test results review
Project FEL-1 FEL-2, Procurement
engineering FEL-3 package
Experimental Proof of Proof of Product Pilot plant – Startup
principle concept Prototype
Tab. 3.13: Typical involvement company parts in product and process innovation.
As percentage per stage.
Discovery and Feasibility and Detailed Product Launch and

Concept Development engineering Process startup
Science 90 10 5 5
Engineering 5 75 75 5
Operation 5 5 75
Marketing 5 5 5
Business 5 5 10 10
Technology Readiness Level (TRL) assessment is very useful in this respect. It is about
quickly assessing of a novel technology where it is in the innovation stage. It is partic-
ularly used for novel technologies resulting from academic research or technologies
provided by startup companies. End-users then use the result of that assessment to de-
cide to pursue the technology (partly) in house or to leave it for further development
outside the company.
There are several technology readiness level (TRL) tables available from NASA [34],
ESA [35], DOE [36], and EC [37]. However, these are all defined by government orga-
nizations and do not everywhere use terms familiar to the industry. They also do not
consider the (mass) manufacturing of the technology, but only focus on the prod-
uct technology itself. To fill this gap in addition a Manufacturing Readiness Level
(MRL) table for novel manufacturing technologies (for existing products) has been de-
fined [38]. Recently the European Association of Research Technology Organizations
(EARTO) developed a TRL level description for both novel products and processes [39].
From these tables, we derived a TRL table for novel homogeneous products and for
their novel manufacturing processes using terms familiar to these industries, shown
in Table 3.14. We added the level “0” so that also ideas, which have not reached TRL 1
can be categorized.
The table is particularly useful to assess novel technologies from universities and
from technology providers. It is also being useful in open innovation projects with
various partners to help to establish consensus on the technology status and what
needs to be done for commercialization of various process options [40].
A warning remark must be made here. The TRL table suggests that novel technolo-
gies at TRL 1–4 are outside the industrial realm. Also in industry, ideas are researched
from TRL 1 to 4.
Tab. 3.14: Technology readiness level definitions for novel products and processes.
TRL Level Description Innovation Stage

0 Idea of novel product or process stated Discovery
1 Experimental proof of key novel element of product or process; Discovery
called proof of principle
2 Concept design of product and process Concept
3 Experimental proof of product concept elements and process Concept
elements
4 Experimental validation of integrated product (prototype) and Feasibility
integrated process (lab scale)
5 Industrial professional’s techno-economics assessment product Feasibility
and process
6 Novel product prototype tested by industry Development
Novel process technology demonstrated by industrial pilot plant
7 Novel product launched in market Implement
First commercial scale demonstration of process plant in operation
8 Full launch product to market Implement
Learning points from demo plant incorporated in commercial
process design
9 Commercial process in operation and evaluation report available Implement
3.6.4 Team formation
3.6.4.1 Team member behavior characteristics required over the stages and teams
Team members should be chosen for their skills but also whether they could work
together. The Belbin group formation method is very successful in forming groups that
cooperate well.
3.6.4.2 Group design with Belbin team roles

Project groups are mostly designed based on disciplines required only. However, if
groups must cooperate closely for the desired results then they should have comple-
mentary personal skills as well. In addition to discipline design the Belbin team roles
can be used to design the group or team. Appendix 1 shows how to apply this method.
Desired behavior in different stage

The desired behavior of people in the various innovation stages is also important. Op-
timism should prevail in the early stages of discovery and concept, as most ideas will
not make it to the commercial implementation and the lack of knowledge should not
hamper people to believe in the success of the project. In the feasibility and develop-
ment stages realism should prevail, as now it is important to show and validate that
the new product and process indeed work. In the detailed engineering and startup
stage pessimism should prevail, as now all details of the design and the startup also
have been right. Pessimistic people are very good in finding detailed concerns why
the novel product and process will not work. All these concerns should be turned into
specific modifications and actions to still get the product and process working.
3.6.5 Intellectual property creation and protection
Intellectual property is something that is new, useful in commercial practice, is doc-

umented and protected from use by others. Intellectual property can be created by
a design, as a design always has new elements, and is nearly always useful when ex-
ploited in practice. It is always documented. Protection from commercial use by others
is then the last step. There are many ways to protect intellectual property. One way is
to keep it secret. For small companies this may be doable. For larger companies this
will be hard. A second way is to file patents. Now the information shown in the patent
is made public but cannot be used by the competitors for business purposes. A third
way is to file copy rights for essential documents containing the innovation. A fourth
way is to file a design right. A fifth way is to file a trade mark.
The main error made on this topic that the novelty is presented orally or in writing
in the open. Open means any public meeting place of any publication in any form.
Even bilateral discussions with no formal nondisclosure agreement are public. Once
the novelty is public no patent protection is possible anymore.
Literature on intellectual property generation and protection is vast. Verloop ar-
gues that intellectual property is a form of intellectual capital, and should be used
defensively, i.e., preventing other companies from commercially exploiting the inno-
vation and offensively by exploiting it in various ways [16].
3.7 Discovery stage
3.7.1 Purpose discovery stage
The purpose of the discovery stage is to obtain ideas and to define the ideas to such a
level that it becomes clear what they are about so that at the end of the discovery stage
it is clear whether they fit to the company strategy can pass on to the concept stage,
or not.
3.7 Discovery stage | 49
3.7.2 Managing projects in discovery stage
The discovery stage can be compared to hatching of an egg. The only essential condi-
tions for success, provided by the hen, are protection and warmth. The actual devel-
opment inside the egg is invisible and autonomous.
The same holds for ideas generation and development in the discovery stage.
Companies highly dependent on a continuous stream of new ideas to feed their in-
novation funnel, protect researchers in their early idea development. Some, like IBM
and 3M, do it by letting researchers spending 10–15% of their time on pursuing their
idea, without requiring a formal budget and reporting.
Other companies have a formal organization such as a Game Changer organiza-
tion where researchers can ask for a budget to pursue their idea, any time of the year,
with a minimal description of the idea. A member of the game changer team will then
help to describe the idea such that it fits in the company strategy and will provide a
budget to perform the most critical test to demonstrate the feasibility of the idea. Often
this is a proof of principle experiment for a novel product or process step.
If the business case looks sufficiently attractive but the idea involves many inno-
vation aspects then the idea may also be analyzed for what specific additional inno-
vations will be required. Once the idea has been proven in some way and the business
case has been sufficiently developed, the researcher will propose the idea for taking up
in the annual research budget round, see also the table generic criteria for stage-gate
decisions.
3.7.3 Creativity methods
In the discovery stage creativity is needed to generate many and good ideas. The lit-
erature on creativity methods for innovation is vast. A good introduction to further
information is by Nijstad [41] and Hermann, [42].
Brainstorming is a good method to generate unusual associations and new per-
spectives to the problem at hand. The most important rules are:
(a) No critical remarks on ideas
(b) Associate and enlarge ideas
Brainwriting is another useful method. It works well for introverted people, as they
can write ideas down on their own. Process engineers in general are introverted; hence
this method is mentioned. The complete method can be found in the Appendix A3.
3.7.4 Discovery stage-gate decision evaluation criteria
What in general needs to be done in the discovery stage is expressed here under ele-
ments of the reporting for the stage-gate decision.
3.7.5 Idea description
If the idea is a product then a functional description of the purpose of the product is
needed. Options for product composition to fulfill the product functionality may be
given. If the idea concerns a process then a block flow diagram with feedstock and
product streams is required.
3.7.6 Proof of principle
The result of a proof of principle experiment for the main critical novelty of product or
process is needed. The challenge for the researcher is to define an experimental setup
that is simple, quick to construct and such that the test result can convince others that
the novel principle works. In general it requires several attempts before the proof of
principle is successful.
3.7.7 Business case
The business case is the business justification of a project [43]. It is very clear from this
definition that it specifically belongs to a specific company. The business case cannot
be established by others outside a company. Or in other words the company owns the
business case. Even better a specified person inside the company owns the business
case. This is not the project leader.
The business case development of an idea in the discovery stage may be done by
answering the following questions [43]:
(1) Does it create value for the company?
(2) Does it fit to the strategy of the company?
(3) Does the company need the idea?
(4) Can the company (together with others) develop the idea?
For a project to pass the discovery stage gate these four questions of the business case
may be used. If positively answered then the preliminary business case for the idea is
established.
3.7.8 Creating value
The question of creating value for the company through radically novel products may
be hard to answer. It may be answered in stating, yes, we believe in its capacity to
create value for the company and society.
A simple economic potential estimate may be provided on the potential economic
contribution per year of the novel product or process, when commercialized. It may be
done by just estimating a sales price per ton and subtracting from that the feedstock
cost per ton of product and estimating the total sales in ton/year.
3.7.9 Strategic fit
The fit of the novel product or process with the company strategy will be indicated.
If it cannot be made fit to the strategy and a strategy change of the company is not
foreseeable then the idea should not be pursued. However, some champions of an
idea are not easily stopped and because of their stubbornness and creativity, may find
ways to pursue the idea.
3.7.10 Necessity of the idea
Even if the novel product and/or process fits the company strategy, still the question
remains of whether the idea is needed. If so, then it will be pursued. If not, it can be
stored and may be retrieved when circumstances change and the idea is needed. This,
however, requires an idea storage and retrieval system that most companies will not
have.
3.7.11 Idea development doable (with others)
The question of whether the idea can be developed by the company with help of others
is also often hard to answer. It requires knowledge of what technology providers and
academic research groups are available to solve some of the foreseen knowledge gaps.
3.8 Concept stage
3.8.1 Purpose of concept stage
The purpose of the concept stage is to obtain a full picture of what the goal of the novel
concept is, what knowledge is needed, what knowledge is available, and which most
critical success factors are to be investigated in the concept stage so that at the end of
this stage, a decision can be taken, based on all criteria, to pursue the concept or not.
For a novel product, this involves developing a customer value proposition, tech-
nical requirements for product and process definition, identification of critical to qual-
ity variables and experiments to proof the concept.
3.8.2 Customer value proposition in concept stage
For a novel product concept, it means, first, that a customer value proposition is made.
This is a marketing statement that describes why a customer should buy a product.
For a business to business product it consists of the sum of total benefits that a vendor
promises a customer will receive in return for the customer’s associated payment [44].
Table 3.15 contains its main elements.
Tab. 3.15: Customer Value Proposition elements.
Customer value proposition elements Description

Product type Single product, or product family, product platform
Technical value proposition Technical advantages over existing product
Product technology Product components
Manufacturing process technology Process options for manufacturing of new product
Materials technology Input materials for the new product
This customer value proposition requires a strong interaction between the product de-
velopers and the business and marketing staff of the company. Design thinking and de-
sign models help these conversations enormously by having drawings and pictures of
the product design, by having both a functional product description (what it does) and
by having several physiochemical alternatives fulfilling those functions. By also mak-
ing basic process concept designs with material balances, first indications of feedstock
cost per alternative are available. Following a checklist derived from human senses
may be useful to make sure that all relevant aspects of the product requirements are
covered:
– Feel to relevant body parts
– Visual appearance
– Smell
– Taste
– Sound
3.8.3 Technical product requirements in concept stage
The primary consumer perception requirements should be translated into techni-

cal requirements. Other functional goals and requirements from the customer value
proposition also should be translated into technical requirements. This translation
may involve several steps, before a list of quantified technical specifications of the
product is obtained. Quality function deployment and house of quality methods may
be used for this translation of customer needs into technical requirements. The sum-
mary of the technical product requirements is called critical-to-quality requirements.
3.8.4 Process concept generation
The new method of process concept generation comprises designing a functional

block-flow diagram with its material inputs and outputs for the chosen product con-
cept. If several product concepts have been generated then at least one functional
block-flow diagram for each product alternative will be made. Major energy inputs
and outputs will also be determined. For each functional (task) block, feasible unit
operations will be selected, or even better, functions will be combined and feasible
functional integrated concept solutions will be defined.
The dominant method of the past, designing with unit operations, does not de-
liver breakthrough innovation anymore. Process design by task integration is emerg-
ing now as the dominant design method, as the advantages in investment and operat-
ing cost, lower environmental impact and inherent safe design have been proven for
hundreds of commercial scale implementations [8]. The design method is described
in detail in Chapter 6 and 7.
3.8.5 Proof of process concept
All key issues are derived from the process concept design and a lab-bench scale re-
search program is defined and executed to remove these key issues and establish in-
formation for a better process concept design.
3.8.6 Selection of best process alternative
The best process alternative will be selected based on SHEETS criteria (see Chapter 4),
with often a high weight factor for lowest feedstock cost and lowest investment cost.
Selection methods are provided in Chapter 7.
3.8.7 Process concept design
For the best alternative, a process concept design will be made containing:
– Feedstock specifications
– Product specifications and quality performance requirements
– Catalyst type and amount
– Byproduct stream specifications
– Stream compositions for each stream
– Heat balances for each process unit
– Selected unit operations with pressures, temperature, and residence time
– Selected equipment types and main sizing aspects
– Choices of construction materials
– Main control strategy is indicated for sensitive process aspects
– Auxiliaries
– Utilities needed
3.8.8 Basic design data generation
Basic design data will be generated for the concept design. Chapter 6 provides details
on what data are to be generated and how this can be done efficiently.
3.8.9 Concept stage gate evaluation
Business and research management will be informed about the results of the concept
stage and about the planning and cost of the feasibility stage. For all criteria informa-
tion will be provided such that a decision to pass the stage gate or stop the project can
be made.
3.9 Feasibility stage
The purpose of the feasibility stage is to verify the feasibility of the product concept
and the process concept for all critical aspects for success.
3.9.1 Concurrent product-process design and testing
The technical product feasibility is proven by experimentally testing the integrated

product. Often this integrated product is called the prototype. Here unexpected effects
of integration may cause the product to fail. Often the failure can be repaired by anal-
ysis of the cause of the failure. Because the tests are carried out at laboratory scale,
modifications are fast and at low cost.
Through the direct customer and end-consumer feedback, both the product de-
sign and the optimal process design can be fixed. The product design includes not
only the optimal composition and structure, but also will contain tolerable composi-
tion levels (including impurities). Also, the option to have multiple material suppliers
validated, will help later in the negotiation of raw materials purchasing and supply
conditions. Finally, the tolerable key product attributes and their relationship with
composition and processing should be very clearly reported for testing in the next
stage: the manufacturing stage.
3.9.2 Microplant in feasibility stage
The best process design concept is integrally tested experimentally at laboratory scale.
This test unit is best called a microplant.
A properly designed microplant contains all major process units in the same line
up, recycle structure, and commercial scale design. It contains all reactions and sep-
arations and it can operate in the commercial design window of process conditions.
Its purpose is to study the reactions and separations under the design conditions,
with recycling. Reaction conversion rates and byproduct formation under recycle con-
ditions are determined to improve reaction kinetics. Separation performance results
are used to validate and improve physical equilibrium models for the separation steps.
Including the recycle flows in the microplant is also very useful. Heavy boiling
species for instance can build up in it, affecting the reactions and the separations.
Knowing this is of enormous value for improving the commercial scale design and
reducing its risk of failure.
Because of its small scale, the construction and operation cost are low. Often is it
made from glass, by which problems such as foaming and fouling are also quickly lo-
calized and the process is quickly modified to solve them. It can be used to test various
process design options [45].
A downscaled microplant is a very cost-effective element to reduce risks and pro-
vide data for optimization of the commercial scale design. It is the smallest scale of
an integrated version of the commercial scale design concept including recycle flows.
The typical production capacity range is 1–100 gram/h and the equipment diameter is
typically 1–5 mm [45].
The emergence of microscale equipment with well-defined residence time dis-
tribution and uniform temperatures has helped enormously to boost the use of mi-
croscale units.
For new pharmaceutical products, these microplants increase the speed of pro-
cess development and thereby reduce the time to market. In some cases, the laboratory
scale process is run for several weeks to quickly provide product for clinical testing.
In fact, a whole new industry branch of very small-scale devices has emerged [46]
with over 274 companies providing these devices for the pharmaceutical industry [47].
A new term, microfluidics, is often used for these devices. The term refers to devices,
systems, and methods for the manipulation of very small fluid flows, as small as micro,
nano, pico and even femtoliter/hr flows [48]. Other terms used are: lab-on-a-chip and
flow chemistry.
The application of microflow systems is increasing also for fine chemicals. In this
industry branch, most of the production is carried out in a batch operated mechani-
cally stirred reactor. Cooling is provided by having a solvent that evaporates from the
reaction liquid, which in turn is cooled by an overhead condenser. The cooled liquid
solvent is returned to the reaction liquid.
This reactor and its operation have large scale-up uncertainties, which are solved
by an experimental program on the batch reactor. Byproduct formation is also often
far more than the amount under ideal conditions, due to the far longer residence time
than required for the reaction itself. Also, the solvent after its use is becoming a waste
byproduct, similarly with the cleaning liquid in between batches.
Boodhoo describes the following microprocess technologies in some detail [50]:
micromixers, microheat exchangers, and microreactors. Of the latter class, multichan-
nel reactors and spinning disk reactors are described in detail. Günther provides an
overview of multiphase microfluidics for chemicals production at laboratory scale [51]
and Kenning covers microfluidic unit operations: extraction, absorption/desorption,
distillation, as well as particle or droplets separation [52].
If successful, the process design is completed with process instrumentation and
control and the product packaging method is designed. The degree of design defini-
tion for an economic assessment is typically 50%. This means that 50% of the invest-
ment is based on the process design and 50% is based on correlations for cost of plot
preparation, concrete structure, and outside battery limits.
Remaining uncertainties regarding process integration and scale-up will be ad-
dressed in a pilot plant in the development stage. The design and cost of the pilot plant
and cold-flow model, however, belong to the feasibility stage. It may well be that the
cost of the pilot plant is so high that the project stops at the feasibility stage gate.
The feasibility of the product and the process design are assessed at the stage gate
using all criteria of Chapters 10–13 and focused on validation of its feasibility for all
those criteria. Details on the feasibility stage gate evaluation will be found in these
evaluation chapters.
3.9.3 Scale-up strategy and information
The first question to answer is what scale-up strategy should be followed. For novel
products with limited patent protection and high added value, such as for pharma-
ceuticals, a short R&D time is very important. Equipment cost is in general a minor
Tab. 3.16: Scale-up methods available for various process equipment. Scale-up methods: Brute
Force (B.F.), Correlation based (Cor.), Modeling based (Mod.).
Equipment Residence time distr. Micromixing Mass transfer

Trickle bed B.F., Cor. + dilution B. F B.F., Cor. + wetting
L, G Packed bed B.F., Cor. B.F. B.F., Cor.
Static mixer L, G. B.F., Cor., Mod. B.F., Mod., Cor. B.F., Cor., Mod
Static mixer G/L B.F., Cor., Mod. B.F. B.F., Cor., Mod.
Tray column Cor.
Plate column Cor.
G/L Bubble column Cor., Modeling Unknown Cor.
Cross flow bub. col. Cor., Modeling Unknown Cor.
Mech st. tank L, G Cor., Mod. Mod. –
Mech st. tank G/L L Cor.; G unknown Unknown Unknown
item in the overall cost. So, easily scalable equipment requiring only one validation
step are preferred. Millimeter flow reactors for which the brute force method can be
applied have here a large advantage.
For bulk products for which reliable production to satisfy the customers and to
minimize the startup cost of the commercial scale process is very important, the scale-
up risk needs to be low. The innovation class table with its risk data of Section 3.3 can
be used to assess the risk involved.
Table 3.16 provides for some, often applied, process equipment available scale-up
methods.
For reaction, the residence time distribution (r.t.d.) is always relevant. For two-
phase reactors both the r.t.d. and the mass transfer is always relevant. If the reactions
are fast and multiple reactions are involved often micromixing is also relevant.
For crystallization, the micromixing and the residence time distribution and the
mass transfer is relevant.
For gas/liquid phase separations such as in distillation and absorption, both the
mass transfer and r.t.d. are important.
A special warning must be given for the scale-up of two-phase mechanically
stirred reactors. These reactors have a highly nonuniform turbulent intensity profile.
Near the impeller high shear and energy dissipation occurs. There the dispersion is
made very fine. Further away the turbulent field is less intense and there coalescence
of the dispersed bubbles or droplets can occur. At scale-up the areas of high and low
turbulent intensity change and often the ratio of high and low intensity changes. This
makes scale-up notoriously uncertain. In emulsion polymerizations carried out in
mechanically stirred tank reactors where the dispersion quality also determines the
final product quality, scale-up is still a high-risk affair. Choosing other equipment
such as static mixers would avoid this scale-up problem.
3.9.4 Scale-up information equipment
For all process equipment scale-up information will have to be provided, so that the
commercial scale design base and its associated risk, is known for each process equip-
ment. In some cases, the risk may be reduced by mitigation measures.
3.9.5 Business case feasibility stage
In the feasibility stage, the business case is much further developed. Also, here the
business case values of all projects are established and so the value of the whole port-
folio of innovation projects [23].
3.10 Development stage
3.10.1 Purpose development stage
The purpose of the development stage is to reduce the risk of introducing the novel
product and process to acceptable levels. This means that at the end of the develop-
ment stage the chance of a successful product launch is high and that the commercial
scale process will meet the product requirements and process capacity and that it can
be started up within the average industry startup time.
3.10.2 Product testing in development stage
For product performance at the client level, different product versions are produced at
a processing scale that is large enough to perform an accurate scale-up of the process
and/or is large enough to produce sufficient amounts for testing the product in real
conditions at the direct customer and/or end user of the final product with the new
ingredient/component. Customer evaluations of the product should reveal the appre-
ciation for, and the value of the product for them. Modifications of the product are still
possible and needed, not only to verify the product functions, but also to maximize the
business opportunity by tuning the added value parts of the product and reducing the
costs (raw materials, processing) of producing the product.
Often a miniplant is run to produce different product versions based on raw mate-
rials (different suppliers, quality levels, process settings, etc.) and to validate the prod-
uct in the industrial customers processes and in the final consumers’ application. For
this purpose, sufficient material for testing should be produced, and this often also
determines the size of the mini- or pilot plant. If this scale is not large enough and a
pilot plant will be necessary, that plant will be constructed and operated in the devel-
opment stage.
Also, critical logistics processes, storage, filling, packaging, transportation, un-

loading, etc. at direct customers may be identified and verified with the product
batches. Again, if major investment cost is involved then that investment will be part
of the development stage.
3.10.3 Process testing in development stage
3.10.3.1 Miniplant purpose and design

The miniplant ideally is also a downscaled version of the commercial scale design
including all recycling, but also now containing the same type of reactors and sepa-
rators with the same number of mass transfer units, residence time distribution, and
the same construction materials as the commercial scale. It typically has a production
capacity of 0.1 kg/h [45].
With this miniplant, further validation of the design concept and of the process
models are established and problems are quickly identified and solved. Potential in-
teraction between the construction materials and the components such as catalysis
by construction material and corrosion of construction material can now also be ob-
served as these plants operate continuously and for a long period (several months).
Additionally, corrosion test material samples are placed in all streams to optimize
for the commercial scale the construction material choices. Automatic process con-
trol and data storage is purchased from a reliable supplier of the distributed control
system.
In some cases, a miniplant avoids the need of a large pilot plant and then saves
a large investment. The pilot plant section demonstrates when a larger pilot plant is
still needed.
The following design documents for a miniplant are required [45]:
– A process flow sheet
– An engineering flow diagram
– All equipment types and sizes
– Objectives of instrumentation
– Specifications for insulation of equipment and piping
– List of measuring points
– Safety concept and logic: how sensors are linked to actuators and safety procedure
3.10.3.2 Pilot plant purposes and design

The first question to answer is: is a pilot plant needed? The pilot plant can be needed
for:
(A) Producing a large amount of product for quality and market testing
(B) Identifying scale-up problems
(C) Validating process models and providing input for mass and energy balance up-
dates
(D) Training operating personnel of the future commercial plant
(E) Improving the estimates of catalyst life times
(F) Testing construction materials for corrosion
(G) Validating process control and operating procedures
A pilot plant is also very useful in getting reliable stream compositions. This increases
the stream knowledge factor F of the startup time correlation from nearly 0 to 1 (see the
section on startup). The difference in startup time of those two F values is on average
3.2 months. The amount of money lost by not having a specification product for that
period in general easily outweighs the cost of a pilot plant.
3.10.3.3 Pilot plant as downscaled version commercial scale design

If a pilot plant is needed then it should always be a downscaled version of the com-
mercial scale plant with all unit operations and all recycle flows. Also, the commercial
scale process conditions such as temperatures, pressures, and residence times should
be inside the pilot plant operating window. Only in this way the tests will be represen-
tative of the previously optimized commercial scale plant.
If the pilot plan is not designed as a downscaled version of the commercial scale
then this means that the commercial scale plant must contain the same process type
units and recycling as the pilot plant. In some cases, this means that multiple units
must be selected in parallel for the commercial scale, because the selected equipment
for the pilot plant appeared not to be scalable. It will be clear that this way of scale-up
means a higher investment cost for the commercial scale process. So it is highly unde-
sirable, although it is still a practice in some companies. If the commercial scale design
also differs from the pilot plant in process type and recycle structure, then the risk of
failure increases enormously. According to Merrow this is one of the major causes of
commercial scale failure, meaning that the startup time is 40% longer than budgeted
for (see startup time prediction) with a 40% additional cost to remedy the failure [33].
A: Pilot plant for producing large amount of product for testing

Often a large amount of product is needed for performance tests for one or more
clients. This testing often is desired by the client even if the product is promised to be
the same and only the process has been modified. This even holds for intermediate
bulk products. Trace amounts not detected by analysis can still affect the product
performance at the level of the client; for instance, when the new trace components
degrade the catalyst in the client process.
If the product is novel, large amounts are often needed for all kinds of market
launch tests [45].
B: Pilot plant to identify scale-up process problems

A pilot plant is needed to identify scale-up process challenges if either [53]:
– More than one new process step is involved
– The process step contains a new step and has a recycle stream over more than one
step
– The process contains a novel solids processing step
– The feedstock is a crude source, whose composition knowledge is incomplete
If the pilot plant is needed but is still omitted, startups in most cases lead to failure,
which means that the process as such cannot be operated and/or meet the design
specifications on product quality and capacity [54]. An additional budget of over 30%
of the investment is often needed to solve the problems [23]. These costs far exceed
the pilot plant cost, which are in most cases less than 10% of the commercial scale
investment cost.
C: Pilot plant to validate process simulation models

Process simulation models are now the main source of detailed information of pro-
cess streams composition knowledge everywhere in the process including the final
product output stream. This information is used to select the construction materials,
to select and size equipment, and to define utility requirements. The startup time cor-
relation parameter F also indicates the importance of this knowledge as shown in the
startup section of this chapter. It is, therefore, important that the simulation models
are reliable. Extensive validation of these models can be carried out by measurements
of stream compositions, temperature, and pressure and comparing them with model
predictions.
D: Training operating personnel of the future commercial plant

A pilot plant can also be used to train operating personnel of the future commercial
plant. Nowadays this training is mostly carried out using dynamic simulation mod-
els, but additional training with a real pilot plant helps to get acquainted with, for in-
stance, taking samples for analysis and seeing what the intermediate stream samples
and product samples look like and how misoperation also shows up in the changing
colors of those samples.
E: Improving the estimates of catalyst life times

Long-term pilot runs under actual process conditions will show the catalyst lifetime
and how the process stream compositions change with deteriorating catalysts. This
type of testing can also be done in a miniplant.
F: Testing construction materials for corrosion

By placing many material test samples in many pilot plant streams and running the
pilot plant for a long time, accurate corrosion rate information can be provided, which
can be used to improve the final materials selections for the commercial scale plant.
Many problems in commercial scale plant operations are still connected to materials
corrosion [53]. This corrosion testing can also be done in a miniplant; see the section
above on miniplant.
G: Validating process control and operating procedures

In process designs with strong dynamic interactions between various process parts,
process control may be difficult to achieve. Testing the chosen process control design
and tuning the process control in the pilot plant can be very important. Operating
procedures can also be tested by process operators.
This validation is particularly important for batch processes, who by nature have
dynamic behaviors. The validation of process control design and operating procedures
is a necessity for them.
3.10.4 Pilot plant engineering, procurement, and construction (EPC) company

choice
The detailed design and construction of the pilot plant should be done with the same
rigor as the commercial scale plant. The selection of the Engineering, Procurement,
and Construction (EPC) company for the pilot plant is, therefore. important. A com-
pany dedicated to pilot plant EPC is strongly preferred. If an EPC company is chosen
that handles both commercial scale projects and pilot plants, then it is likely that the
pilot plant will get a far lower priority and lower qualified personnel, as the size of the
pilot plant project in money terms is a fraction of a commercial scale project. This in
turn can mean that the agreed timing is not adhered to and that the pilot plant will
contain errors. For this reason, an EPC company only concerned with pilot plants is
the strongly preferred choice.
3.10.4.1 Equipment scale-up effects determination

Commercial scale equipment will be larger than preceding test equipment in the de-
velopment stage. In general, this means that the hydrodynamics in the equipment will
be different. The following transport phenomena will change with the change in hy-
drodynamics and in turn can affect the performance of the equipment:
3.10.4.2 Phenomena in equipment sensitive to scale-up and influencing

performance
– Residence time distribution
– Micromixing
– Mass transfer
– Dispersion of a second phase
– Heat transfer
– Momentum transfer
For most standard pieces of equipment these transport phenomena are known as a
function of geometric sizes, fluid velocities, and fluid properties; often in the form
of correlations. The relations between the phenomena and the performance are also
known and often expressed in models.
A lot of conventional reactor equipment and gas-liquid contact equipment correla-
tions for the transport phenomena are available in textbooks such as Perry’s Chemical
Engineers’ Handbook [55].
A word of warning is needed here. Often the mass transfer correlations assume
implicitly that the residence time distribution of the fluids is both plug flow, while
deviations from plug flow occur. Often these mass transfer correlations have been ob-
tained under the conditions of plug flow. At scale-up the deviations from plug flow can
become larger and then the commercial scale performance is worse than predicted us-
ing the mass transfer correlation.
For tray type gas-liquid columns, often the residence time distribution of the liq-
uid is assumed to be completely back-mixed, as well as for the gas phase. For very
large diameter trays with one or a few down comers for the liquid phase this assump-
tion is sometimes not valid. Technology providers of large distillation trays have this
knowledge available for their clients.
If the relation between the equipment size and the phenomena are not known
then that relation must be determined. There are two established methods for deter-
mining the relation:
Method A: The phenomena are determined in the same piece of equipment but used
in a different process.
Method B: A large scale of that equipment is constructed and the relevant phenomena
are studies in that piece of equipment. To keep the cost low often simple fluids
such as air and water are used, rather than the actual fluids. Such a test facility is
called a mockup or cold flow model.
3.10.4.3 Mockup design and testing for hydrodynamic scale-up effects

Nowadays computational fluid dynamics (CFD) flow models are often made for equip-
ment with complex geometries and complex hydrodynamic flow behavior affecting
performance upon scale-up. These models are then used to design and optimize the
equipment geometries of the commercial scale. These models, therefore, need vali-
dation. This validation is carried out in so-called mockups where inert fluids such as
water and nitrogen are used and mixing, residence time distribution, and mass trans-
fer are measured and compared with the model predictions.
3.10.4.4 Development stage gate evaluation

The purpose of development stage gate evaluation is to decide whether to go to the de-
tailed design stage or not. As the detailed design stage is very costly, this is an impor-
tant decision. The economic evaluation of the commercial scale design for this stage
gate is, therefore, very important.
Formal evaluation methods found in the evaluation chapters can be used as in-
put for the gate decision. It is advised to consider all criteria for the decision, not just
the economics. The gatekeepers for the decision should be experienced and have the
authority to take the go/no-go decision.
3.11 Detailed design stage
3.11.1 Detailed product design
The product has already been designed in the development stage. However, product
pricing adaptation to latest developments in the market, details of packing, logistics,
and transport planning and if needed additional buildings, are now also to be defined.
3.11.2 Detailed process engineering
The detailed process engineering stage comprises detailed process design, procure-
ment of all process equipment, and construction of the process. These three steps are
in general executed by Engineering, Procurement, Construction (EPC) contractors. For
a detailed description of the activities of this stage and how the manufacturing com-
pany cooperates with the EPC contractor, the reader is referred to other books such as
Bakker’s [23].
The starting point for the detailed design is the commercial scale concept design
provided by the manufacturing company. If only a preliminary concept process design
is delivered then the EPC contractor will make a concept design. In general first a pro-
cess design is made with sufficient definition to allow an investment estimate of ± 30%
with which the manufacturing customer can decide to continue the project or not and
then a design allowing a ± 10% accurate estimate. These two levels of definition and
accuracy required differ a little from company to company. Clarification between the
customer and the EPC contractor is always needed. Also, a complete evaluation for all
other criteria for now and future trends on safety, health, environment, social accep-
3.11 Detailed design stage | 65
tance, and legal items of the local country (license to operate) is carried out as part
of concept design. The customer then decides to go for detailed engineering, procure-
ment of equipment, and construction, or not.
Nomenclature of process concept designs varies enormously between companies
and clarification should always be asked for. Process concept design with 30% capital
investment inaccuracy, is also called front-end loading 1 (FEL-1). The next concept
design with 20% inaccuracy, is also called FEL-2, and the third process concept design
with 10% inaccuracy is also called FEL-3 [23].
Detailed engineering is an enormous effort with many engineering specialists in-
volved. Here only some main elements are mentioned.
(1) A complete piping and instrumentation diagram (P&ID) is made for every part
of the process. It shows all piping connections to all equipment and all controls
and instrumentation. This is an enormously important piece of information for
process construction and safe and proper operation. To get a feeling for the im-
portance, consider that, on average, piping concerns 20% of the total process in-
vestment cost. A large part of that cost is making specifications for each pipe on,
construction material, tracing or not, sizing, and precise 3D location.
(2) A complete process description is made for each unit operation containing dimen-
sions, capacities, temperatures, pressures, construction material, chemical and
physical knowledge, and data.
(3) A 3D plant layout plan is made showing both all equipment and piping.
(4) A detailed commissioning and startup instruction manual is made.
(5) Equipment detailed specifications for purchasing are defined.
(6) Detailed drawings for special equipment, to be custom made, are sketched.
3.11.3 Choice of EPC contractor
The choice of the EPC contractor for process design and construction is very important.
The selection should be made by the project leader of the manufacturing company.
Any involvement by business leadership in the selection of the individual contractors
is inappropriate. Any involvement of the lead sponsor’s purchasing organization in
the selection of contractors is likely to be catastrophic. Even more important is the
selection of the right contractor team, a team that the manufacturing company team
can work with creatively [33].
The EPC contractor should have experience in the relevant process industry
branch. For the food industry branch for instance, that contractor will know how
to prevent contamination, how to ensure complete cleaning of equipment, and how
to avoid undesired micro-organism growth.
For a breakthrough novel process unit within the overall process, additional con-
tractors to the main contractor may be involved, with specific knowledge of designing
for the materials flowing through the process unit and with knowledge of designing for
similar stream compositions. That contractor will, for instance, design in additional
measures to prevent or removing local fouling, or select at critical locations an instru-
ment that is robust to fouling.
3.11.4 Demonstration plant
For a radically novel process that in the end will be built for very large capacities and
with a large investment, first a demonstration process is in most cases designed and
operated. This demonstration process is in general 1–10% the capacity of the final com-
mercial scale. It may be 10–30% of the full-scale investment cost.
The main reasons for having this demonstration plant are to reduce real and per-
ceived risks. The risk can be technical and/or business-related (uncertain new market
development).
The technical risk of a novel process (and product) is reduced simply by the lower
investment cost. The higher investment cost per ton of product for this demonstration
scale may also be earned back later by incorporating the learning points of the demon-
stration plant in the final full capacity plant. The savings through these learning points
can be small, but also be as much as 40% of the final investment, depending on how
much redundancy has been built in by the engineers in the demonstration plant [53].
The redundancy measures can be operating safety margins, oversized equipment, or
even spare equipment.
The business risk due to new market development is uncertain. Sometimes this
demonstration plant is needed, to convince investors with no experience in the pro-
cess industry (inside or outside the company) that reliable processing is feasible and
that promised return on investment can be made.
It is very important that up front it is made certain that the demonstration plant
will make a profit. It is also important that the lower return on investment of the
demonstration plant is reported clearly to the business department of the company.
If either is not carried out the commercial scale plant will not become reality.
An example of a demonstration plant prior to the final full-size plant is the Shell
GTL plant in Bintulu. It has a capacity of 14,700 bbl./d of liquid products. It started
up around 1990. Its capacity is a factor of 100 smaller than the commercial scale
PEARL plant in Qatar of 140,000 bbl./d of liquid products. The latter plant started up
in 2011 [56].
3.12 Process startup and product launch
3.12.1 Introduction
A product market introduction and in particular a consumer product introduction re-

quires an enormous amount of planning and preparation. Here are some main ele-
ments for consumer product introduction. A business to business product introduc-
tion will contain similar elements, but to a lesser extent.
3.12.2 Panel for product testing
The consumer product will be tested by a panel first consisting of company personnel
and later by a panel of outside potential customers of the product. If it is a food product
the panel will be asked to mention the look, the smell, mouth feel, and the taste. If it
for instance is a laundry soap the potential users will be asked to read the instructions
and use it in a washing machine. All reactions of the panel will be monitored and the
product and instruction may undergo a final adaptation, based on the panel feedback.
Here is an example of such a panel test for a new, more concentrated laundry
detergent. Because of the more concentrated detergent, far less is needed per laun-
dry batch. The external panel members, however, could not believe that so little soap
would work and still were tempted to add (nearly) the same amount as before.
3.12.3 Product launch planning
Detailed planning will be made by the producer for all elements of the product launch.
Market intelligence will be used for the product launch planning. It will include how
much, when, and where to supply the product for the first time. A prelaunch will nearly
always be planned in which the product will be launched in a certain location (a town,
or a region) only. The purpose of the prelaunch is to validate market expectations and
to detect imperfections in the logistics. After the successful prelaunch, the full prod-
uct introduction will start. Contract preparation and signing in time with all relevant
stakeholders is of course an essential element in the planning.
3.12.4 Matching marketing and manufacturing timing
It is of utmost importance that the planning is made by intimate contact between mar-
keting and manufacturing to fit the manufacturing production availability and capac-
ity to the market planning.
In some cases, the pilot plant will be used to make sufficient product for panel test-
ing or even the prelaunch. Up front, in the design of the pilot plant years earlier than
the product prelaunch, special attention will also be paid to product packaging. This
has all been considered by having intimate contact between marketing, development,
and manufacturing.
For consumer product oriented companies, this regular exchange of information
and matching marketing with development manufacturing is a common practice.
For business to business manufacturing oriented companies that move to consumer
products, this will be new, and structured connections between marketing, develop-
ment, and manufacturing are often absent, causing risk of mismatching the product
prelaunch and manufacturing.
3.12.5 Information to supply chain and customers
All relevant stakeholders in the supply chain (transport, storage, shop) and customers
will be informed about the prelaunch and the launch of the new product by proven
and novel media means. The product description for each stakeholder is an important
aspect, and the fine-tuning of the product introduction price is an important element
of the product introduction to the market.
3.12.6 Process implementation
In the implementation stage of a novel process for an existing business to business

product, the startup should be viewed as an experiment that is not allowed to fail. If
it fails then the client does not receive the product and in turn cannot produce its own
product. That client will look for other suppliers, as security of supply is enormously
important to stay in business.
This means that precaution measures must be taken. The measures are regarding
the process design, construction, preparation, startup, and regular operation.
3.12.7 Recognition of new commercial implementations requires special

preparation
The first step in any implementation is to recognize that something new is to be im-
plemented. Most often failures are due to not recognizing that the process is new and
perceiving that the process is the same as before. An analysis of projects containing re-
vamps, expansions, and green fields, highly innovative and lowly innovative, revealed
that the most successful projects in time and cost were innovative projects. However,
the worst projects were also innovative, but in the latter cases the technical difficulty
was grossly underestimated. The innovative projects were treated as standard tech-
nology and in many cases adequate testing of the process was absent as well as sound
basic data packages [57].
Heuristic: A process is novel if it has not previously been in operation with that
feed, that product, that catalyst, that piece of equipment, or those conditions at
commercial scale
This heuristic rule is often violated. A case I have experienced was a crystallization
process step applied for decades in the beverage industry that was introduced at a
bulk chemical plant. Because the process step was considered commercially proven,
smooth operation was expected. The process step was interrupted every few weeks
and requiring cleaning before regular operation could be continued. After questioning
it appeared that in the beverage industry cleaning was carried out every week as part
of the normal operating procedure. The bulk chemical process, however, was expected
to run continuously for four years.
Another case I have experienced is the use of a new batch of catalyst that is per-
ceived as the same as before. The same procedure for startup is followed and then
all kinds of problems appear, because in reality the catalyst behaves differently from
the previous catalyst. What can happen if novelty is not recognized is dramatically
illustrated in the Shell case hereafter.
3.12.8 Case of Shell Moerdijk where new catalyst charge leads to explosion
A dramatic explosion occurred at the Styrene process at Shell in Moerdijk, the Nether-
lands in 2014, because a new catalyst was charged to a reactor. Fortunately, nobody
was injured, because nobody was present in the plant. The damage, however, was so
large that it took one year and a half to rebuild the process.
The catalyst was perceived as an improved version of the previously used cata-
lyst. Performance tests under production conditions had been carried out at lab scale.
However, nobody had considered the consequences of the difference between chemi-
cal composition of the catalyst during preconditioning and startup and no formal new
startup and operation risk assessment was carried out. The new catalyst contained 5%
Chromate VI instead of the 0.2% in the old catalyst.
The preconditioning of the catalyst in the commercial reactor occurred with ethyl-
benzene, as always had been used before. Ethyl benzene was considered an inert
flushing liquid just to heat up the catalyst bed. In reality, the chromate reacted exother-
mically with the ethylbenzene causing in the end a runaway reaction resulting in an
explosion destroying large parts of the process. As always with a calamity like this
there were additional causes resulting in the explosion. The spray installation for the
trickle-bed reactor did not function optimally for distributing the ethylbenzene by
which local hotspots in the catalyst bed occurred that could not be noticed by the
operators. At these hotspots the copper oxide in the catalyst started to react with the
ethylbenzene, causing more heat to be produced, which caused evaporation of the

ethylbenzene and thereby an enormous pressure increase. In the design, this high
pressure was not considered and a pressure relief valve was not part of the design.
In addition, the operators of the process at that time had never carried out a startup
of this reactor. They did not notice differences such as an unusual pressure profile,
a strongly fluctuating ethylbenzene flow, and a controller alarm. Manual operation
rather than automatic control operation also contributed to the explosion [58].
3.12.9 First commercial scale process startup
3.12.9.1 Startup preparation

From literature on industrial process startups, Harmsen extracted four critical success
factors for the design stages and six for the startup stage [59].
The design success factors are:
– Project is identified as new.
– An integrated pilot plan, as downscaled version of the commercial scale process
has been used
– R&D, detailed design and construction personal have meeting to ensure that all
knowledge is integrated in the design and construction.
– Scale-up knowledge for unit operations is available.
The six startup preparation measures follow here:

– Potential Problem analysis
– Complete startup team
– Operators trained in startup and normal operation
– Precaution measures
– Startup plan
– Documentation
With these preparation measures, he analyzed ten Shell chemicals startup cases. It
appeared that for cases where the process design success factors were fulfilled and
where all preparation measures were taken the startup time was 5–10 times shorter
than the startup time predicted with the startup time correlation [59].
3.12.9.2 Startup manual

An important element of successful implementation is having an operation manual
and a startup manual. Process startups still fail often due to a lack of proper startup
procedure and lack of operator training [53]. The startup manual includes the precom-
mission period when all individual equipment is cleansed and tested using inert gas
References and further reading | 71
Tab. 3.17: Startup time correlation derived from industrial cases [54].
tstartup = 3.3 + 3.7 N − 3.2 F + S

Parameter Dimension Description
t startup Months Startup time from moment start all feeds till design targets met
N – Number of new process steps
F – Fraction of mass and heat flows that are known
F = 0, if mass and heat flows are not known
F = 1, if mass and heat flows are all precisely known
S Months Solids processing parameter
S = 0, if no solids are processed
S = 0.7, if refined solids such as plastics are produced
S = 10.8, if raw solids are fed to the process
and liquids. Also process controls are tested as much as possible for the limits of the
process conditions.
Independent project analysis has made a statistical correlation for the startup
time of a new process derived from their large database of startups of commercial scale
processes. Most of them concerned solids processing units but also others [54]. The
startup time correlation is shown in Table 3.17.
A process step is new, if it has not been applied before at commercial scale for
that specific input or output. A process step is here defined as a section of the process
such as a reaction section or a purification section with a defined input and output.
A process step will include one or more unit operations.
This startup time correlation can be used to plan the startup and the related mar-
keting and sales planning. It can also be used to justify a certain development effort,
such as having a pilot plant to get the value of F from near 0 to near 1. The difference in
startup time of those two conditions is on average 3.2 months. The amount of money
lost by not having a specification product for that period in general easily outweighs
the cost of a pilot plant.
Further reading on how to make innovation projects successful is provided by
Savelsbergh [60].
[1] Goffin K, Mitchell R. Innovation management, strategy and implementation using the pen-
tathlon framework, 2nd edn, Basingstroke: Palgrave Macmillan, 2010.
[2] Cooper RG. Winning at New Products: Creating Value Through Innovation, New York: Basic
books, 2017.
[3] Arora A. Implications for energy innovation from the chemical industry, NBER, 2010.
[4] Valencia RC. The Future of the Chemical Industry by 2050, Hoboken: J. Wiley, 2013.
[5] Seider WD, et al. Product & Process Design Principles, 3rd edn, Hoboken: John Wiley & Sons,
USA, 2010.
[6] Almeida-Rivera C, Bongers MM, Zondervan E. A Structured Approach for Product-Driven Pro-
cess Synthesis in Foods Manufacture, Computer Aided Chemical Engineering, 39, 2017, 417–
441.
[7] Little AD. Pathway to Innovation; A Global Benchmark study April, 2010. Available from:
http://www.adlittle.com/downloads/tx_adlreports/ADL_InnoEx_Report_2010.pdf. Last re-
sourced 15 November 2010.
[8] Harmsen GJ. Dutch Chemical Technology, Top of the world, lecture for AIChE, section BENELUX,
the Hague: oral presentation, 2006.
[9] Jaber MY. Learning Curves: Theory, Models, and Applications, Michigan: CRC Press, 2017.
[10] Swanson RM. A vision for crystalline silicon photovoltaics. Progress in Photovoltaics, Research
and Applications, 14, 2006, 443–453.
[11] WBCSD, CEO Guide to the Circular Economy, 2017. Available from: http://www.wbcsd.org/
Overview/Resources. Last resourced 10 January 2018.
[12] Chakravorti B. What Businesses Need to Know About Sustainable Development Goals, HBR,
2015.
[13] Cheng B, et al. Corporate Social Responsibility and Access to Finance. Strategic Manage-
ment Journal 2017. Last resourced 10 January 2018. Available from: https://dash.harvard.edu/
handle/1/9887635.
[14] Thomson Reuters. ASSET4 ESG scores on credit views key ESG scores on over 3,200 global
companies, 2017. Available from: https://customers.reuters.com/community/fixedincome/
material/ASSET4ESGSCORES.pdf.
[15] Gorak A. Delft Skyline Debate, Chemical engineering and processing, Special Issue Vol. 51,
2012.
[16] Verloop J. Insight in Innovation, Amsterdam: Elsevier, 2004.
[17] Mehdi Miremadi M, et al. Chemical innovation for the ages, 2017. Available from:
http://www.mckinsey.com/industries/chemicals/our-insights/chemical-innovation-an-
investment-for-the-ages. Last resourced 10 January 2018.
[18] Palucka T. The Wizard of Octane: Eugene Houdry never trained as a chemist but he made the
greatest advance in the history of petroleum chemistry, American Heritage of Invention & Tech-
nology 20(3), 2005, 36–45.
[19] Harmsen GJ. Reactive distillation: the frontrunner of industrial process intensification: a full
review of commercial applications, research, scale-up, design and operation, Chemical Engi-
neering & Processing, Process Intensification, 46(9), 2007, 774–780.
[20] Walpole H. Letter to Horace Mann, 28 January 1754. Available from:
http://www.thefreedictionary.com/serendipity, Last resourced 17 September 2017.
[21] Drent E, et al. Polyketones, Encyclopedia of Polymer Science and Technology, 3, John Wiley and
Sons, 2001.
[22] Synthon, 2017. Available from: http://www.synthon.com/Corporate. Last resourced 10 January
2018.
[23] Bakker HLM, et al. Management of Engineering Projects – People are Key, Nijkerk: NAP, 2014.
[24] Nagji B, Tuff G. Managing your innovation portfolio, HBR, 90(5), 2012, 67–74.
[25] Meesters G. DSM technology platforms, 2017. Available from: https://www.dsm.com/
corporate/about.html. Last resourced 25 January 2017.
[26] Markham SK. Traversing the Valley of Death: A practical guide for corporate innovation leaders,
2014. Available from www.TraversingTheValleyOfDeath.com. Last resourced 10 January 2018.
[27] Cooper RG. Where are all the breakthrough new products?: Using portfolio management to
boost innovation, Research-Technology Management, 56(5), 2013.
[28] Cooper RG. Stage-gate systems: A new tool for managing new products, Business Horizons,
33(3), 1990, 44–54.
[29] Cooper RG, Edgett SJ. Best practices in the idea-to-launch process and its governance, Re-
search Technology Management, Vol. 55, 2012, 43–54.
[30] ConQuastor. Innovative Business Process Management, Utrecht: ConQuastor BV, 2013.
[31] Verganti R. Design-Driven Innovation: Changing the Rules of Competition by Radically Innovat-
ing What Things Mean, Boston, USA: Harvard Business Press, 2009.
[32] Kahneman D. Thinking, Fast and Slow, London: Penguin Books, 2012.
[33] Merrow ED. Industrial Megaprojects, Concepts, Strategies, and Practices for Success, Hobo-
ken: John Wiley & Sons, 2011.
[34] Mankins JC. Technology Readiness Levels: A White Paper, NASA, Office of Space Access and
Technology, Advanced Concepts Office, 1995. Available from: http://fellowships.teiemt.gr/wp-
content/uploads/2016/01/trl.pdf. Last resourced 15 January 2018.
[35] ESA, Technology Readiness Level (TRL) The ESA Science Technology Development Route. Euro-
pean Space Agency, Future Missions Office, Technology Preparation Section. Technology readi-
ness levels (TRL) (PDF). European Commission, G. Technology readiness levels (TRL), HORIZON
2020 – WORK PROGRAMME 2014–2015 General Annexes, Extract from Part 19 – Commission
Decision C(2014)4995.
[36] DOE, Technology Readiness Assessment Guide (DOE G 413.3–4). United States Department of
Energy, Office of Management, 2011.
[37] EC2014 Technology readiness levels (TRL) (PDF). European Commission, G. Technology readi-
ness levels (TRL), HORIZON 2020 – WORK PROGRAMME 2014–2015 General Annexes, Extract
from Part 19 – Commission Decision C(2014)4995.
[38] Graettinger D, Caroline P, et al. Using the Technology Readiness Levels Scale to Support Tech-
nology Management in the DOD’s ATD/STO Environments: A Findings and Recommendations
Report Conducted for Army CECOM (CMU/SEI-2002-SR-027). Carnegie Mellon Software Engi-
neering Institute, 2002.
[39] EARTO, 2014, The TRL Scale as a Research & Innovation Policy Tool, EARTO Recommendations,
2014. Available from: http://www.earto.eu/fileadmin/content/03_Publications/The_TRL_
Scale_as_a_R_I_Policy_Tool_-_EARTO_Recommendations_-_Final.pdf. Last resourced 15 Jan-
uary 2018.
[40] Harmsen J. Novel sustainable industrial processes: from idea to commercial scale implementa-
tion, Green Processing and Synthesis, 3(3), 2014, 189–193.
[41] Nijstad BA, Paulus PB. Group creativity: Innovation through collaboration, New York: Oxford
University Press, 2003, 326–338.
[42] Herrmann D, Felfe J. Effects of Leadership Style, Creativity Technique and Personal Initiative on
Employee Creativity, Brit J Manage, 25, 2014, 209–227.
[43] Rademaker E. 101 gouden tips voor de business case, Teteringen: Van Ierland Uitgeverij, 2007.
[44] Anderson JC, et al. Customer Value Propositions in Business Markets. Harvard Business Re-
view, 84(3), 2006, 90–99.
[45] Vogel GH. Process Development From the Initial Idea to the Chemical Production Plant, Berlin:
Wiley-VCH, 2005.
[46] Yole. Micro fluidic applications, 2015. Available from: http://www.yole.fr/iso_album/
microfluidicapplications_emergingappli_yole__june2015.jpg. Last resourced 15 January 2018.
[47] FluidicMEMS, microfluidics lab on a chip, 2016. Available from: http://fluidicmems.com/list-of-
microfluidics-lab-on-a-chip-and-biomems-companies. Last resourced 15 January 2018.
[48] Sekhon BS, et al. Microfluidics technology for drug discovery and development – an overview,
Int. J. PharmTech Research, 2(1), 2010, 804–809.
[49] Schwalbe T. Microstructured reactor systems, CHIMIA, 56(11), 2002, 636–646.
[50] Boodhoo KVK, Harvey AP (eds). Process Intensification for Green Chemistry, Chichester: John
Wiley & Sons, 2013.
[51] Günther A, Jensen KF. Multiphase microfluidics: from flow characteristics to chemical and ma-
terials synthesis, Lab Chip, 6, 2006, 1487–1503.
[52] Kenig EY, et al. Micro-separation of fluid systems: A state-of-the-art review, Separation and
Purification Technology, 120, 2013, 245–264.
[53] Harmsen J. Industrial Process Scale-up: A practical innovation guide from idea to commercial
implementation, Amsterdam: Elsevier, 2013.
[54] Merrow EW. Estimating start-up times for solids-processing plants, Chem. Eng., 24, 1988,
89–92.
[55] Perry RH. Perry’s Chemical Engineers’ Handbook, 8th edn, New York: McGraw-Hill, 2008.
[56] Helvoort T, et al. Gas To Liquids Historical development of GTL technology in Shell, Amsterdam:
Shell Global Solutions Int, 2014.
[57] De Groen T, et al. 2 × 2 Your choice of projects, twice as cost effective, twice as fast, a guide for
key decision makers in the process industry, Leidenschendam: napDACE, 2002.
[58] Biesboer F. Explosie Shell illustreert falend risicomanagement: onvoorzien scenario, de inge-
nieur, 2016, 50–53.
[59] Sekhon BS, et al. Microfluidics technology for drug discovery and development – an overview,
Int. J. PharmTech Research, 2(1), 2010, 804–809.
[60] Savelsbergh C, Martijn Jong M, Storm P. Project success: creating excellent projects. How can
I make my project successful? Maastricht, AMI 2011. Resourced from: http://ami-consultancy.
com/wp-content/uploads/2016/04/Project-Success-creating-excellent-projects.pdf. Last
resourced 15 January 2018.
4 Designing for innovation
4.1 Introduction
This chapter provides theory about design for innovation, an integrated method for
breakthrough concurrent product-process design, planning for design and experi-
mentation for innovation, setting criteria for design, and shows roles of modeling in
design.
This chapter is meant for industrial product and process designers and PDEng
students to generate novel product-process solutions. The provided thinking modes
and methods facilitate breakthrough concurrent product-process designs.
For graduate students, step-by-step design methods for novel products and pro-
cesses are described in Chapters 6 and 7. Modeling and optimization is treated in Chap-
ters 8 and 9. Evaluating designs is treated in Chapters 10–12.
4.2 Design thinking
4.2.1 Design for innovation theory
Design as a verb, is a problem-solving thinking activity. It starts with assigning a spe-

cific problem and ends with presenting a solution to the client [1]. The design approach
to a problem is to lay out possible solutions and try to improve upon them. The ap-
proach is mainly cycling between two thinking modes, a creativity mode and an anal-
ysis mode. These two steps are repeated until a satisfactory solution is obtained. The
focus is thus on getting a specific solution to a specific problem.
Dorst [2] shows design thinking (and other types of thinking modes) in a scheme,
which is summarized in the Table 4.1.
Abduction-1 and abduction-2 are the types of thinking carried out in the field of en-
gineering design. Understanding these differences help companies to organize their
innovation and prepare an organization environment, so that both designers and sci-
entific researchers can cooperate. It also helps students, first educated in scientific
Tab. 4.1: Types of thinking to solve the unknown (based on [2]).
Type of thinking What (thing) How (working principle) Result

Deduction Known Known Unknown
Induction Known Unknown Known
Abduction-1 Unknown Known Known
Abduction-2 Unknown Unknown Known
https://doi.org/10.1515/9783110467741-004
76 | 4 Designing for innovation
research and subsequently educated in engineering design to understand the large

differences in the modes of thinking they must acquire.
In abduction-1, which may be called normal abduction, the engineering designer
generates intermediate design solutions to the ‘what’, and simulates the result of the
intermediate solution (by using knwowledge of ‘how’) and by adjusting the solutions
of ‘what’ until the desired result is obtained. Simulation can be just a sketch and
visualizing its working principle (‘how’). There are many iterations between solu-
tion generation, simulation, and evaluation [2]. Conventional process design follows
abduction-1 reasoning. The product is specified. Most feeds are specified. The chem-
ical reaction stoichiometry for the conversion of the feed to the product is known.
Working principles for reaction and separations (the ‘how’) can be selected from text-
books. Unit operations can be selected from textbooks, while sizing is determined
using available relations. Chapters 6 and 7 deal systematically with conventional
design for novel products and processes.
In abduction-2 both the what (the desired end solution) and the how (the work-
ing of the end solution are unknown. The what and the how are then obtained by
cycling between what and how to simulate intermediate solutions. Often the inter-
mediate solutions also involve reframing the solution space by thinking first which
problem is hardest to solve, redefining the framework of that problem, and then con-
tinuing to find intermediate solutions. The reframing of the solution space may also
be done several times. This type of design thinking is particularly fruitful for solving
complex problems for which no off-the-shelf solutions exist, and which do not lend
themselves to solutions generated by scientific research because they cannot be as-
signed to a specific discipline.
In concurrent product process design abduction-2 design thinking is applied, as
both the product and its related manufacturing process is unknown at the start. Sec-
tion 4.3 provides methods to help the designers in generating combined solutions.
4.2.1.1 Design and risks

An additional view on what characterizes design is from Petroski [3]. He views en-
gineering design as a succession of hypotheses that such and such an arrangement
of parts will perform a desired function without fail. Petroski stresses that a design
result is always novel and thereby always will have a risk of failure. Petroski also
remarks that implicitly the designer assumes that the design will not fail if it is used
as intended.
The risks and how to reduce them associated with these two aspects of design are
treated in this book explicitly by advocating the testing of novel designs and writing
startup and operation manuals for commercial scale implementation.
4.2 Design thinking | 77
4.2.1.2 Design links with society and nature

The International Council of Societies of Industrial Design (ICSID) defines design (in-
cludes services, processes, and systems) as such: “Design is a creative activity whose
aim is to establish the multifaceted qualities of the objects, processes, services, and
their systems in whole life cycles. Therefore, design is the central factor for cultural
and economic exchange.” [4].
The exchange between society and designers is treated extensively in this book
in Chapter 2 and in Chapter 4 showing the relations between modal aspects of reality,
including society and nature on the one hand and on setting design criteria derived
from these aspects on the other hand.
4.2.1.3 Further reading on design

The field of design thinking is rapidly expanding. Good entries are Dorst [2], Ver-
ganti [4], Verkerk [5] Lawson [6], and Bont [7]. The latter treats designing all kinds of
industrial products and services in complex contexts.
4.2.2 Design knowledge types
Knowledge of designing can be categorized in many ways. The authors found the cat-
egorization of Vincenti fitting with their view on the knowledge of designing. Vincenti
distinguishes the following categories as presented by Verkerk [5] (relevant chapters
in our book are mentioned for each category):
(a) Knowledge of fundamental design concepts; elements a design is made of.
This knowledge is explicitly treated in Section 7.2 on synthesizing product and
process designs, and Chapters 8 and 9 on modeling.
(b) Knowledge of criteria. The kinds of demands a design should meet.
This subject is treated in Section 4.5 of this chapter, in Chapter 6, and in Chap-
ters 10–12.
(c) Knowledge of quantitative data.
In our book, these are called basic design data, and are treated in Chapter 6.
(d) Knowledge of theoretical methods and models.
Chapters 8 and 9 deal with modeling for products and processes in detail.
(e) Knowledge of practical considerations.
This aspect is treated in Sections 4.6 and 4.7.
(f) Knowledge of design approaches.
This main topic is explicitly treated in Chapters 4–12.
Experienced designers have gained large amounts of knowledge in all these cate-
gories. Novice designers do not have the notion of these types of knowledge. Entry
courses on process design treat the knowledge types (a), (b), (c), and (d) explicitly.
Knowledge types (e) and (f) are obtained when designers execute design projects.
4.2.3 Differences between design thinking and scientific research
The main differences between design thinking and scientific research thinking are
found in Table 4.2. The first four items have been obtained from Verganti [4]. The oth-
ers have been obtained from Lawson [6].
Tab. 4.2: Differences between creative design and scientific research (based on [4] and [6]).
Parameter Creative design Scientific research
Number of ideas Create numerous ideas One vision or hypothesis

Peer values Beginner approach to problems Knowledgeable scholar approach
Focus Variety and divergence Convergence on one paradigm
Work culture Neutral: if it works it is good Personal vision
Goal Specific solution General theory
Stopping criterion Good enough Best
Method Many (anything that helps) Hypothesis and experiment
Reliability Judgment and calculation Experimental validation
Essence Prescriptive Descriptive
4.3 Designing for innovative products and related processes
4.3.1 Introduction
This section provides methods for designing both the novel product and the related
novel process effectively and efficiently. It covers all innovation classes and for all in-
novation stages. It has been used by PDEng students at TU Delft and also in industry.
It facilitates team work planning.
4.3.2 Product-process design method: Delft Design Map (DDM)
The presented design method was first defined and developed by Grievink and
Swinkels in the period 2002–2014 [8]. In this period, it was also further refined and
tested by case work inside a consumer product company by Grievink and Almeida [9].
It combines all important elements: the value chain (with subsystems supply chain
and market), product, process, design cycle steps, and innovation stages.
To aid the designer, all these elements have been systematically placed in the Delft
Design Map (DDM); shown in Table 4.3.
The first column in this table lists the design steps from the engineering design
cycle [8] and [9]. These ‘Design (Cycle) Steps’ range from 1 Scope to 7 Report.
The top rows of this table lists 12 ‘Design Levels’ shown horizontally in the DDM
(3rd row, Table 4.3). The design levels are grouped into substages (framing, supply
chain imbedding, process technology, process engineering, and final design):
Framing:
– ‘Project Framing’ (PF)
Supply chain imbedding:

– Customer Wants – Product Concept (CW-PC)
– Product Concept – Property Function (PC-PF)
– Input-Output structure (I-O)
– Subprocesses (SP)
Process technology:
– Task Network TN)
– Unit Network (UN)
– Process Integration (PI)
Process engineering:
– Equipment Design (ED)
– Operability Integration (OI)
– Flowsheet sensitivity and Optimization (FO)
Final:
– Final Design (Final)
The design levels have a clearly ordered sequence that is synchronized with the inno-
vation stages (DDM, first row) defined in the Stage GateTM framework [10].
Each cell in the resulting table stands for a set of focused design activities to be
carried out. A cell is called a ‘Design Activity Space (DAS)’. Each activity space is con-
fined by the design level and design step. The required activities in each space are
thereby relatively limited and straightforward. Advice for the activities to be executed
in each design activity space is indicated with a section reference number of this book.
4.3.3 Explaining the Delft Design Map for product-process design
The DDM table’s purpose is threefold. First the DDM table is used to quickly find infor-
mation on what design activity to perform and how. The second purpose of the table
is facilitating the planning of the work to be done. The third purpose of the table is
to help categorize and check the design result, and whether all activities have been
carried out.
In this paragraph, the engineering ‘design (cycle) steps’ and the ‘design levels’ are
described in further detail. Recall that the design (cycle) steps are carried out for each
Tab. 4.3: Delft Design Map structure (based on [8] and [9]).
Stage gateTM : Concept

Substages: Framing Supply chain imbedding
Design (PF) (CW-PC) (PC-PF) (I-O) (SP)
levels ⇒ Project Consumer Wants Product Input-Output Subprocesses
Framing – Product Concept Concept – structure Subpro-
(product perfor- (product) Products, feeds, cesses,
mance), Property wastes, utilities, action
quality function Function process functions blocks,
technology,
⇓ Design steps I/O
structures
1 Scope 6 6 6 6 6
(of design) ⇓ ⇓ ⇓ ⇓ ⇓
2 Knowledge 6.5 6.5 6.5 6.5 6.5
(about ⇓ ⇓ ⇓ ⇓ ⇓
objects)
3 Synthesize 7 5.2, 5.4, 5.5, 5.2, 5.4, 7.2.3 7.2.3

(alternatives, ⇓ 7.2.2, 8.1–8.3 5.5, 7.2.2, ⇓ ⇓
structure, ⇓ 8.1–8.3
scale) ⇓
4 Analyze 7.3 7.3, 8.4–8.6 7.3, 8.4– 7.3, 9 7.3, 9
(physical ⇓ ⇓ 8.6 ⇓ ⇓
behavior) ⇓
5 Evaluate 7.4 4.5, 7.4, 10–12 7.4, 8.7, 7.4, 10–12 7.4, 10–12
(performance ⇓ ⇓ 10–12 ⇓ ⇓
vs specifica- ⇓
tions)
6 Select 7.5 7.5 7.5–7.6 7.5 7.5–7.6
(alternatives) ⇓ ⇓ ⇓ ⇓ ⇓
7 Report 13 13 13 13 13
(decisions
& results)
design level, and that each combination of a design step and a design level forms a
cell in the DDM, called a Design Activity Space (DAS). The numbers in each cell of the
DDM table refers to a chapter or section in this book.
Tab. 4.3: (continued)
Feasibility Development
Process technology Process engineering Final
(TN) (UN) (PI) (ED) (OI) (FO) (Final)
Task Network Unit Network Process Equipment Operability Flowsheet Final
Allocation, Purification, Integration Design Integration sensitiv- design
targeting, conversions, Energy, Structure, Safety, ity and Product &
mechanism, separations, solvents, water, size, monitoring, Optimiza- process
rates of task assembly emissions, materials, control, tion (PFD)
integration in stabilization production time reliability flexibility,
compartments availability
6 6 6 6 6 6
⇓ ⇓ ⇓ ⇓ ⇓ ⇓
6.5 6.5 6.5 6.5 6.5 6.5 Design
⇓ ⇓ ⇓ ⇓ ⇓ ⇓ data,
heuristics
& models
7.2.3 7.2.3 5.6,5.7,5.8, 7.2.3 7 9 Flowsheet
⇓ ⇓ 7.5 ⇓ ⇓ ⇓ & control
⇓ structure
7.3, 9 7.3, 9 5.8, 9 9 9 9 Mass &

⇓ ⇓ ⇓ ⇓ ⇓ ⇓ energy
yield
factors
7.4, 10–12 7.4, 10–12 7.4, 10–12 7.4, 7.4, 10–12 7.4, SHEETS
⇓ ⇓ ⇓ 10–12 ⇓ 10–12 evaluation
⇓ ⇓
7.5 7.5 7.5–7.6 7.5 7.5 7.5–7.6 Selected

⇓ ⇓ ⇓ ⇓ ⇓ ⇓ design(s),
3.12
13 13 13 13 13 13 Go/
No Go?
4.3.3.1 Design (cycle) steps description:

In the bullet list below the design (cycle) steps are described, with a few examples of
what design aspects and activities are relevant for selected design levels. The design
levels will be elaborated in more detail after the bullet list. Chapter 6 covers items (1)
and (2), Chapter 7 covers items (3–6), and Chapter 13 covers item (7) reporting.
(1) Set the design scope, involving:

– Design goal
– Constraints
(2) Acquire and assess knowledge about potential building blocks (depending on de-
sign level):
– Target market size and dynamics
– Competitor activities/patents
– Raw material supply specifications
– Pure component properties (physical, safety, health data) for components,
and combined streams
– Thermodynamic equilibria, kinetics
(3) Synthesize alternative designs (networks/structures) to meet the design scope by:
– Selecting suitable building blocks/objects at this design level
– Varying order and connectivity of these building blocks
It is important to aim for a quantity of solutions first before moving to step 4,
Analyze, to avoid the too-early elimination of design options that are also valu-
able in combination with other options. Creativity methods and tools discussed
by Tassoul [11], ‘TRIZ’ methodology [12], and synthesis heuristics (Section 4.7.1)
are sources to help create a large number of design options.
(4) Analyze the physical-chemical behavior of generated alternative designs (net-
works):
– Estimates, computations, and/or model simulations connecting resource
flows to building blocks
– Expert analysis by sketches and drawings of product and process block flow
diagrams
The aim of this step is to quantify the design options behavior at the design level
at hand. Aim to arrive at quantitative data like yield estimates mass and heat bal-
ances, separation efficiency, utilities used, etc. as much as possible. These quanti-
tative data are the input into the next design step, evaluate, where the quantitative
data are translated into performance criteria.
(5) Evaluate the performance of alternative solutions against the SHEETS (see Sec-
tion 4.5.3) sustainability goals and constraints:
S: Safety
H: Health
E: Environment
E: Economy
T: Technology
S: Social
For the supply chain related design levels the estimates will be less accurate than
for the estimates at the equipment design level; it is crucial to make every attempt
to quantify the performance of the various design options for each design level, as
it will make clear where design options perform well or not, and where the design
focus should be drawn to in the subsequent design levels.
(6) Select a limited number of preferred solutions from the alternatives for the next
design level.
– Eliminate/weed out unfeasible options.
– Keep promising parts/interesting aspects of eliminated options for implemen-
tation in other options.
(7) Report all design outcomes, drawing, mathematical models, analysis and evalua-
tion results, key decisions, underlying rationales, and numerical data for transfer
to the next design level.
– Write, store, and share activity reports for key design activities at each design
step (see Chapter 13 on communication).
4.3.3.2 Executing design activities in all twelve design levels

In this section design activities for the design levels in the framing, supply chain
imbedding, process technology, process engineering and final substages of the Delft
Design Map will be shown.
1. Design level PF: Project Framing (substage: framing)

Important design activities in the PF-Project Framing design level can be summarized
as follows:
– Frame the product and/or process design task in the overall business innovation
project (‘BOSCARD’: Background, Objectives, Scope, Constraints, Assumptions,
Reporting, Deliverables):
– Background of business context
– Marketing and supply chain aspects and opportunities
· 6P-model (proposition, promotion, position, packaging, price, prod-
uct)
– Design driver (company profile, market pull, technology push)
– Design type (new product family and process platform, platform deriva-
tive, manufacturing improvement, platform development)
– Objectives
– Scope
– Constraints
– Describe the foreseen design approach.
– Design methodology
– Assumptions
– Project organization
– Team members and roles
– Reporting
– Deliverables and milestones

– Project and time planning
– Result: Project Brief
2. Design level CW–PC: Consumer Wants – Product Concept (substage: supply

chain imbedding)
– Identify direct product users.
– Identify stakeholders in product/process life cycle.
– List (qualitative) stakeholder/consumer needs.
– Translate (qualitative) needs into product performance attributes (quality func-
tion) and their quantitative metrics.
– E.g., using Quality Function Deployment (QFD: House of Quality [HoQ])
methodology (see Section 8.3)
– Rank product performance attributes.
– Performance attributes (quantitative) specifications
– Targets and acceptable ranges
– Make relationships explicit between different product attributes (using HoQ
method).
– Identify the key (product) design challenges.
– Result: House of Quality (1): product performance (quality) attributes
mapped onto consumer wants
3. Design level PC–PF: Product Concept – (product) Property Function (substage:

supply chain imbedding)
– Create/synthesize product concepts (with existing or novel product application
processes, aiming to fulfill the product (quality) performance metrics.
– Find thermodynamic phases and physical scales to compose different product
structures.
– Find chemical/biological compositions of these various phases.
– Analyze the behavior of the structured products using mathematical model-
ing and/or experimentation.
– Evaluate product design options against performance metrics.
– For each product design option also review:
– Tolerances on variability of product structure and compositions per grade
– Expected economic life span of product(s) and evolution of its price range
– Overview of suitable alternative feeds to make the product(s)
– Recovery of materials from spent product as secondary feed for manufactur-
ing or in a circular economy
– Result: House of Quality (2): product property function
– Product composition, structure, physical properties – combined with the

product application process, mapped onto the product performance (qual-
ity) attributes
4. Design level I-O: Input-Output structure (substage: ‘supply chain imbedding’)

– Imbedding of the process in a manufacturing supply chain and choice of future
site
– Specification of products and options for cogeneration of chemical products and
energy
– Selection of chemical product grades to be manufactured and available feeds
– Demarcate the process from its environment by setting boundaries (“battery lim-
its”).
– Specification of production modes (campaigns) and operation mode(s)
– Estimation of mass balances based on conversion yield factors
– Identification of possible waste streams
– Estimation of expected economic and physical life span of process plant
– Duties and SHEETS (Safety, Health, Environment, Economy, Technology, Social)
specifications for process plant and external permits
– Result: Input-Output Diagram (incl. feeds, (by)products, solid/liquid/vapor
(S/L/V) waste streams, technologies
5. Design level: subprocesses allocation (substage: supply chain imbedding)

– Decompose the main process into subprocesses (see NOTE) with high level man-
ufacture process functions – for supply chain mapping.
NOTE: these subprocesses should be envisaged as large supply chain blocks (not as a
split of a single product plant) to be located at different locations, and/or run by sepa-
rate legal entities, and/or using a single technology and/or selling their products to more
customers. It is not meant as the split of a single production plant into reaction, sepa-
ration, and polishing steps. Each subprocess will be further designed designing its own
tasks network and unit network design spaces. For maximum process intensification, the
subprocesses should be defined quite broad as is realistically possible.
– Choose the use of key technologies per subprocessing function (licensing).

– Decide on the making, buying, or selling of intermediates or product components.
– Design the connectivity and recycle structure between subprocesses.
– Decide on (re-)use of standard processing equipment for short process life span.
– Or: design subprocesses with innovative processing operations for a long eco-
nomic life.
– Result: Block Diagram with selected subprocesses and processing actions
(and different feeds, (by)products, S/L/V waste streams; including prelimi-
nary mass balances (yield factors) and evaluation regarding gross margin,
transportation cost estimates, etc.
6. Design level TN: Task Network (substage: process technology)

– Task network: allocation, targeting, and integration
– Decomposition of subprocesses into a network of elementary tasks (alloca-
tion)
A task achieves a change in the state of matter from an inlet to a target outlet state, so
demanding a processing duty (targeting). Per task, a (bio) chemical and physical mech-
anism (e.g., reactions, phase transfer) and operating conditions are selected.
– Balancing of rate processes and finding space-time requirements

– Identification of compatible tasks suitable for integration in spatial compartments
– Sizing of compartments, with due consideration of various production modes
– Results: task network(s), represented in a Task Block Diagram
7. Design level UN: Unit Network (substage: process technology)

– Create spatial structures (e.g., cascades) and operating modes (e.g., fed-batch) for
the (task integrated) compartments.
– Integrate compartments in operational units achieving the main processing func-
tions (e.g., conversion, separation).
– Adjust connections and recycle structures between operational units.
– Result: process unit network at mesolevel, represented in a Unit Block Dia-
gram
8. Design level PI: process integration (substage: process technology)

– Efficiently match of internally consumed and generated physical resources by cre-
ating a network of exchanges between process units, involving mass (solvents),
and various forms of energy (thermal, mechanical, power).
– Obtain any remaining unbalanced duties from the utility systems.
– Consider the operability aspects of the resource exchanges networks.
– Deal with ceilings on integrated emissions from the plants (e.g., SO2 , CO2 ).
– Account for multiproduct integration aspects with respect to unit run times and
storage.
– Set targets for the system’s availability as built-up from availability of units.
– Result: Resource exchange networks at the macrolevel of the entire subpro-
cess.
9. Design level ED: Equipment design (substage: process engineering)

– Select and size existing type of equipment, which can host process unit opera-
tions.
– Or: design better performing new equipment for task integrated operational units.
– Spatial structures and space-time requirements fix the main equipment di-
mensions.
– Choose materials to safely accommodate processed materials and conditions.

– Set targets for ‘RAM’ (= Reliability, Availability, Maintainability) for equipment
items.
– Fix multiplicity and overdesign of identical equipment (for size, safety & reliabil-
ity).
– Identify (inequality) constraints confining feasible window of operation of equip-
ment.
– Determine process variables in equipment to be controlled with sensors and ac-
tuators.
– Result: Equipment Specification sheets.
10. Design level OI: Operability integration (substage: process engineering)

– Apply a safety analysis, evaluation and adaptations (e.g., DOW Fire & Explosion
Index).
– Perform flexibility analysis and feasible windows of plant operation.
– Specify key features of an operator – process information interface with time
scales.
– Select instrumentation for process performance & product quality monitoring.
– Develop models to convert measured plant data to process management informa-
tion.
– Design a plant-wide stabilizing control structure per operating mode.
– Check on feasibility of mode switches, start-up & shut-down.
– Determine ‘RAM’ features and run-time per production mode.
– [Optional: design an optimizing control structure for product quality and through-
put.]
– Result: Process Flow Diagram optimized regarding process control.
11. Design level FO: Flowsheet sensitivity and Optimization (substage: Process
engineering)
– Select the key performance indicators (KPIs) of the process plant.
– Identify sensitivities of KPIs to main design variables of process units. Perform an
uncertainty analysis of the design and make necessary adaptations.
– Do a Pareto trade-off optimization of some KP’s for final designs.
– Result: Optimized Process Flow Diagram (PFD)
12. Design level Final: Final design (substage: final)

– Consolidate and compile:
– All design data, heuristics and models
– Product composition and structure
– Process Flow Diagram, control structure, equipment design
– Mass & energy balances, including yield factors

– SHEETS Performance metrics (Safety, Health, Environment, Economy, Tech-
nology, Social)
– SWOT (Strengths, Weaknesses, Opportunities, Threats) of the design and risk
analysis and mitigation suggestions
– Evaluate:
– Project results
– Project design process
– Recommendations for next steps:
– Go/No Go? To Manufacturing & Product Introduction Stage-GateTM phases.
– Write and file Final Report.
– Close design project.
4.3.4 How to plan and execute design activities using the Delft Design Map
The general way of working with the DDM is based on the presumption that the de-
sign results in the preceding design level (N − 1) are available for transfer to design
level (N). Here, the design is further expanded, for which the steps of the engineering
design cycle are executed (again). The design results so obtained in design level (N)
are passed on to design level (N + 1).
For a complete novel product and process design the team would indeed start at
the first design level and carry out the design activities as indicated by the engineering
design steps (see Figure 4.1). Then move to the next design level and repeat the cycle of
design steps again. This can be repeated until all design levels have been covered. This
is indicated with the arrows in Table 4.3. No design process is strictly sequential, so
knowledge and understanding generated in the later design levels can be ‘ploughed
back’ into earlier design levels to improve the design. The circular setup of the design
steps around the design team aims to suggest that steps can be worked on simultane-
ously and iteratively by the team members, rather than in a strictly sequential manner.
As long as the information is reported, classified and shared between the team mem-
bers, the design results can be captured.
Another way of working with the DDM is to first assess the design case at hand,
and then use the DDM to assess what design levels and design step combinations (De-
sign Activity Spaces: cells in the DDM) are crucial at the start of the project. Execute the
design activities, and use the results to plan further navigation through the DDM, es-
pecially when a complete novel product and process design is not required. The DDM
can help the team making a navigation plan and allocate resources accordingly. Yet
a different way is to use one of the predefined ‘fingerprint DDM tables’ relating to a
design case’s innovation class. This will be further explained in Section 4.3.5.
Almeida [9] describes the strengths and weaknesses of the strictly sequential, the
simultaneous, and the ‘anticipating sequential’ approaches. The strictly sequential
Knowledge
of objects
Synthesize
Scope alternatives
objects
& specs
Analyze
Design physical behavior
Design
Level Design
Level
N Level
N+1
N+2
Evaluate
performance
Report Select
decisions best
& results alternatives
Fig. 4.1: Design steps of the generic engineering design cycle. Photo: Creative Commons.
approach is the most intuitive and, therefore, currently most adopted, but it also the
one that needs multiple iteration cycles. Attempts are being made for simultaneous de-
signs of products and processes, but this requires the capability of formulating math-
ematical descriptions of the product’s physical and chemical functions in use and the
manufacturing process to produce the required product structure/composition. For
some cases, this modeling is feasible and Section 4.7 provides a method for this mod-
eling. However, for most product-process cases, modeling solves only a part of the
design problem.
As this field has only recently begun to further develop, the current best bet is the
anticipating sequential approach that can use mathematical optimization at restricted
design levels, but follows the sequential approach to a large extent.
The key aspect is to stress the anticipating sequential approach as a hybrid to
concurrent design of product and process: you do not necessarily move through the
design levels in a sequential way and without recycling. Multiple design levels can
be started, especially when novel product/manufacturing processes are targeted and
new knowledge and insights need to be acquired that potentially and most probably
are affecting each other mutually.
The design levels are merely a demarcation of product and process aspects that
need to be addressed, but the design team needs to maintain overview of the impacts
the different design levels have. Moving back and forth between design levels for anal-
ysis, evaluation, and certainly also synthesis steps is key for the overall best result,
and the Delft Design Map caters very well to this [9].
4.3.4.1 Practical benefits of working with the Delft Design Map

The practical benefits of using the DDM are:
– Sure-footed and faster design processes due to having a clear design structure
– More inventive power by greater awareness of the design levels, design steps, and
design options
– Better options for supply chain integration and process intensification at multiple
scales in the process
– Comprehensive coverage of process integration of all shared resources
– Applicability to processes for structured products
– Better integration of operability aspects into design
– Proper documentation of design decisions and rationales
– Easier reuse of a design for retrofitting and revamps using such documentation
4.3.5 Design planning with Delft Design Map for various innovation classes
Chapter 3 provides a table with innovation classes for products and processes with
their markets and supply chains, including several examples. In this paragraph, we
are linking each innovation class to the design focus that is required. For several in-
novation classes, so-called fingerprint tables are defined using the DDM. The design
activities are classified as ‘key design focus’ (K), ‘optional design focus’ (o), ‘default
design activity’ (D), and ‘check implications of design decisions’ (C). With these fin-
gerprint tables the designer can quickly see where the main design efforts should be
focused.
The fingerprint tables have been derived from observing about 200 PDEng prod-
uct/process innovation design projects carried out for industrial partners in the past
decade by Delft University of Technology (TU Delft).
4.3.5.1 Fingerprints of Delft Design Map for design task planning
Tab. 4.4: Legend belonging to fingerprint Tables 4.5–4.10.
Legend Design levels

K: Key design focus PF: Project Framing UN: Unit Network
o: Optional design focus PC/CW: Product Concept/ PI: Process Integration
Consumer Wants
D: Default design activity PC/PF: Product Concept/ ED: Equipment Design
Property Function
C: Check implications of I-O: Input-Output structure OI: Operability Integration
design changes
_: No activity SP: Subprocesses FO: Flowsheet sensitivity and
Optimization
TN: Task Network Final: Final design
Tab. 4.5: Fingerprint for innovation class 6: radically new

(New market, completely new product, new supply chain, and new process).
Stage gateTM : Concept Feasibility Development

K: Key design focus Supply chain Process Process Final
Framing
o: optional design imbedding technology engineering
focus
D: Default design
activity
C: Check implications
design changes
_: No activity
Design levels ⇒ (PF) (CW- (PC- (I-O) (SP) (TN) (UN) (PI) (ED) (OI) (FO) (Final)
⇓ Design steps/ PC) PF)
activities
1 Scope K D D D D D D D D D D D
2 Knowledge D K K K K K K D D D D D
3 Synthesize K K K K K K K K K K o K
4 Analyze K K K K K K K K K K o K
5 Evaluate K K K K K K K K K K o K
6 Select K K K K K K K K K K o K
7 Report D D D D D D D D D D D D
Tab. 4.6: Fingerprint for innovation class 5: radically new

(Existing market, new product [new platform], new supply chain and new process).

Framing

focus
D: Default design
activity
design changes
_: No activity
activities
2 Knowledge D K K K K K K D D D D D
3 Synthesize K K K K K K K K K K o K
4 Analyze K K K K K K K K K K o K
5 Evaluate K K K K K K K K K K o K
6 Select K K K K K K K K K K o K
Tab. 4.7: Fingerprint for innovation class 4: new

(Existing market, modified product [platform derivative], modified supply chain, and new process).

Framing
focus
D: Default design
activity
design changes
_: No activity
activities
2 Knowledge D D D K K K K D D D D D
3 Synthesize K _ _ K K K K K K K o K
4 Analyze K C C K K K K K K K o K
5 Evaluate K C C K K K K K K K o K
6 Select K _ _ K K K K K K K o K
Tab. 4.8: Fingerprint for innovation class 3: new

(Existing market, existing product, existing supply chain, and new process).

Framing

focus
D: Default design
activity
design changes
_: No activity
activities
2 Knowledge D D D D D K K D D D D D
3 Synthesize K _ _ C C K K C K K o K
4 Analyze K C C C C K K C K K o K
5 Evaluate K C C C C K K C K K o K
6 Select K _ _ C C K K C K K o K
Tab. 4.9: Fingerprint for innovation class 2: novel

(Existing market, existing product, existing supply chain, and modified process).

Framing
focus
D: Default design
activity
design changes
_: No activity
activities
2 Knowledge D D D D D D D D D D K D
3 Synthesize K _ _ _ _ _ _ _ _ _ K K
4 Analyze K C C C C C C C C C K K
5 Evaluate K C C C C C C C C C K K
6 Select K _ _ _ _ _ _ _ _ _ K K
Tab. 4.10: Fingerprint for innovation class 1: conventional

(Existing market, existing product, existing supply chain, and capacity increase process).

Framing

focus
D: Default design
activity
design changes
_: No activity
activities
2 Knowledge D D D K K D D D D D D D
3 Synthesize K _ _ K K _ _ _ _ _ _ K
4 Analyze K C C K K C C C C C C K
5 Evaluate K C C K K C C C C C C K
6 Select K _ _ K K _ _ _ _ _ _ K
This Delft Design Map has been developed over decades with strong interaction be-
tween developments in the global field of product and process design, developments
in industry, and developments in teaching at Delft University. Its complete historic
development is found in Appendix A4.
4.4 Planning for design and experimentation in innovation
Design and experimentation both play important roles in innovation. The major meth-
ods for planning both activities are provided in Chapter 3. The Delft Design Map, how-
ever, can also be used for planning both design and experimentation. In all cases
where knowledge cannot be obtained from written sources, and where analysis can-
not be done by calculations, experimentation will be needed. Validating designs will
also be done by using experimentation.
There is a general efficient approach to planning design and experimentation,
namely by making a design first and listing all assumptions needed for that design.
For all assumptions that are critical to success and for which no written data can be
obtained, an experimental plan is made. If the execution of that experimental plan
is expensive and takes a long time, then the design may be altered in such a way that
the assumption is not needed any more, and the related experimental program is then
avoided. In this way only the most essential experiments are carried out.
This working method is particularly fruitful if the concept designer and the ex-
perimentalist maintain strong interactions. As soon as the experimentalist foresees
expensive long lasting experiments the designer can modify the design and ask again
an estimate of time and effort for the major assumptions to be tested.
4.4.1 Decision sequence ranking method
Information and knowledge from many different knowledge fields are affecting the
design. For strong interactions between many fields of information, a decision ranking
method can be of value. This was proven for an industrial case with eight different
fields of information by Korevaar in 2004 [13]. Each field is judged by its influence
on the performance of the design by the team members with a simple score: strong,
weak, or no. Each field is also scored on active, or passive related to the other fields.
With these scores, the fields are classified as:
(I) High performance influence and highly active
(II) High performance passive
(III) Low performance active
(IV) Low performance passive.
4.5 Embedding design by criteria and context setting | 95
The field I experiments are performed first. If their result is negative the whole design
is stopped and a new design is made.
4.5 Embedding design by criteria and context setting
4.5.1 Introduction to the purpose of criteria and context setting
In designing a product, a process, or any artefact, the designer tries to take all relevant
aspects into account, and in such a way that the new artefact fits well in reality. Only
then the new artefact will be successful. At the start of the design, this can be done by
deriving design criteria from a description of reality that is relevant to the design. This
subject is treated in Section 4.5.2.
4.5.2 Comprehensive list of modal aspects for defining criteria from modal aspects
of reality
The most comprehensive description of reality – in terms that are easy to understand
and that are relevant to design problems – is provided by Verkerk [5]. He provides a
list with 15 modal aspects of reality, shown in Table 4.11.
This table also contains suggestions of what design criteria type can be defined
for each modal aspect for a design (2nd column). Guidelines for this procedure are
provided for each modal aspect hereafter.
Tab. 4.11: Modal aspects and design criteria. Modal aspects from Verkerk [5].
Modal aspect Criteria type Examples

Belief Agreement with belief Halal, Kosher
Moral Moral values Safe, health, sustainable
Juridical Legal Within the laws
Esthetical Esthetic Attractive appearance
Economic Economic Profitable, efficient
Social Social Socially accepted
Linguistic Communication Clear reporting
Shaping Shape Correct 3D shape
Logical Logical Correct model calculations
Sensitive Sense Tasty product
Biotic Biological No patogen contamination
Physical/Chemical Properties and principles Technically feasible
Kinematical Flow rates Heat and mass flows correct
Spatial Geometry Within plot area
Numerical Numerical Mass balances closed
Regarding the modal aspect ‘belief,’ the generation of design criteria needs elaborate
consideration. The governing beliefs in the countries in which the product and the pro-
cess will be present will have to be considered for acceptance. The obvious criterion
for food products and processes for Muslim belief for instance is whether it is halal.
However, for nonfood products where the feed materials for the manufacturing pro-
cess are from animal origins, this criterion also holds. Even the choice of location of
the process can be subject to belief aspects. A process for Liquefied Natural Gas (LNG)
was planned for a location in the north part of Australia. Late in the project, when
its location was publicized, it appeared that the location was considered holy ground
by Aboriginal inhabitants. The location had to be changed due to their protests. As a
consequence, the design also had to be modified, due to the different conditions at the
new location.
Moral values play a strong role in safety, health, environment, social, and sustain-
able development criteria. Most companies have developed specific design criteria for
these aspects, and report on these in their annual sustainability reports.
Social acceptance of new products and processes are of course of enormous im-
portance. If a product is not accepted, there will be no market for it. Social acceptance
is very difficult to predict. In most cases it needs testing.
The linguistic model aspect is about human communication. A design in itself
is information to be communicated to others. Therefore, the linguistic model aspect
is very important. Clear reporting, presenting and obtaining feedback is vital for any
design.
The shaping aspect of design is about the three-dimensional shape of product and
process. The shape of a product is important for its use, transport, and storage. The
shape and outline of the process must fit within the given plot limits.
The logical aspect is about reasoning, modeling, and calculations regarding the
design. Mass and heat balances for instance must be correct.
The sense aspect is particularly important for product design. In most design
cases for each sense, at least one criterion will need to be defined.
The biotic aspect is particularly important for biological and food products
and processes. Contamination by pathogenic micro-organisms for instance must
be avoided. Also for process design biotic criteria are important, for example, in
preventing micro-organism induced corrosion.
Physical and chemical aspects concern all product and process designs. The tech-
nical feasibility of a product or a process will always depend on the physical and chem-
ical properties of the materials selected.
The kinematic aspect is about movement. It is particularly important for flows
inside the process. If the velocity becomes higher than the speed of sound, unwanted
effects may occur.
The spatial aspect is the space taken up by the product and process, and by the
material transport related to the product and process design.
4.6 Role of modeling and simulation in concurrent design | 97
The numerical aspect is involved in all computations made for the design. The
calculations should be based on correct mathematical equations.
4.5.3 SHEETS criteria list
For students to learn to scope a design a limited set of criteria should be used. Often
such a limited set is also used inside companies. This limited set can be seen as the
minimum criteria set for each innovation project and its design. In this book the set is
called SHEETS. It concerns the following criteria types:
– Safety
– Health
– Environmental
– Economic
– Technical
– Social
This list also incorporates the triple P (3P) definition of sustainable development of
People (Social), Planet (Environmental), and Prosperity (Economics). Chapters 10–
12 provide detailed descriptions of these SHEETS criteria and how to use these for
evaluating designs.
4.6 Role of modeling and simulation in concurrent design
Models and simulations play various roles in design. In the discovery and concept
stage it is a fast way of determining the feasibility of an idea or concept. The famous
inventor Tesla could imagine or simulate in his mind the working of a novel concept,
like the dynamo. Only when he could envisage it working, would he construct a proto-
type. Some product developers work in a similar way. They say things like, ‘I envisage
this to be working.’ For novel process concepts, experienced engineers can also imag-
ine the working of a complex process. To the authors’ knowledge, a few guidelines
can be given to facilitate this type of imagination and simulation. One guideline is to
consult an experienced developer about your idea and discuss. A second one is to take
time to visualize the working of your idea using drawings, schematics, or mockups as
presented next.
A more common way of modeling a concept is to produce a drawing, construct
a mathematical model, and (mathematically) simulate the working of the product or
process. In this way problems with the concept surface soon, and modifications can
be made to overcome the problems. This is an important role in the analysis of prelim-
inary design solutions.
In the radical concurrent product-process design, modeling can also play this role.
As the synthesis of a product and its process, concepts are carried out in a concurrent
manner, the mathematical modeling role is more complex. A workable design flow di-
agram involving modeling and simulation of product and process has been generated
by Grievink, Swinkels, and Almeida. A simplified version is shown in Figure 4.2 below
(see also Section 8.5).
Design decision variables Specification of product functions Additional Inputs
molecules, ingredients
phase structure Product synthesis product in use
composition product model generation conditions
Specifications of product & required feeds

flowsheet
structure
Process function model
functional
unit sizes process model generation
physical parameters
flowsheet structure: flow diagram
process steady state & dynamic behaviour
performance optimisation & trade-offs disturbances
performance
specifications
process system states
processing product states
conditions process performance
Product property functions generation

physical parameters
Top-down,
causal flow of
product states product properties
information
external
Product quality & performance functions generation conditions of
product use
product quality factors product performance
Fig. 4.2: Concurrent product-process-concept design flow diagram.
Specifications of product functions are input into the product synthesis step. A prod-
uct model is built and its function is simulated based on external use conditions and
design decisions. The design decision variables are molecule identities, which phase
structure and composition are identified and manipulated. By model simulations the
behavior of the product concept in its use/application is predicted. By trial and er-
ror, the product synthesis and model analysis steps are carried out until a satisfactory
product concept is obtained.
The resulting product composition is then taken as the specification for the prod-
uct the process concept design should deliver. A process concept design, based on var-
ious process functions is then synthesized and analyzed for its behavior. Again, this
4.7 Exploiting experience in design (design heuristics) | 99
is done by simulating with a functional model: a process model. After several itera-
tions of process design modifications and simulation an acceptable process design is
obtained. The design results are the process system states, physical conditions of the
process functions, flows between the function blocks, the process performance, and
the product state.
In the third block (product property function), all (physical and chemical) product
properties are predicted by a model that contains all relevant interactions between the
constituting materials and the product structure.
In the fourth block (product quality and performance functions), the product
quality and product performance in use is predicted. The product state, (physical)
product properties, and the external conditions for using the product are the inputs
for predicting the product quality factors and product performance.
Chapter 8 deals with product modeling, simulation, and optimization in detail.
Chapter 9 deals with process modeling. The connectivity between product and process
modeling and simulation will be covered in Section 8.5.
4.7 Exploiting experience in design (design heuristics)
4.7.1 Strength of design heuristics
Designing using heuristics is enormously fast. Particularly in the concept stage, this
facilitates making several designs quickly, and selecting the best. The resulting con-
cept design will also be more reliable than those made by using theory only.
Experience can be exploited for design in the following ways:
(a) By using rules of thumb (design heuristics)
(b) By asking experienced people for advice and feedback
(c) By going back in your memory on previous cases and deriving knowledge to use
for the case at hand
(a): Design heuristics is coded knowledge obtained from many experienced design-
ers. Many heuristics for process concept design are provided by Seider [14] and Dou-
glas [15]. For process design in all innovation stages, Koolen is also a very good re-
source [16].
Heuristics for product design are scarce. Most of this type of knowledge is avail-
able within product innovation companies, and strongly context bound: to product
design in foods, cosmetics, pharmaceuticals, etc. Martin provides heuristics for each
of these product types [17].
(b): Obtain advice and feedback from experienced people. They have a lot of knowl-
edge from experience and insight, but this is not easily transferred to others. This
knowledge is called tacit knowledge. By stating the design solution, or a design prob-
lem, and asking them to go back in their own memory, give them time (can be minutes,
can be days) on similar cases and probe them for stories during conversations or reflec-
tion on decisions. Through this process, a lot of useful information can be obtained.
An example from Harmsen’s own experience is presented here. I had made a
trickle-flow fixed bed reactor design and showed it to an experienced engineer. He
said: “What is the bulk crushing strength of the catalyst and is it below the maximum
pressure drop that could occur when the bed is in operation, and when dust fouling
has happened on the bed?” I looked up the bulk crushing strength of the catalyst par-
ticles, and it appeared to be even below the pressure drop of the empty bed. I needed
to redesign the reactor and had to select a larger catalyst particle size to further reduce
the pressure drop.
(c): Consult your own experiences. By asking yourself the same question as you would
ask experienced people, you will find that you have more knowledge then you imag-
ined you had.
If while designing using heuristics, a conflict of rules occurs, then rethink the
whole design in such a way that the conflict disappears. Philosophically this is the
same as solving the conflict of a thesis and an antithesis, by moving the whole prob-
lem to a different level, or different context by which synthesis is possible.
4.7.2 Weaknesses in using design heuristics
Heuristics are based on truths related to a historically fixed context (sometimes called
orthodoxies). Some heuristics for instance are based on the context that electricity is
an expensive energy form. If in the future electricity becomes very cheap (albeit for
certain periods during the day), caused by excess wind energy produced, solar cells
becoming very cheap, etc., then these heuristics have become invalid.
4.7.3 Tapping into experience
Experiences shape our intuition. The intuition is used for practical purposes including
design purposes. It is a very fast and a very effective form of thinking for all kinds of
design activities. This is the reason you see so many experienced employees in design
departments.
The intuitive form of thinking has one disadvantage: it blocks the individual
pieces of memory about specific historic cases. The individual learning points of his-
toric cases, however, can also be of interest to the design at hand. Kahneman [18]
provides a simple method of drilling through the intuition barrier. His method is: Ask
the experienced person (may be yourself), to go back into his own past and “envision”
the cases that were a success, and those that were failures. The person will fall silent
for a minute or more, as he goes back in his memory. Then ask the question, what
is the learning point relevant to the design case at hand. With this method, drilling
through the intuition barrier is easy. Kahneman calls this the outside view, as it gives
us views beyond the barrier of intuition.
Heuristics for breakthrough products and their processes are not publicly avail-
able. Heuristics for product design and for process design are provided in Chapter 7.
4.8 Industrial example of design driving innovation
The prime example of a design driven breakthrough process innovation is the counter-
current extractive reactive distillation for the production of methyl acetate by Eastman
Chemical. Agreda [19] describes this innovation in detail with all steps from concept
generation to commercial scale implementation.
The implemented design is a one-column process containing two countercurrent
extractions, a reactive distillation, and distillation sections at the top and bottom. It is
dynamically controlled allowing for stoichiometric feeds of the two chemicals. At that
time there was no equivalent concept of a complex combination of phenomena in a
single column. Even simple reactive distillation was just being researched at the time
of conception.
The following steps were taken in this design driven research, development, and
implementation.
Step 1: A conventional design with classic unit operations would consist of two
reactors and eight distillation columns. This design was tested at pilot plant scale.
Because of its complexity, it was deemed desirable to search for a more economical
process, requiring far less unit operations, hence; a challenging design goal was as-
signed.
Step 2: Many ideas were considered and computer simulations were used to test
the ideas. After close examination of the problems of azeotropic distillations and the
chemical equilibrium reaction, the concept was generated to use countercurrent ex-
tractions to avoid azeotropic distillations and to use reactive distillation to avoid reac-
tion equilibrium limitations and to carry this all out in a single column. The top and
bottom part of the column were designed as classic distillations to purify the product
streams. An economic analysis revealed that the process was sufficiently economically
attractive to pursue and laboratory testing was recommended.
Step 3: Laboratory experiments were carried out to generate kinetics and phys-
ical properties. The feasibility of the concept was also tested in a laboratory setup.
These tests also confirmed the simulation calculations. The liquid holdup for reaction
appeared to be a very critical factor for success. New economic analysis showed favor-
able results and large scale testing was recommended.
Step 4: A bench scale test glass column was built with a diameter of 0.1 m and a
height of 9 m. With this column in operation, corrosion tests were carried out as well
as detailed measurements of component concentration profiles over the height of the
column.
Step 5: A commercial scale design was made using the data from the bench-scale
tests. Steady state and dynamic computer simulation programs were developed to
generate data for the design of the column. Again an economic analysis was made.
The results reinforced the expectation that it could be a viable commercial process. A
down-scaled pilot plant was then designed, constructed, and tested. Its dimensions
were a diameter of 0.2 m and a height of 30 m. Special precaution was taken to ensure
adiabatic operation. A dynamic control was also installed to allow stoichiometric feed
ratio as intended for the commercial scale. Again corrosion tests were also carried out.
The results proved that steady state production of methyl acetate of the required pu-
rity could be made over a significant period of time. Again an economic analysis was
made of the commercial scale process.
Step 6: The decision was taken to make a detailed design, to construct and to op-
erate the commercial scale process. The commercial scale process contained an ad-
ditional methanol recovery column to allow for a lower temperature bottom flow re-
ducing potential corrosion and to minimize chemical losses during upsets of the op-
eration. Additional measures were taking concerning sufficient liquid holdup on the
reaction trays.
Step 7: Operation procedures were written with special attention to startup. Plant
commissioning started with inspection and water checkout. Many problems were
identified and corrected at this stage.
The plant started up in May 1983. Many problems including froth related problems
had to be solved in the startup phase. However, no production losses were incurred.
Agreda attributes this to the pilot plant testing. The plant now operates routinely at
rates above 100% of the design. The design production rate is 22,700 kg/h. It is about
five times lower in investment cost and energy cost compared to the conventional pro-
cess setup with 11 unit operations [19].
It is clear from his description that at the beginning it was not known how the
process would work, nor what the process concept would be. Suitable simulation tools
were not even available, but were developed during the research and development
stages. Also a novel control method had to be developed. This innovation would not
have happened if a design approach had not been followed from the start and through
all subsequent stages to commercial implementation. It illustrates the power of design
driven design. The case also shows the power of the stage-gate approach and having
the commercial scale solution always in mind, and from there, designing whatever
research and development is needed.
This design driven success led Siirola, the principal design engineer of Eastman
Chemical, to develop a process synthesis method based on tasks (functions) rather
than classic unit operations [20]. It is a nice illustration where first, design solutions
are generated and then through reflection, a general design method is derived. Or as
Henderson put it: “Science owes more to the steam engine, than the steam engine
owes to science” [21].
Abbreviations
6P Proposition, Promotion, Position, Packaging, Price, Product
BOSCARD Background, Objectives, Scope, Constraints, Assumptions, Reporting,
Deliverables
C Check implications of design changes
D Default design focus
DDM Delft Design Map
ED Equipment Design
FO Flowsheet sensitivity & Optimization
HoQ House of Quality
I-O Input–Output
ICSID International Council of Societies of Industrial Design
K Key design focus
LNG Liquid Natural Gas
o optional design focus
OI Operability Integration
PC-CW Product Concept – Customer Wants
PC-PF Product Concept – (Product) Property Function
PI Process Integration
PFD Process Flow Diagram
RAM Reliability, Availability, Maintainability
QFD Quality Function Deployment
SHEETS Safety, Health, Environment, Economy, Technology, Social (aspect)
S/L/V Solids/Liquids/Vapors
SP Sub Processes
SWOT Strengths, Weaknesses, Opportunities, Threats
TN Task Network
Triple P People, Planet, Profit
TRIZ Theory of Inventive Problem Solving – (Russian acronym)
UN Unit Network
[1] Dorst K. Understanding Design, 175 reflections on being a designer, Amsterdam, Netherlands:
BIS Publishers, 2006.
[2] Dorst K. Frame Innovation: Create new thinking by design, Cambridge MA, USA: MIT Press,
2015.
[3] Petroski H. To engineer is human: The role of failure in successful design, New York NY, USA:
Vintage books, 1992.
[4] Verganti R. Design-Driven Innovation: Changing the Rules of Competition by Radically Innovat-
ing what Things Mean, Boston MA, USA: Harvard Business Press, 2009, 4.
[5] Verkerk MJ, et al. Philosophy of Technology, New York, USA: Routledge, 2016.
[6] Lawson B. How designers think – The design process demystified, 4th edn, Oxford USA: Else-
vier, 2006.
[7] De Bont C. Advanced design methods for successful innovation, Den Haag, Netherlands: De-
sign United, 2013.
[8] Grievink J, Swinkels PLJ, Van Ommen JR. Basics of Process Design. In: De Jong W, Van Om-
men JR (eds). Biomass as a Sustainable Energy Source for the Future: Fundamentals of Con-
version Processes. 1st edn, Hoboken NJ, USA, Wiley & Sons Inc., 2014, 184–229.
[9] Almeida-Rivera C, Bongers P, Zondervan E. A structured approach for product-driven process
synthesis in foods manufacture. In: Martin M, Eden MR, Chemmangattuvalappil NG (eds). Com-
puter Aided Chemical Engineering Vol 39: Tools for chemical product design – from consumer
products to biomedicine. Amsterdam, Netherlands: Elsevier BV, 2017, 417–441.
[10] Seider WD, et al. Product & Process Design Principles, 3rd edn, Hoboken NJ, USA: John Wiley &
Sons Inc., 2010.
[11] Tassoul M. Creative facilitation. Delft, Netherlands: VSSD, 2009.
[12] TRIZ – a powerful methodology for creative problem solving, 2017. Resourced from:
http://www.mindtools.com, https://www.mindtools.com/pages/article/newCT_92.htm.
[13] Korevaar G. Sustainable Chemical Processes and Products: New Design Methodology and
Design Tools, Delft, Netherlands: Eburon, 2004.
[14] Seider WD, et al. Product & Process Design Principles, 3rd edn, Hoboken NJ, USA: John Wiley &
Sons Inc., 2010.
[15] Douglas JM. Conceptual design of chemical processes. New York NY, USA: McGraw-Hill, 1989.
[16] Koolen J. Design of Simple and Robust Process Plants, Weinheim, Germany: Wiley-VCH, 2001.
[17] Martin M, Eden MR, Chemmangattuvalappil NG, eds. Computer Aided Chemical Engineering
Vol 39: Tools for chemical product design – from consumer products to biomedicine. Amster-
dam, Netherlands: Elsevier BV, 2017.
[18] Kahneman D. Thinking, Fast and Slow, London, UK: Penguin Books, 2012.
[19] Agreda VH, Partin LR, Heise WH. High-purity methyl acetate via reactive distillation, Chemical
Engineering Progress, 1990, 86(2), 40–46.
[20] Siirola JJ. Industrial applications of chemical process synthesis, In: Anderson JL, Bischoff KB
(eds). Advances in Chemical Engineering Vol 23: Process Synthesis. Academic Press, 1996,
1–92.
[21] Henderson LJ. Quote in Moore, WJ. Physical Chemistry, 4th edn, London, UK: Longmans Green
and Co., 1962.
|
Part B: Design generation
5 Special design approaches
5.1 Introduction to special design approaches
This chapter is about special design approaches. Each approach has a specific lens
through which the problem and its solution are looked at. The purpose of the chapter
is to enhance the creativity of the designer by this special view on the problem and
to improve the quality of the design solution. Each approach has its specific solution
methods.
Some of the approaches, such as design for biomimicry, design for industrial ecol-
ogy, and design for renewable energy sources are very suited for the discovery and
concept stage. Other approaches such as design for energy efficiency, design for hu-
man factors, and design for six sigma are very suitable for the feasibility and detailed
design stage.
5.2 Design for biomimicry
The term ‘biomimicry’ appeared for the first time in 1981 and was popularized by sci-
entist and author Janine Benyus in her book Biomimicry: Innovation Inspired by Na-
ture [1]. She defines biomimicry as a “new science that studies nature’s models and
then imitates or takes inspiration from these designs and processes to solve human
problems.” Since then an array of articles and books have appeared, in which the idea
is further developed. Here are the basic ways to apply biomimicry.
The simplest way is to take a natural material with a very special property and
use its specific property such as strength, hardness, or chemical reactivity and design
a product with the same material for an application. The industrial process for making
the material may also be copied from the way nature makes it.
A second way is to take a physical principle from nature and use that principle to
make a novel product. An example is the use of a nanostructure that reflects sunlight
in such a way that the material shows a specific color. The same nanostructure, but
with different materials can then be designed and made to provide color on the surface
of the material. In this way the need to paint is avoided.
A third way to apply biomimicry is to take a natural system’s method and use the
same method for an industrial system. An example is the natural system for closing
the material cycles for carbon, oxygen, and water. Through photosynthesis, carbohy-
drates are formed from carbon dioxide and water, which in turn are used by other nat-
ural species as an energy source by oxidizing them back to carbon dioxide and water.
This third way has been developed into industrial ecology and its subsystem, indus-
trial symbiosis. This way is treated separately in this book in the section on industrial
ecology.
https://doi.org/10.1515/9783110467741-005
108 | 5 Special design approaches
A fourth way to apply biomimicry is to be inspired by nature in the way it has

solved problems. Perhaps this is the most fruitful method. It has the advantage that it
is not rigidly held to the specific natural method. Such rigidity can lock in the design-
ers to solving the problem in a fixed natural way. A classic example is the idea that a
flying machine should be flying by moving wings such as birds do. This led to many
failures. Only by leaving this idea and separating the propulsion from the airlift did
flying machines become a success.
The inspiration to improve the airplane wings for energy efficiency by looking at
the design of bird wings, however, is still applied by wing designers.
For an elaborate description of biomimicry for product and process design and
other usages, the reader is directed to references[1–3].
5.3 Design for industrial ecology
Industrial ecology is the study of the flows of materials and energy in industrial and
consumer activities, of the effects of these flows on the environment, and of the in-
fluences of economic, political, regulatory, and social factors on the flow, use, and
transformation of resources. Core elements are [4]:
– The biological analogy
– The use of system perspectives
– The role of technology change
– The role of companies
– Dematerialization and ecoefficiency
– Forward looking research and practice
Unfortunately, this handbook provides little information of use for designers. To my

knowledge there is also no other book on designing for industrial ecology available.
5.4 Design for circular economy
The widely accepted definition of a circular economy is the one by the Ellen MacArthur
Foundation: “a circular economy is one that is restorative and regenerative by design,
and which aims to keep products, components and materials at their highest utility
and value at all times, distinguishing between technical and biological cycles.” [6]
Figure 2.3 of Chapter 2 shows the two closed mass cycles and their mostly applied
process steps. This figure can be used by designers for inspiration to design for closed
mass cycles for the entire life cycles of products and processes.
Cramer has developed a practical set of ten guidelines. She calls it the 10 R Ladder
of Circularity [8]:
5.6 Design for industrial symbiosis | 109
Refuse: Do not use native material, but use refused material.

Reduce: Use less material.
Renew: Use waste material in a new way.
Reuse: Use waste material again.
Repair: Repair the device or material.
Refurbish: Revive the material.
Remanufacture: Make a new product from used materials.
Repurpose Reuse product for other function.
Recycle: Recycle material for use.
Recover Recover some value from used material such as energy.
The ladder should be started at the top. If a certain step has been tested and cannot
be further applied, a lower step should be tested.
As the used material often has a zero or even negative price, while the new product
has a higher market price, profits can often be foreseen. In general, startup companies
exploit this potential earning power, as shown in her webinar [8]. Cramer estimates
that in the Netherlands alone, 7 billion €/year can be made by circular economies,
just considering water, energy, and materials [7].
5.5 Design for cradle-to-cradle
Braungart’s cradle-to-cradle design method for ecoeffectiveness is based on three es-

sential elements [10, 11]:
(A) Design for recycling all product components staying within the technology cycle
(see Section 5.4).
(B) Design for end-of-cycle products obtained from biological origins and ending in
the biosphere, such that all decomposed components end up in the environment
after their use, acting as nutrients for plants to grow.
(C) Use solar as the sole primary energy source.
His method is used for novel consumer product design and focuses on protecting the
environment.
5.6 Design for industrial symbiosis
A subset of industrial ecology is industrial symbiosis. It is concerned with transferring

the biology analogy of the ecosystem to the industrial system. Chertow investigated
22 industrial symbiosis cases and concluded that environmentally and economically
desirable symbiotic exchanges are all around us and now we must shift our gaze to
find and foster them [12].
Table 5.1 provides an overview of the main elements of industrial symbiosis [13].
Tab. 5.1: Industrial symbiosis elements derived from nature.
Nature ecology system elements Derived industrial symbiosis elements

Food chain and closed material recycles Closed material cycles
Many different species connected in network Industry and society parts connected in network
Primary energy from sun Solar energy based
Energy stored and transported via food Energy in intermediate materials
The designer can use the Dow Chemical method finding and implement industrial
symbiosis synergies [14]. It contains the following steps:
Step 1: Perform a high-level feasibility study of potential cases with environmental,
social, and economic benefits and select cases with positive outcomes based
on these three criteria.
Step 2: Research material properties and potential applications for those cases.
Step 3: Locate potential suppliers and users of streams and define potential connec-
tions.
Step 4: Obtain interest from partners in the project both from technical and economic
points of view.
Step 5: Finalize project.
Step 6: Monitor and improve project.
Magan describes an actual industrial symbiosis case (which he calls a byproduct syn-
ergy case) and from which he derives following steps [14]
Step 1: Organize facilitated meetings between potential partners of an industrial site
and explain the concept and the potential benefits. The meeting should be
chaired by an independent person to avoid perceived bias for a certain partner.
Step 2: Form working groups of engineers from potential partners to analyze options
and prioritize best cases.
Step 3: Overcome barriers, such as legal, technical, economic, communication, com-
peting priorities, geographic, and perception challenges by maintaining com-
munication and cooperation within the working groups.
Step 4: Make action plans.
In Magan’s experience, steps 1–4 can be carried out within a year. After that first
year quarterly meetings should be held to implement action plans and bring in new
companies [15].
Synonyms used for industrial symbiosis are:
– Byproduct synergy
– Eco-industrial development
– Eco-industrial parks
– Industrial ecology
– Industrial ecosystems
5.7 Design for renewable energy sources | 111
5.7 Design for renewable energy sources
5.7.1 Present bulk chemicals production from fossil fuel resources
Present industrial processes are still mainly driven by fossil sources. This is mainly
performed by burning fossil energy resources such as natural gas in furnaces and us-
ing the heat in the form of steam or a hot medium to provide the energy for reaction,
separation, extrusion, and mass movement.
For the formation of hydrocarbon products and their derivatives, the reaction en-
ergy is provided as follows. First the hydrocarbons are transformed into olefins by the
steam cracker process. This is a highly endothermic process where the required reac-
tion energy is provided by burning fossil hydrocarbons in a furnace. By placing the
reactor tubes in that furnace, the reaction heat is directly provided to the tubes.
The resulting olefins are then used in subsequent reactions to produced plastics,
and virtually all other chemicals. All these latter reactions are exothermic, so they then
produce the product, and heat. That heat stems from the energy stored in the olefins.
The advantage of exothermic reactions is that the reactions run to complete conversion
at moderate temperatures as the main way to obtain a negative free enthalpy is the
negative reaction enthalpy.
5.7.2 Design for renewable energy based on product-process chain methods
When renewable energy sources are used to produce chemicals and other products,
ways must also be found to deliver energy for driving the reactions and for driving the
other steps in the process.
Renewable energy is provided from the sun, directly or indirectly. Solar radiation
is abundantly available but at a low energy density. It is so low that it must be con-
centrated first to be applied in industrial scale processes. A proven way is to use pho-
tovoltaic cells to convert solar radiation into electricity that can be transported and
concentrated.
Another way is using Concentrated Solar Power (CSP) technology. Using mirrors,
solar radiation is concentrated and can then be used to generate a very high tempera-
ture that can be transferred to a medium such as molten salt, which then can be used
to drive a chemical reaction or converted subsequently to electricity.
Wind mills are also used to harvest solar radiation and convert it into electricity.
Yet another way is to grow biomass, which then in turn can be used as a renewable
energy source. It can also be directly used as feedstock for chemical production.
In all cases the renewable energy should be used efficiently and at low investment
cost for the production of chemicals and other products. This requires new innovative
solutions. An example of such a solution is the production of ethylene glycol from
carbon dioxide, water, and solar radiation described in Dautzenberg’s lecture [16].
The first technology is electricity from Concentrated Solar Power (CSP). Already
99 of these are in commercial operation and 28 were under construction in 2017.
The second technology is high-pressure water electrolysis. This technology pro-
duces hydrogen and oxygen from water at low cost, as no compressor is needed. Sev-
eral Japanese patents have been published.
The third technology is the electrochemical catalysis of carbon dioxide and hydro-
gen to ethylene glycol. The company Liquid Light has developed this technology [17].
The combination of these three technologies means that from the inputs carbon
dioxide, water, and solar radiation, ethylene glycol can be made. As the primary feed-
stocks have little to no cost it is likely that this ethylene glycol production could be
competitive with fossil based production.
It is clear from this example that renewables based products and processes can be
designed and developed cost effectively, but they require a lot of considerations, such
as the basic elements of solar radiation, electrolysis, and catalysis. The latter two may
be combined.
Murzin provides process design methods for products from renewables [18].
5.8 Design for energy efficiency
‘Energy efficiency’ is a generic qualification that describes the use of a minimum of ex-
ternal energy resources to obtain a desired service or an amount of product. For chem-
ical products and processes the common energy resource is power, any form of fuel or
utilities that is derived from these two main resources. Their use is generally related to
the amount of product produced or feed processed to get the ‘process specific energy’
(in J/kg), as shown in Figure 5.1. Aiming for energy efficiency, a low process specific en-
ergy minimizes both the environmental footprint (use of resources and CO2 -emission)
and the operating cost to obtain the required external energy resources. For products,
the energy efficiency of the intended service of the product may also be relevant, like
fuel savings by low rolling resistance tires [19]. This section focuses on processes and
it gives a brief introduction how to assess their energy efficiency and how to improve
it.
Fig. 5.1: Generic process input-output

scheme.
Fig. 5.2: Example Sankey diagram: sugar extraction (Courtesy Cosun R&D).
Energy efficiency can be achieved in any form of energy that plays a role in the pro-
cess to produce the desired product. Changes in composition and the physical state of
material streams are achieved by chemical and physical processes. All of these cause
changes in energy states and as we deal with practical processes, they have their spe-
cific irreversibilities. The most energy efficient process is the reversible process. Ther-
modynamic approaches are available to determine the minimum energy resources re-
quired to perform the desired process in a reversible way. The ultimate low process
specific energy found this way is the ultimate target to aim for in energy efficient de-
sign. This is the basis for ‘exergy analysis’ and design approaches to minimize the
exergy loss, also referred to as lost work [20] [21]. Exergy losses in a process can nicely
be visualized in a Sankey diagram (Figure 5.2), which shows the exergy supply to the
process, the losses in each operation within the process, and the exergy sinks in prod-
ucts, waste streams, and losses to the environment [22].
The exergy analysis gives useful insight in the ‘irreversibilities’ that are impor-
tant in a process that is available. This may guide changes to the process to make
it more reversible and improve the energy efficiency. It relates, however, energy ef-
ficiency to an ideal and unreachable target. For designers it is generally more relevant
to get guidance about what is practically feasible. One source of practical guidance is
the current state of the art: existing operating plants have a proven process specific
energy. Values for known processes are available in the literature. They may also be
available from governmental and consultancy sources that monitor the efficiency of
specific processes and industrial sectors [23]. Typical process specific energy values
can be used as alternative targets for new facilities, but these values should be treated
as the maxima that can be allowed for a new process, as it will have to compete with
the existing ones.
Traditional (petro) chemical bulk industries have been keen on energy efficiency
especially since the 1970s. For these processes the main energy transformation is re-
lated to physical heating and cooling. This has resulted in a practical tool kit to analyze
such processes, determine practical energy efficiency targets, and design the heat ex-
changer network, the system that performs the required physical heating and cooling.
This tool kit is known as Pinch Analysis [24] and is the main commonly used analysis
and design method for energy efficiency in the industrial practice. We will illustrate
the approach using a simple example case.
Example 5.8.1 (Example of reactor heating and cooling). Consider the scheme of Fig-
ure 5.3, with the stream data given in Table 5.1. Cold feed S1 is heated up and fed to a re-
actor. The reactor product S2 is cooled down. Stream S1 we refer to as a ‘cold stream’ as
it needs to be heated up. Stream S2 we refer to as a ‘hot stream’ as it needs to be cooled
down. Exchanger HX1 has a duty of 5 kg/s × 2 kJ/kg/°C × (100 °C − 50 °C) = 500 kW,
HX2 a duty of 5 kg/s × 1.6 kJ/kg/°C × (105 °C − 45 °C) = 520 kW. Both the heating
and cooling duty may be supplied by an external source: steam to heat S1 and cooling
water to cool S2. The process specific energy can now be defined as:
steam duty supplied 500 kW 100 kJ
process specific energy = = = (5.1)
product flow rate 5 kg/s kg product
Obviously, it should be possible to make this process more energy efficient and lower
the process specific energy. If we look at the overall heat balance we find a heating
requirement of 500 kW and a cooling requirement of 520 kW. The net result suggests
a net cooling requirement of 20 kW. No heating would be required. Clearly, this ap-
proach is too simple and only based on the first law of thermodynamics (heat con-
servation), while it ignores the second law of thermodynamics. This second law re-
Fig. 5.3: Simple reactor system with preheat and

aftercooling.
Tab. 5.2: Example 5.8.1 stream data.
Stream Source Target Mass flow Heat capacity

°C °C kg/s kJ/kg/°C
S1 50 100 5 2
S2 105 40 5 1.6
Fig. 5.4: Simple reactor system with feed-effluent heat

exchange.
quires that for any feasible physical heat transfer the supply heat has to have equal
or higher temperature than the demand heat. For a reversible heat transfer the sup-
ply (hot) and demand (cold) temperatures are the same, but in practice the exchanger
area has to be limited in size which produces the requirement of a minimum temper-
ature difference (well) above zero. This becomes clear from the generic heat transfer
relation: A = Q/U/DT, in which A is the heat transfer area (m2 ), Q the duty (kW), U is
the overall heat transfer coefficient (kW/m2 /°C) and DT is the effective temperature
difference (°C) that is equal to the logarithmic mean temperature difference for a pure
countercurrent heat exchange [25].
The reactor effluent is a little hotter than the reactor feed and can be used to heat
this feed. If we try to supply all the heat (500 kW) required in Stream S1, as illustrated
in Figure 5.4, it becomes clear that Stream S2 is inadequate to do that. After cooling
500 kW, S2 is cooled to 42.5 °C, which is below the supply temperature of the cold
feed S1. Apparently, there is a limit to the heat transfer from the hot product to the
cold feed.
To find this limit we draw the hot supply and cold demand in an enthalpy – tem-
perature plot, as illustrated in Figure 5.5. In this plot we can graphically construct
the feasible overlap where the hot (supply) curve is hotter (higher temperature) than
the cold (demand) curve. When we allow the curves to touch, we allow the thermody-
namic minimum approach of 0 °C and we get the ultimate heat transfer that is feasible.
The resulting graph shows the following:
– It is possible to transfer 440 kW from the hot product to the cold feed;
– We will need 60 kW additional heating for the cold feed to reach the target reactor
inlet temperature;
– We will need 80 kW additional cooling for the hot product to reach the target prod-
uct temperature.
This is the best thermodynamic solution for the given system but as the temperature
approach is 0 °C it will require an infinite transfer area. For any practical design it may
be used as a reference. It will be necessary anyway to have three exchangers as shown
in the scheme depicted in Figure 5.6.
For a practical design the temperature approach should be selected to allow the
application of real heat exchangers. Choose for example a minimum temperature dif-
ference of 10 °C. Note that a realistic value depends on the exchanger type that can be
applied; for the more common types the minimum difference is 10 to 20 °C, but more
Fig. 5.5: Enthalpy–temperature plot Example 5.8.1 with minimum temperature difference of 0 °C.
Fig. 5.6: Heat-integrated reactor system.
advanced exchangers allow a minimum difference as low as 1 °C. With this selected
value of 10 °C we can do a similar graphical construction, see Figure 5.7, to determine
the heat that can be transferred in the enthalpy – temperature plot. The resulting hot
to cold process stream transfer is 360 kW, the heating requirement 140 kW and cooling
requirement 160 kW.
Obviously, we can do this exercise for any other minimum temperature difference.
If we increase the temperature difference we will reduce the process-to-process heat
transfer and increase the need for additional heating and cooling. This will make the
process less energy efficient. Additionally, we will need less transfer area for process-
to-process heat exchange (S2 to S1) as the duty goes down and the temperature differ-
ences go up. Meanwhile, the transfer area for the additional heating and cooling (by
utilities) will increase as these duties go up. The net effect is case dependent.
Fig. 5.7: Enthalpy–temperature plot with minimum temperature difference of 10 °C Example 5.8.1.
Example 5.8.1 shows that a heat curve plot can be used to determine realistic targets
for process-to-process heat exchange and the need for external heating and cooling re-
sources. Pinch Analysis is an extension of this simple analysis for processes with any
number of hot and cold streams. To make this pinch analysis all cold process streams
are lumped in a cold composite stream. Plotted in a temperature – enthalpy plot this
gives the ‘cold composite curve,’ Figure 5.8, which shows the heat that is required in
the process at each temperature. In the same way all hot streams are lumped to the
‘hot composite curve.’ Graphical construction, similar to the heat curves of the exam-
ple, as illustrated by Figure 5.9, allows the determination of the maximum overlap,
depending on the minimum temperature difference that we allow between the hot
and cold composite curves. The point at which the composite curves are closest to-
gether to the selected minimum temperature difference is called the ‘pinch,’ which is
considered the key area in designing an energy efficient heat exchanger network.
Pinch Analysis is the thermal analysis of a process to determine realistic targets
for process-to-process heat transfer, heating and cooling by external energy resources
similar to the analysis for one hot and one cold stream in the example. Commercial
software like Supertarget [26], Aspen Energy Analyzer [27], and various process sim-
ulation packages support such analysis and provide easy interfaces to generate these
curves from a process model. The composite curves are a powerful visualization to
show what heat recovery (process-to-process heat exchange) is possible and what
quality of external energy resources or related utilities are necessary. It also gives an
indication how difficult it will be to get the desired heat transfer between the hot and
cold process streams to and from the utilities. Figure 5.10 illustrates that if curves are
Fig. 5.8: Cold composite curve.
Fig. 5.9: Typical composite curves showing pinch, process-to-process heat exchanger, heat demand,
and release to external sources.
Fig. 5.10: Composite curves for easy and complex heat transfer.
5.9 Design for human factors | 119
far apart and the pinch is a narrow region (left) the heat transfer will be easier than
when the curves are close together and have a large parallel region (right). Based on
this insight it is possible to design an efficient utility system that converts the primary
external energy resources to the utilities required in the process. Also it is possible to
derive from the composite curves process modifications that will ease or enhance the
heat transfer and finally allow for a more energy efficient design.
Once the process is fixed and the utilities selected, the composite curves give guid-
ance in designing the heat exchanger network to heat balance the process following
the Pinch Design Method. Most pinch related literature focuses on this last design
stage, but in practice this network design is only a small part of the process design
work. Pinch Analysis is useful in the entire process design work and it is recommended
to apply it early when the first process concepts are evaluated to identify the scope in
which to improve the energy efficiency of the process, especially when physical heat
transfer is important in this process. Excellent explanations of Pinch Analysis and the
Pinch Design Method can be found in Smith [24] and Kemp [28].
5.9 Design for human factors
Processes and products are sometimes two sides of the same coin. A process facil-
ity is in essence a large machine that is operated and interacted with by personnel
and can thus be regarded as a sociotechnical system. Several large scale incidents
with fatalities have occurred in the industry where human factor aspects such as poor
human-machine interface, lack of understanding of human error, lack of situational
awareness, etc. were the cause. Human factors are all those things that enhance or
improve human performance in the workplace. Human factors engineering (HFE) fo-
cuses on the application of human factors knowledge to the design and construction
of sociotechnical systems. The objective is to ensure systems are designed in a way that
optimizes the human contribution to production and minimizes potential for design-
induced risks to health, personal or process safety, or environmental performance [29].
Good HFE will increase HSE (health, safety, and environment) performance and
reduce the likelihood of error, which in turn helps to improve the operational cost
of the facility. There are a variety of HFE studies that can be performed depending
on the project phase. OGP [29] has produced a guidance in which the most common
ones are described per project phase. Studies that can be performed during the basic
engineering phase are for example:
– Working environment health risk assessment (WEHRA);
– Valve criticality analysis (VCA);
– Vendor package screening;
– Task requirements analysis (TRA);
– Human machine interface (HRI) requirements analysis;
– Control room requirements analysis;
– Control system and alarm management analysis;

– Safety critical task inventory;
– Critical task analysis;
– Human error ALARP demonstration.
One HFE study discussed here as an example and typically performed during basic en-
gineering and detailed engineering studies is called the “working environment health
risk assessment” (WEHRA).
The goal of the WEHRA is to identify work tasks that may be associated with po-
tential risks to health. The purpose is to uncover these risks, investigate if they can be
avoided with proper design and suggest measures to reduce the risk to an acceptable
level. The first step is the mapping and creation of an activity chart, where the nature
of the activity, area of work, duration, and frequency of the work tasks are described;
see Table 5.3.
Tab. 5.3: Example WEHRA activity chart for a single item.
No. Area Equipment Activity Duration Frequency No. Persons

1 Cellar deck Choke Replacement of 8 hours 2 times per 1 on board the
actuator lifetime for installation
each choke
To identify the events/tasks that may lead to any potential health hazard, a ranking
table is used. In this table, all activities involving personnel are listed and evaluated
per working environment factor used. The scoring and evaluation is done in a meeting,
where several disciplines can provide their input. The working environment aspects
that are covered are:
– Noise;
– Vibration (whole body, hand-arm);
– Ergonomics;
– Indoor climate;
– Outdoor operations/climate;
– Lighting;
– Radiation;
– Chemical substances and preparations;
– Biological agents;
– Psychosocial and organizational conditions.
A qualitative risk ranking is performed for each working environment factor; see Ta-
ble 5.4. Evaluations are noted down and actions formulated for follow-up in the de-
sign.
Tab. 5.4: Example row of a ranking outcome that is created for each activity in the WEHRA activity
chart.
Noise Vibra- Ergonomics/ Climate Lighting Radiation Chemical Biologi- Psych./

tion access cal org.
2 1 5 2 1 1 1 1 1
5.10 Design for six sigma
The term ‘design for six sigma’ is used in two ways. The first is just about making the
design such that the six sigma (sigma stands for standard deviation) value of the prod-
uct output is within the specifications that the customers agreed to. This means that
the chance that at any time the product is outside the specification limits is extremely
small. This method is applied in the detailed design stage by using all statistical infor-
mation on input parameters and choosing the design parameters such that this input
variation is so limited that the product variation defined by six sigma is within the
product specification.
The second way is a method to improve the quality of the generic business pro-
cesses related to design and innovation used inside a company. It is derived from the
quality improvement method six sigma. The latter is focused on improving existing
business processes using statistical data and information.
For further reading the reader is referred to textbooks by Cudney [31] and Yang [32].
[1] Benyus, J. Biomimicry: Innovation Inspired by Nature, New York, William Morrow Paperbacks,
2002.
[2] D. Baumeister, Biomimicry Resource Handbook: A Seed Bank of Best Practices Createspace,
New York, Independent Publishing Platform, 2014.
[3] Martin-Palma RJ. Engineered Biomimicry, New York, Elsevier, 2011.
[4] Ayres RU, Ayres LW. A handbook of Industrial ecology, Cheltenham USA, Edward Elgar Publ.
Ltd, 2002.
[5] Graedel TE, Allenby BR. Industrial ecology, 2nd edn, Upper Saddle River, N.J., Pearsen Educa-
tion Inc. 2003.
[6] Ellen MacArthur Foundation. Circular economy scheme. Sourced from:
https://www.ellenmacarthurfoundation.org/circular-economy/interactive-diagram. Last re-
sourced: 15 January 2018.
[7] Ellen Macarthur foundation. Sourced from: https://www.ellenmacarthurfoundation.org/. Last
resourced: 15 January 2018.
[8] Cramer J. Circular cities, plenary lecture, 10th World Congress Chemical Engineering, WCCE10,
Barcelona, 2017.
[9] Cramer J. Circular cities webinar. Sourced from: http://www.nzwc.ca/events/circular-cities-
amsterdam/Documents/Webinar-JacquelineCramer-Mar14-17.pdf.
[10] Braungart M. Eco-effectiveness Cradle-to-cradle design method, plenary lecture, Watervisie

Congres te Rotterdam op 18 februari 2016.
[11] Braungart M, McDonough W. Cradle to cradle – Remaking the way we make things, New York:
North Point Press, 2002.
[12] Chertow, Marian R. Uncovering industrial symbiosis. Journal of Industrial Ecology 11(1), 2007,
11–30.
[13] Harmsen J. Sustainable Development and the Process Industries, NPT, 2013, 18–19.
[14] Wu Q. Industrial Ecosystem Principles in Industrial Symbiosis: By-Product Synergy. Chapter 14,
Harmsen J, Powell JB (eds.), Sustainable Development in the Process Industries: Cases and
Impact, Hoboken NJ: John Wiley & Sons, 2010.
[15] Magan A, Olivetti E. By-product Synergy Networks: Driving Innovation Through Waste Reduc-
tion and Carbon Mitigation, Chapter 6, Harmsen J, Powell JB (eds.), Sustainable Development
in the Process Industries: Cases and Impact, Hoboken NJ: John Wiley & Sons 2010.
[16] Dautzenberg F. A journey towards Green(era) technologies, Kivi lecture, Technical University
Eindhoven, 2017.
[17] Liquid Light. Sourced from: http://cen.acs.org/articles/93/i43/Liquid-Light.html. Last re-
sourced: 15 January 2018.
[18] Murzin DY (ed). Chemical Engineering for Renewables Conversion, Advances in Chemical Engi-
neering, 42, 2013, 2–371.
[19] Pike E. Opportunities to improve tire energy efficiency, The International Council on Clean
Transportation, White paper 13, 2011.
[20] Dincer I, Rosen MA. EXERGY: Energy, environment and sustainable development, 2nd edn,
Amsterdam: Elsevier, 2012.
[21] Smith JM, et.al. Introduction to Chemical Engineering Thermodynamics, 8th edn, New York,
McGrawHill, 2018.
[22] Sankey diagram. Sourced from http://www.sankey-diagrams.com. Last resourced: 15 January
2018.
[23] Johansson TB, et al. GEA, Global Energy Assessment – Toward a Sustainable Future, Cam-
bridge: Cambridge University Press, 2012.
[24] Smith R. Chemical process design and integration, Chichester: John Wiley & Sons, 2005.
[25] Thulukkanam K. Heat exchanger design handbook, 2nd edn, Boca Roca: CRC Press, 2013.
[26] Sourced from: http://www.kbcat.com/energy-utilities-software/supertarget. Last resourced:
15 January 2018.
[27] Sourced from: http://home.aspentech.com/en/products/pages/aspen-energy-analyzer. Last
resourced: 15 January 2018.
[28] Kemp IC. Pinch Analysis and Process Integration – A User Guide on Process Integration for the
Efficient Use of Energy, 2nd edn, Oxford: Butterworth-Heinemann, 2007.
[29] OGP. Human Factors Engineering in Projects, Report 454, 2011. Sourced from: https://
humanfactors101.files.wordpress.com/2016/02/human-factors-engineering-in-projects.pdf.
Last resourced: 15 January 2018.
[30] Edmonds J. Human Factors in the Chemical and Process Industries: Making it Work in Practice,
1st edn, New York: Elsevier, 2016.
[31] Cudney EA, Agustiady TK. Design for Six Sigma: A Practical Approach through Innovation,
Cleveland Ohio: CRC Press, 2016.
[32] Yang K, El-Haik BS. Design for Six Sigma: A Roadmap for Product Development, 2nd edn, New
York: McGraw-Hill 2009.
6 Scoping the design
6.1 Introduction to scoping the design
Scoping the design involves defining the design goal, its constraints, its scope, and its
context. This set of actions is carried out in discussion with the client of the designer
as described in general in Chapter 4. That chapter is well suited for experienced de-
signers. This chapter, with its specific guidelines, is particularly suited for students
learning to design. The focus of this chapter is on the concept and feasibility innova-
tion stages but advice for the other stages will also be provided.
6.2 Defining design goal and name
6.2.1 Purpose design goals and names
A design goal is a clear short statement on what should be achieved by the design. In a
business environment this will be obtained after several rounds of discussion with the
client. In design education it will, in a simple form, be provided by the teacher. The
students should then redefine it so that each design team member and the teacher
agree on the design goal. The design goal statement is crucial in communicating with
anyone what the design is about. A vague design goal will lead to a mediocre design.
A clear design goal is a good start to a superior quality design. A good name will further
enhance the communication inside the team and with others outside the team.
After the design goal is defined the design should be given a name. The discussion
in the design team to generate and decide on the design name will also help to create
a good team spirit.
6.2.1.1 Design goal setting

The design goal should be a very clear short statement that the internal client (busi-
ness case owner) and all participants of the innovation group understand and agree
upon. It is the shortest summary of what the design is about. For product concept de-
sign the goal can be formulated in terms of what functions the product has to have. An
example is: design a chewing gum that does not foul the streets. The process design
will state how much of what product has to be produced.
For starting a breakthrough innovation it is very important to define a challenging
design goal statement. It should state the aim of the project in very concrete terms.
It takes time to formulate this problem statement but you only can work efficiently
if you know what your goals are. If it is not concrete, sooner or later conflicts will
arise between people working in the concept stage who cannot decide what the better
solution is [1].
https://doi.org/10.1515/9783110467741-006
124 | 6 Scoping the design
If the goal statement is not different enough from the goal statement that gave rise
to the current commercial process then the end result of the whole innovation process
is that you merely rediscover what was good about the original process. All innovation
work has then been in vain [1].
The goal statement should therefore contain a quantitative strong reason for
change. The best way of achieving this is by explicitly referring to the best commer-
cial process or product and stating in the goal and also in the constraints how much
better the novel solution should be. This means that a good reference case needs to
be defined. Section 6.4 discusses this subject in more detail. The main purpose of this
challenging design goal is to stimulate design team members to come up with new
solutions.
The design goal stated in the concept stage may be modified at the start of subse-
quent stages. For product designs this modification will often be needed, due to added
insight from the interaction between the market desires and the product composition.
For process designs the goal often stays the same or is only slightly modified.
6.3 Defining the design scope (system levels, boundaries, and

context)
A major choice when scoping the design is the choice of system level in which the
design will take place. All system levels relevant to design are described in Chapter 2
and a conscious choice of the system level for the design should be made. This choice
will involve a discussion with the client. Unconsciously, a system level may already
have been selected by the client. The designer can then show the other system levels
to the client and suggest advantages for including other system levels in the design
scope. In a design course the teacher may first choose a system level in which the
students have to make a design and then choose a higher system level and again ask
the students to make a design. In this way students become aware of the importance
of choosing the system level.
Even within a system level the boundaries for design still have to be chosen. The
boundaries define which elements are outside the design objective and are treated as
context (a given or a constraint). These boundaries will also be defined in discussion
with the client. A simple example of a process design is whether to include storage of
feedstock and product in the design scope or not.
After the system level has been chosen for which the design will be made, it is use-
ful to define the other system levels that form the context of the design. If, for example,
it is decided to take the chemical reactor and the catalyst level as the design scope of
the process, then the other higher system levels will not structurally be changed by
the new design. However, the size of mass and energy flow to and from the surround-
ing system levels may change, due to the new reactor and/or catalyst design, if the
amount of byproduct relative to the main product is reduced.
6.4 Defining constraints (specifications)
6.4.1 Design constraints
A design constraint in general consists of a type, a parameter, and its upper and lower
limits. These constraints may also be called design specifications. A good starting
point for defining constraints is to use the generic criteria check list provided in Chap-
ter 4 and decide for each criterion what type of constraint and what quantified lim-
its should be included. For a concept design by students the reduced constraint list
SHEETS of Chapter 4 can be used.
In addition to these there will be constraints imposed by the company strategy.
If the company has an explicitly written down innovation strategy then defining the
strategy related constraints is easy. This may however not be the case. The written
strategy may not be available, or the strategy may not be sufficiently defined to cover
the specific innovation design and project. Once the design is made and presented, a
conflict with the implicit strategy may be revealed. Then the strategy may be further
refined and the design constraints adapted to this new situation.
Here is an example from the author’s experience. A team had to design a sus-
tainable process for ethylene production. The design solution by the team was to take
biomass as a feedstock. Via fermentation and dehydration ethylene was produced.
When the design result was presented to top management, the reply was, that an ex-
isting output stream from a crude oil should be used as feedstock and that the sole use
of biomass as feedstock was outside the company strategy. That latter statement was
until that time not written down. Shortly afterward, it was included in the company
strategy.
This example shows that a strategy can also be underdefined. A stage-gate ap-
proach in which concept design is presented to management in the discovery stage will
quickly reveal this underdefinition, resulting in a more defined strategy and thereby
avoiding large research and development efforts on projects that do not fit the busi-
ness strategy.
In the discovery stage the constraints will be limited to that which fits the company
strategy and perhaps that for which potential value for the company can be created.
For the concept stage the minimum set of constraints that should be used is SHEETS:
– Safety
– Health
– Environment
– Economics
– Technically feasible
– Socially accepted
In addition it is becoming more and more common to include a constraint on the con-
tribution to sustainable development. A full set of constraints inside a company will
be longer. The full set provided in Chapter 4 may be used as a checklist to make sure
that no constraint type is forgotten.
It is good to consider however that at the start of the concept design the problem
will still be underdefined and ill-defined. In some cases the problem is first defined by
the client of the designer and during some meetings, due to the knowledge provided
by the designer, the problem is better defined. Some preliminary solutions may be
presented to the client to test whether the problem is properly defined. If so, the prob-
lem is specified further into a goal and constraints (specifications). This subtle game
between designer and client has been described by an experienced designer as such:
“Our job is to give the client, on time and on cost, not what he wants, but what he never
dreamed he wanted- and when he gets it he recognizes it as something he wanted all
the time.” [1]. Read also in Chapter 4 more about this subtle game. For students it is
very important to learn that design problem definition is not just accepting what the
teacher provides as the design problem to work on, but redefining it. In education a
dedicated effort has to be spent on this aspect.
For the feasibility stage a complete set of constraints should be defined and dis-
cussed with internal and external clients, as now the product and the process for com-
mercial scale production has to be defined. The same holds for the development and
detailed design stages. The table on all aspects relevant to a design as presented in
Chapter 4 can be used to generated a complete set of criteria and can also be used as
a checklist to determine whether an item was forgotten.
6.4.2 Product specifications
Product specifications are of tremendous importance in both product design and pro-
cess design. In product design the specifications will be developed through all in-
novation stages by interaction with marketing and customers. It is crucial that the
developed specifications reflect the customer needs, differentiate the product from
competing products, and are technically as well as economically realizable. In the ini-
tial project stages the specifications represent the hopes and aspirations of the team,
often without being constrained by what can be technologically achieved. During the
course of the project the specifications are further refined to include actual technolog-
ical constraints and expected production costs by making difficult trade-offs between
desirable product characteristics.
For consumer products in particular it is important that the product to be devel-
oped will be appealing to all senses. Here is the list of senses to ensure that for each
sense, product specifications are made:
– Sight
– Feel
– Hearing
– Smell
– Taste
6.4.3 Process specifications
Process specifications will also have to be made. These will start with product out-
put specifications, which are mostly obtained from the business part (marketing and
sales) of the organization. If the product is radically new then not all characteristics
of the new product are captured in quantified specifications. Some product specifica-
tions are then likely to be defined/provided later on, when the product prototype has
been tested by the client, resulting in additional specifications. The amount of product
to be made will also be part of the output specification.
For radically new products with functional concept specifications, the process de-
signer is advised to make multiple process designs based on assumptions on the prod-
uct composition and to discuss the process designs and their potential consequences
for product performance properties with the product developer.
Input process specifications also start from the business part of the company,
which provides which types of inputs are allowed and desired and also which part
of the supply chain is now considered inside the design scope. These pieces of infor-
mation will be used to define the input specifications. In addition, the process concept
designer will further define and refine the input specifications using the relations with
the process output specifications based on yield figures. As a starting point a 100%
theoretical molar yield of product on feedstock may be used, or empirically obtained
molar yield data from scouting experiments. The designer should be aware that often
the precise yield dimension (molar or mass based) is not reported in the experimental
results. A check on the precise yield definition will be needed. How the output and
input specifications have been obtained will also be part of the design report. Chap-
ter 13 contains more details of what and how process outputs and inputs should be
reported.
6.4.3.1 Identifying reference cases to improve design constraints

Identifying a reference case and defining constraints for the new design with respect
to this reference case is a very powerful method to make sure that the final design so-
lution is far better than the reference case. The reference case is the best existing com-
mercially available product and/or process, upon which the design solution should
deliver a significant improvement.
For a radically new product design, defining a meaningful reference case can be
complicated. Often the novel product has no single commercial reference product. The
reference cases may be obtained by observing the functionality of the novel product
and then finding existing products fulfilling the same functionality as group, so that
a combination of existing products is selected as the reference case.
For a novel process design the reference case selection is relatively easy. It should
be the best process in commercial operation for the same product. If that process is
in in the company where the innovative design takes place then all information will
be easily available. If the process is operated by a competitor then information is less

easy to obtain. In some cases the information can be gathered from environmental
reports on that process that had to be made public.
For a novel product or process concept design it is important to set a few hard-
to-meet constraints related to the reference case. An example is setting the constraint
that the novel design should be a factor of 4 times better than the reference case. The
reason for this requirement is the same as for defining the product goal, namely setting
only challenging constraints for the initial new concept design will ensure that the
final design in the detailed stage will still be superior to the reference case.
A final remark on constraints: it is not very important whether the statement on
the requirement to improve significantly relative to the reference case is put under the
design goal or in the constraint section, or in both. The only guideline is: the design
goal should be short. The number of constraints should be many.
6.5 Generating basic design data
For a design, input information to facilitate the design will be needed. This type of
input information is called basic design data. These data need to be generated. This
is done by searching in the literature and patent sources, by making best guesses,
by making calculations, and by carrying out experiments. Not all information will be
generated prior to the design start. Some will be generated during the design, when it
appears to be needed to complete the design. For each subsequent innovation stage,
more of these basic design data with an increasing level of detail will be generated. The
ultimate challenge here is to generate just the minimum amount of data required to
make reliable decisions in each stage, as generating more will result in unnecessary
project expenses for data that may never be used when a project is not continued.
This chapter provides guidelines on generating these data. The focus is on the first
innovation stages up to and including development.
Basic design data includes not only data, but also information, knowledge, un-
derstanding, and wisdom as defined by Ackhoff [2]. Here is a short elaboration on the
elements covered by basic design data:
– ‘Single discipline’ design data is data that can be easily understood by people
skilled in that discipline, such as physical property data, safety data, and eco-
nomic data. Single discipline data in general consist of a parameter (symbol), an
amount (figure), a dimension, and may contain a reference. An example is: the
density of water is 1000 kg/m3 (at 277 K). A further distinction is between scale
independent data, such as physical properties, and scale dependent data, such
as investment cost, which depend on the scale (production capacity) of the de-
sign.
– ‘Information’ means relations between data. An example is reaction kinetics; pro-

viding information on the relation between reaction rates, concentrations, tem-
perature, and catalyst type.
– ‘Knowledge’ means applying the data and information to the design. For example,
it could mean selecting a best residence time distribution concept, such as plug
flow, and using reaction kinetics information for reactor concept design.
– ‘Understanding’ means appreciating why the design choices have been made. It
is cognitive and analytical. It is about explicit reasoning on why the whole or a
major design is made as it is. It involves for instance the design choices made on
a particular separation and reaction combination using safety and economic cri-
teria.
– ‘Wisdom’ means evaluated understanding. It calls upon data, information, knowl-
edge, and understanding and in particular on moral and ethical codes. Wisdom
is in general very seldom described in basic design data, but when it is practiced,
it is of enormous importance.
An example from the author’s experience on the role of wisdom is a design for shale
mining and oil production in Morocco in the eighties by the company I worked for. One
of the design questions was, what are the environmental constraints? It appeared that
the Moroccan government had a general set of environmental emission constraints
for business activities. Experienced process designers in the company felt insecure
about this set, and initiated an investigation into the local environment for the min-
ing and the conversion process. The investigators reported that a migration bird sanc-
tuary was near the spot of the mining and that it could be disturbed by the mining
and processing activities. The company then decided to make sure that this would not
happen and also decided to apply European environmental standards for the design.
The reasoning was that causing environmental problems in Morocco could damage
the company’s global image of being a responsible business.
It is very important that generated data are reported, discussed, stored, and made
retrievable to all project members. Usually such a database can be best established in
the concept stage. Chapter 7 deals with the subject of reporting, storage, and retrieval
in detail. Table 6.1 provides an overview of potential sources to generate basic data rel-
evant for chemical product and process design. In the following sections, guidelines
for basic design data generation in each innovation stage are provided.
6.5.1 Ideation stage
The main objective in the ideation stage is to identify which data are available and
need to be obtained when deciding to pursue an idea further in the concept phase.
Data generation in the ideation stage is therefore limited to those that can be easily ob-
tained from literature sources or simply and quickly derived using simple equations
and correlations or best guesses made by an experienced researcher or process de-

velopment engineer. Physical properties, safety, health, and environmental data will
be obtained from literature sources. If not available, educated guesses will have to be
made. Consulting experienced employees of the company is often the quickest way of
obtaining the required data and information. Economic data such as earning power
can be derived by substracting feedstock cost from sales prices; for more information,
see Chapter 10 on economics.
6.5.2 Concept stage
The main challenge in the concept phase is to just generate the required data to be able
to decide on the most promising design concept(s). There is a huge risk in wanting to
generate lots of nice to know data, which are in the end discarded because a different
concept is chosen or the project is abandoned as a whole. Because in the concept stage
quick decisions have to be taken and the budget is limited, the emphasis should be on
rapidly finding these data, even at the expense of accuracy and reliability.
Some physical-chemistry data are always needed in the concept stage. Crucially
lacking physical and physical-chemical data may be generated by specific experi-
ments. These experiments can be done in-house but specific experiments such as
physical basic data can also be obtained from specialized outside laboratories. Alter-
natively modern flowsheet computer packages that have a library of physical property
models can be used to quickly generate physical properties. Care should be taken in
using these models. They need a specialist to choose the best model for a particular
property. It may also involve additional experiments to validate the calculation re-
sults. More detailed physical-chemistry data measurement and modeling studies are
usually performed in the feasibility stage.
Generation of chemical kinetics data in the concept stage for process concept de-
sign is in general restricted to:
– The chemical stoichiometric equations of the main reactions involved
– Order of magnitude reaction times required for specified temperature and concen-
tration windows
– Experimental results of proof of principle experiments showing products made
and the yields of product on feedstock for a particular catalyst, some tempera-
tures, and reaction times.
Economic data in the concept stage is in general limited to the purchase prices of feed-
stock and sales prices for products. For new products these prices will not be available.
Pricing new products in the concept stage is an important step for innovative product
companies. The reader is referred to the classic article by Dean [3] and to a McKinsey
article [4] to get some insight into the art of new product pricing. Detailed methods for
economic evaluations of novel processes and products are found in Chapter 10. Here
we simply state the following guidelines:
(1) Estimate the new unit product value for the customer and do not base it on man-
ufacturing cost.
(2) Determine the feedstock cost per unit of product.
(3) Estimate the Earning Power EP0 .
Safety, Health and Environmental (SHE) data are also very important for first a rapid
analysis of preliminary concept designs and for evaluating concept designs when fi-
nalized. The product or a process that is inherently unsafe will not pass the concept
stage gate. A product or process with no SHE information will also not pass the con-
cept stage gate, or if it does then it will not pass the feasibility stage gate.
6.5.3 Data generation feasibility stage
Data generation in the feasibility stage is focused on detailing the required informa-
tion for the selected design to such a reliability level that it can be used to establish a
feasible commercial scale product and/or process design and when needed also a pi-
lot and/or demo plant design. Alternative technical options for each process step will
be contemplated. Selection criteria will be generated, and the best options selected.
Equipment will be sized and stream compositions calculated. All these data need to be
documented, stored, and made retrievable. For a plant design inside an existing pro-
cessing site it is important to have all local plant and site related boundary conditions
documented.
For reliable experimental physical property generation and modeling, large enter-
prises will have their own specialist group. Smaller companies will turn to specialized
research institutes to obtain key pieces of information. Similarly much more detailed
and reliable chemical kinetics data need to be obtained as the best reactor concept
with its connected separation concept plus recycling can only be designed with sound
kinetics. Here is a checklist of required chemical kinetic data:
– (Bio)chemical reactions
– Possible reactants/pathways
– Stoichiometry
– Conditions (pressure, temperature)
– Yields, selectivity vs conversion
– Product distribution
– Side reactions
– Reaction rates
– Equilibrium
– Phases (gas, liquid, multi)
– Exothermic/endothermic
– Residence time, space velocity
Tab. 6.1: Sources for basic design data.
Data Type Reference books Web based sources

Physical- – The properties of gases and – Properties of organic compounds
Chemistry data liquids, McGraw-Hill 2004 – Chemexpert Chemical Directory
– Materials Science and Engineering – NIST Chemistry Webbook
Handbook, 4th Ed, CRC, 2004 – Beilstein online database
– CRC Handbook of Chemistry and – Knowledge Centre (Bio) Chemical
Physics, 97th Edition, 2016 Engineering
– Dechema Chemistry Data Series,
Dechema (2004)
Technology and – Ullman’s Encyclopedia of – www.sciencedirect.com
scientific Industrial Chemistry – www.scopus.com
information – Kirk Othmer’s Encylopedia of – SciFinder
Chemical Technology – Wikipedia
– Encyclopedia of Chemical – www.google.com
Processing and Design – Patents
– Riegel’s Handbook of Industrial – http://nl.espacenet.com/
Chemistry – www.patsnap.com
– www.nexant.com
– www.ihs.com
– www.intratec.us
– www.cheresources.com
SHE data – BP process safety series – KCCE: databases
– Handbook of chemical health and – ChemWatch (product safety)
safety – INCHEM (chemical safety and
– Handbook of chemical compound environmental health)
data for process safety – International chemical safety card
– Chemical safety handbook (by CAS number)
– Property estimation methods for – MSDS
chemicals – http://chemiekaarten.sdu.nl/
– Handbook of environmental health chkonline/
– Handbook of reactive chemical – http://hazard.com/msds
hazards – http://www.ilpi.com/msds
– EU Directives and Law (for
industry)
– http://eur-lex.europa.eu/
– http://ec.europa.eu/enterprise/
– www.msdssearch.com
Economic data – www.icis.com/chemicals
(market, pricing) – www.nexant.com
– www.ihs.com
– www.orbichem.com
– www.intratec.us
– www.business.com/directory/
chemicals
– www.alibaba.com
Data Type Reference books Web based sources

Equipment cost – Product and Process Design – www.matche.com/equipcost/
Principles, 3rd Edition, John Wiley – www.mhhe.com/engcs/chemical/
& Sons (2010) Chapter 22 peters/data/ce.html
– Plant Design and Economics for – www.equipnet.com
Chemical Engineers, 5th Edition, – www.ipps.com
McGraw-Hill (2004) Chapters – www.alibaba.com
12–15
– Dutch Association of Cost
Engineers
Also market and economic information will be detailed further to establish the busi-
ness case of the project and assess its economic viability. Important aspects to be in-
cluded are:
– Local society and cultural dimensions of market
– Market share, competitive position
– Desired production rate
– Desired product purity
– Product sale price versus purity specifications
– Feedstock materials compositions and their prices
– Location
– Climate
– Depreciation agreement or policy
– Feedstock prices
– Product sales prices
– Investment cost estimate
– Utilities availability, specifications and price
– Process cost structure
– Raw materials availability and composition
A minimum set of safety, health, and environmental data for the feasibility phase com-
prises:
– Safe operating window for the process parts
– Toxicity of all components applied for product and process
– Life cycle assessment of environmental impact
– Waste disposal options
Waste streams/emissions
– Environmental legislation
Chapter 11 provides comprehensive information on safety and health data and Chap-
ter 12 discusses the sustainability data needed for evaluation.
6.5.4 Development stage
During the development stage a large number of data for many design object aspects
have to be generated on top of the already generated data in previous stages. These
data concern:
– The commercial scale product and process and their related product prototype
and pilot and/or demo plant
– Data generated by prototype and pilot plant testing to validate the feasibility of
the novel product and process
– The marketability of the novel product
– Intellectual property protection
Table 6.2 provides a checklist of items for which basic design data have to be generated
in the development stage.
Process performance data

Phenomena that cannot be predicted from modeling are: corrosion, fouling, foaming,
catalyst decay solvent decay, and others. These potential problems should be part of
the reasoning for having the pilot plant; see also Chapter 3 on reasons for having a
pilot plant. The pilot plant and the testing program should be such that observations
and measurements on corrosion, fouling, foaming, and catalyst decay can be made.
Tab. 6.2: Basic design data items of the development stage.
Item Obtained from

Safety performance: product and process Chapter 11 Literature and experiments
Health performance product and process Chapter 11 Literature and experiments
Environmental performance LCA data Chapter 12 Literature
Sustainable development performance Chapter 12 Literature
Economy performance: see Chapter 10 Calculations
Product technology data
Process technology data
Technology performance process Pilot plant test results
Technology performance product Prototype product test results
Product market data Market surveys
Product sales data Market and client surveys
Intellectual property Design rights, trademark, patents, copyright
[1] Dorst K. Understanding Design, 175 reflections on being a designer, Amsterdam: BIS Publishers,
2006.
[2] Ackoff RL. From Data to Wisdom, J. Appl. Sys. Anal., 16, 1989, 3–9.
[3] Dean J. Pricing policies for new products.,Harvard Business Review, 54(6), 1976, 141–153.
[4] Marn MV, et.al. Pricing new products, McKinsey Quarterly August 2003. Sourced from:
https://www.mckinsey.com/business-functions/marketing-and-sales/our-insights/pricing-
new-products. Last resourced: 6 October 2017.
7 Executing designs
7.1 Introduction to executing designs
This chapter describes methods and guidelines for executing novel product and novel
process designs for innovation classes 4 and lower, as defined in Chapter 3. Its focus
is on designs for discovery and concept stages, and is particularly useful for MSc stu-
dents and for young professionals.
Figure 7.1 shows all major steps in executing any design in any stage. The first step;
scoping is treated in Chapter 6. This chapter provides general methods and guidelines
for the steps: synthesize, analyze, evaluate, and report.
The design steps synthesis and analysis, have strong interactions. An initial syn-
thesis result is analyzed by how well it meets or fits the design scope. The difference
between the analysis and evaluation steps is, that in the analysis step, the synthesized
design options are often modeled and qualitative and/or quantitative data generated
(e.g., mass and heat balances) and simulation outcomes are compared with the design
performance requirements (like atomic efficiency, economic margin, etc.). These syn-
thesizing and analyzing steps will be repeated to the satisfaction of the design team
until a design is achieved that meets the scope goal and criteria. Then the result will
be evaluated by people from outside the design team and the result will be reported.
Methods and guidelines for all these steps are provided in Sections 7.2–7.5 of this chap-
ter.
In the discovery and concept stages it is recommended to create several design al-
ternatives and then select the best (or the top two or three), to proceed with. The other
alternatives are reported and noted for optional use later. Therefore, also the subject
of ranking and selecting best designs is added in Section 7.6. Special attention is also
Scoping
Synthesize
Analyze
Evaluate
Report Fig. 7.1: Design steps.
https://doi.org/10.1515/9783110467741-007
paid to modify the design such that it is robust to uncertainties and its intellectual
property is protected. Guidelines on these subjects are also in Section 7.6
How to analyze and optimize a selected design by making models and simulating
a design performance is covered in Chapter 8 and 9. Specific guidelines on how to
evaluate a design for the important criteria of safety, health, environment, economics,
technical feasibility, and social acceptance are found in Chapter 10–13. Guidelines on
how to report are found in Chapter 14.
7.2 Synthesizing preliminary solutions
Generating preliminary solutions by design is called (design) synthesis. This synthe-

sizing can be done by using heuristics and by following specific steps and guide-
lines. For products and processes synthesis in general heuristics information is pro-
vided in Section 7.2.1. Guidelines for synthesizing products and structured products
are provided in Section 7.2.2. Guidelines for synthesizing processes are provided in
Section 7.2.3.
7.2.1 Synthesizing solutions in general using heuristics
Synthesizing design options in each design level is about generating preliminary solu-
tions. There are many ways to generate preliminary solutions. Here are some practical,
but very general heuristics to generate (synthesize!) preliminary design solutions:
(1) Look at the design scope generated by the methods of Chapter 6. Read the design
goal. Then scribble down any idea that comes to mind that may partly meet the
design goal. The scribbling may be a simple sketch.
(2) Ask yourself: what does the solution have to do to (partly) meet the design goal?
Write down what the solution must do. Verkerk [1] calls this the qualifying func-
tion. Do not bother determining how the solution would or if it could work. If you
think you need more than one function to meet the design goal, then write down
those functions.
(3) Look for information on existing solutions (literature, patents, interviews of peo-
ple) that may partly fulfill the qualifying function. Write down these partial solu-
tions.
(4) Deconstruct existing solutions into constituting functions. So, make an abstract
description of functions.
(5) Fill in some columns on the left-hand side of the Delft Design Matrix of Chapter 4
with keywords.
(6) Put the design aside and do something completely different.
138 | 7 Executing designs
(7) Anytime an idea comes to mind, note it down directly in your smartphone, voice
message or a piece of paper at hand.
(8) Organize brainstorming, brainwriting sessions (or any other creativity method as
proposed by Tassoul [2] to generate solutions.
(9) Use any of the following words: substitute, combine, adapt, modify, put to other
use, eliminate, rearrange part-solutions to improve the solution (SCAMPER) [3].
(10) Generate families of (part-)solutions with each family having the same ‘ancestor.’
An ancestor is the common element of one family of solutions [3].
(11) Apply the ‘TRIZ’ creative problem solving methodology [4].
7.2.2 Synthesizing products
7.2.2.1 Product synthesizing design using simple heuristics

For novel product design of innovation class 4 all synthesis methods of Section 7.2.1
can be applied. As the new design is a modification of an existing product a lot can
be learned from deconstructing the existing product, hence; method numbers 3 and
4 are in general fruitful.
Here is an example from a group of students that for the first time in their life
had to make a product design. The design goal was to design a chewing gum that
would not foul the streets. They looked for information through literature searches,
patent searches, and by asking a chewing gum manufacturing company to provide
publicly available information about chewing gum. They deconstructed the conven-
tional chewing gum, into the gum that has an elastic plastic chewing function and
taste additions. They concluded that the plastic gum of polypropylene is the part that
fouls the street. It stays on the street for 28 years on average, if it is not removed.
They then looked for a polymer material that would depolymerize into water-
soluble components and found a polyvinyl alcohol with the desired hydrolyzation
rate. They also made the chewing gum very attractive by putting the taste components
in microglobules. These depolymerized in the mouth, by which the taste components
were slowly released into the mouth [5].
7.2.2.2 Synthesizing physically structured products in concept stage

First the difference between physically structured products and other products is ex-
plained. Two types of products can be distinguished in our book:
– Simple homogeneous products down to the molecular scale
– Structured products homogeneous down to a few microns and then heteroge-
neous and structured products
(Heterogeneous products such as cars and devices, which are heterogeneous on every
scale, are not treated in this chapter.)
Simple homogeneous products are gases, liquids, and solids, such as polymers
and ceramics. Particulate matter such as crystalline powders also qualify for this type.
Physically structured products are for instance emulsions, foams, and gels (see
Chapter 8, Table 8.5 for more examples). They have one or more thermodynamic
phases, with at least one of the phases having distributional features at the tiny
and microscales. For instance, some chemical species in the product may have a
distributed molecular structure (e.g., polymeric chain length distribution). In addi-
tion, the product may be made up of some coexisting thermodynamic phases with
a geometrical distribution on a microscale, such as particle size- and/or shape dis-
tributions. ‘Structured’ products have properties that are related to the ratios of the
thermodynamic phases, the chemical composition of these phases, as well as their
internal spatial distributional structure (e.g., droplet size distributions in emulsions).
Designing physically structured chemical products and a more precise characteriza-
tion of the inner structure are presented in the following section.
7.2.2.3 From functional product specification to product structure

The complexity of the internal product structure arises from the presence of multiple
constituents and from the distributive properties of its constituents. Examples are:
(1) Supramolecular structures such as polymers with molecular chain length distri-
bution
(2) A thermodynamic phase present as particles or droplets of varying size
(3) Dispersion of multiple phases
The joint effect of all distributive and lumped properties of the internal structure deter-
mines the practical performance features of the product. These performance features
and product costs determine the competitiveness of the product on the market in meet-
ing customers’ expectations. An example of practical performance requirements for a
structured product such as a crystalline material, is given by Bermingham [6]:
• For downstream handling: ability to filter, to wash, to dry, and to let flow, freedom
from dust, mechanical strength;
• For customer applications: no caking in storage, dissolution rate, mechanical
strength, porosity, freedom from dust, aesthetic appearance;
• For food, some practical performance requirements [7, 8] are, in order of importance:
safety > nutritional value > taste > smell > mouth-feel.
Also, Cussler and Muggeridge [9] and Wesselingh [10] give a number of diverse exam-
ples of (prioritized) practical performance requirements.
When developing and designing a new product and its associated manufacturing
process, a top-down approach is followed from market analysis to process design, as
indicated by downward arrows in Figure 7.2.
Customer expectations/experience
Product functions and performance
Suitable product components
Product composition & structure
Processing conditions along path of formation
Sequencing of process units and operating conditions
Fig. 7.2: Product and process design information downflow and physical causal effects upflow.
The product functions and performance requirements must be derived from the fu-
ture wishes and expectations of customers. To support such translations, a system-
atic procedure has been developed for product design in general. This procedure is
called ‘House of Quality’ [11], also known as ‘Quality Function Deployment.’ It allows
one to crosscheck to what extent customer wishes are covered by the generated set of
functional specifications of the product and if the specifications in the set are more or
less in harmony (free of gross contradictions). A similar translation can be made from
product functions to the selected components that will comprise the product. These
components can be of a physical, chemical, biological, mechanical, or electronic na-
ture. During product design and development, the product functions are materialized
into a physical structure for the product. The use of the House of Quality approach is
also discussed in further detail in Section 8.3.1.
The bottom-up arrows in Figure 7.2 show the physical cause and effects. The selec-
tion of process units with associated operating conditions and equipment design and
their sequencing into a process flowsheet determine the path of processing conditions
inside the process units when going from feed to product. The processing conditions
involve the distributions of thermodynamic phases, pressure, temperature, concentra-
tions, residence times, et cetera. The processing conditions along the path from feed to
product influence the formation of the internal structure of the product. This internal
structure involves chemical composition per thermodynamic phase, the spatial dis-
persion of the phases, and any internal distributive properties of the thermodynamic
phases of which the product is made up (e.g., droplet size distribution).
The interactions between the ‘down’ flow of design information and the ‘up’ flow
of physical cause and effect indicate that having proper information about the physical
and chemical structure of a chemical product is essential to be able to synthesize the
processing plant. The internal product structure is truly the conceptual linking pin
between product design and process design. Knowing such a structure enhances the
effectiveness of the design process as the physical features of the product serve as a
functional specification for the process design.
What are the important generic physical features? Cussler and Moggridge [9,
p. 130] list three central physical features of a chemical product:
– Structural attributes, involving physical and mechanical properties based on in-
ternal product structure
– Changes in thermodynamic equilibrium, induced by external variations (T, pH
changes)
– Key rate processes, which can occur (reactions, mass and heat transfer, fluid flow)
They capture ‘internal product structure’ by means of four main items (pages 129–130):
– Chemical composition
– Physical geometry
– Chemical reactivity
– Product thermodynamics
When putting a product to its application, it is necessary to trigger the product activity
by means of external controls, such as:
– Solvents
– Temperature change
– Chemical reactions
– Physical changes, like pressure, electrical field, and surface tension
E.g., the acidic environment of a stomach triggers the decomposition of the coating of
a drug and the release of its chemically active substances.
The reader is invited to also use the Delft Design Matrix of Chapter 4 to further
synthesize the product and its process, as it is noticed that in virtually all product
design books, the outline of the internal product structure is rather short and ver-
bally informal. These outlines are not sufficient for conceptual product and process
design purposes. It is especially not so when process simulations are used, requiring
a quantitative representation of the product structure. In Section 8.4.3 a mathematical
modeling approach for multilevel structured products is given to analyze the product
concepts.
7.2.3 Synthesizing processes
7.2.3.1 Radically novel process synthesis

Innovative process design follows the abduction-2 thinking described in Chapter 4.
The desired output (the product) is known but the process type is unknown. Radi-
cally innovative processes are in general obtained by methods of process intensifica-
tion [12].
The industrially most successful method of process intensification is that of func-
tion identification and integration into a single unit, such as in reactive distillation.
The following steps can be helpful in obtaining radically novel process designs.
(1) Identify chemical conversion routes to transform input molecules into product
molecules.
(2) Define functional blocks to transform inputs into outputs.
(3) Connect blocks by mass flows represented by arrows from inputs to outputs.
(4) Check by simple atom balances that all input atoms are present in the output.
(5) Look for options to combine functions and use of streams.
Ad Rule 1: Perform literature searches to obtain options for the desired chemical con-
versions.
Ad Rule 2: Functional blocks can be selected from the following list and placed in a
figure; see Figure 7.3:
– Transport
– Mixing
– Reaction
– Heat exchange
– Separation
– Forming
Ad Rule 4: Atom balances can be quickly made as they are not made or destroyed.
Often atom balances reveal missing conversion and separation blocks.
Ad Rule 5: Often functions can be combined. Examples are mass movement and mix-
ing (a pump is also a mixer), reaction and separation (reactive distillation, hybrid com-
bination of membrane, and distillation).
Fig. 7.3: Most frequently used process functions in process design.

7.2.3.2 Modified process synthesis

The first four steps for radically novel design are also useful for modified process syn-
thesis. The difference starts at step 5 where for each functional block a conventional
unit operation is chosen to execute the function.
7.2.3.3 Media choice

The choice of media (solvents, absorption liquid, etc.) is often not considered in pro-
cess synthesis. Often the media chosen in the scouting experiments, such as a solvent,
a crystallization medium, an absorption medium, or a stripping agent are taken as
fixed for the process design. But using these media has all kinds of adverse effects on
cost, safety, health and environment. Additional separation to recover the media will
be needed, as well as additional storage. Diffusive emissions to the atmosphere will
increase due to the additional equipment with their flanges and pumps.
It is far better to consider using available process streams such as feed streams,
product streams, and process streams inside the process for a solvent, absorption liq-
uid, crystallization medium. In bulk chemicals production, where processing costs are
very important, only available media inside the process are used. So, it can be done.
Here are some heuristics:
Media heuristic 1: need a solvent: consider all process streams including feedstock
and product streams.
Media heuristic 2: want to strip a component from a stream: Consider using distilla-
tion or an available process gas stream for stripping.
Media heuristic 3: want to wash a component from stream with water: Consider using
distillation, or extract component for an available process stream.
7.2.3.4 Unit operation choice

After the required process functions have been defined, unit operations can be se-
lected, fitted to the functions. In general, the most important unit operations of a pro-
cess are reaction and separation. These two units define the product yield and thereby
the amount of feedstock needed for the required amount of product. Heuristics are
found in the next section. Selection of other unit operations for mass flow, heat ex-
change, and product shaping can best be carried out by experts inside and outside the
company. For the concept design stage, confection types such as centrifugal pumps,
shell-and-tube heat exchangers, and single screw extruders can be chosen from the
Handbook of Chemical Engineers [13]. In the detailed design stage, dedicated equip-
ment selection can be made by experienced design engineers, involving equipment
providers.
7.2.3.5 Heuristics for reaction plus separation synthesis

Defining and choosing the reaction section in combination with the separation section
is very important for obtaining high product yields on inputs. For the reaction section
itself the best residence time distribution must be chosen as well as the degree of con-
version in combination with the separation and recycling of unconverted materials.
Here are useful and easy ways to apply heuristics obtained from Levespiel [14] and
Douglas [15]:
HR1: Single irreversible reaction: plug flow concept reactor
HR2: Single autocatalytic reaction such as in fermentation: back mixed concept reac-
tor
HR3: Single reaction and adiabatic exothermic: back mixed reactor followed by plug
flow reactor
HR4: If unwanted reaction is high order in B then distribute feed B along plug flow
reactor.
HR5: Consecutive reaction: plug flow reactor with small conversion followed by sep-
aration and recycle unconverted component back to reactor.
HR6: Product inhibits reaction: plug flow reactor with small conversion followed by
separation and recycle unconverted component back to reactor.
7.2.3.6 Heuristics for reactive distillation synthesis

Reactive distillation is now also a conventional unit operation for which heuristic rules
for selection are available. Reactive distillation is a promising option if [17]:
HRD1: The reaction time is less than 30 minutes and the distillation temperature pro-
file is compatible with the reaction temperature. By changing the distillation pres-
sure, the distillation temperature may be changed to required reaction tempera-
ture.
All reactive distillation hydrogenations applied in refineries at very large scale
obey this rule.
If furthermore either of the following conditions apply:

HRD2: There is a consecutive reaction forming an unwanted byproduct.
HRD3: The reaction is an equilibrium reaction.
HRD4: The distillation has an azeotrope.
7.2.3.7 Heuristics for separation synthesis

For selecting promising separation concepts, Table 7.1, based on heuristics obtained
from PDC can be applied [16]. The top technique has the lowest cost. In descending
order, the techniques become more expensive. So, by checking from top to bottom
which technique can be applied, the most promising option is selected.
Tab. 7.1: Separation techniques selection hierarchy.
Separation techniques Practical solution examples

descending order more expensive
Phase separation Decantation
Phase generation single step Flash, condensation, crystallization
Phase generation multiple steps Distillation, membrane
Separation with recycling L/L Extraction, extractive distillation, adsorption
ABCD S1 A
BCD
S2 B
CD
S3 C
Fig. 7.4: Functional binary separations block flowsheet for four products from single feed.
7.2.3.8 Process separation synthesis from complex input

Process separation synthesis in the discovery and concept stages is facilitated by first
making a simple block-flow diagram, connecting input streams to output streams.
Each block can be defined as a binary separation. Here is an example for one input
stream to be separated in four product streams. Figure 7.4 shows one block flowsheet
for this case. Table 7.2 shows all 15 conceivable alternative block flowsheets and all
25 alternative separation block functions.
In brainstorm sessions with industrial experts for each separation task, several
potential separation solutions are to be generated for each block. Typically about five
to eight separation techniques for each block can be defined, resulting in typically 100
different flowsheet separation options.
7.2.3.9 Choice of batch processing versus continuous processing

The choice of batch-wise processing versus continuous processing is important as it
determines most of the subsequent design choices. The orthodoxy is that regarding
economics, large scale processing with production rates > 5 kt/y is always continuous
and that small-scale processing < 0.5 kt/y is batch [15].
The basic reason why batch is often chosen for small scale production rates is
that most or all process steps are carried out in the same equipment. So, the amount
of equipment is reduced. For large scale processing the size of the equipment for batch
processing becomes much larger than for continuous processing as the idle time for
Tab. 7.2: All alternative flowsheets and all binary separation functions for four product streams from
one input stream.
Flowsheet label 1st Separation and label 2nd Separation and label 3rd Separation and label
FS1 A/BCD S1 B/CD S2 C/D S3
FS2 A/BCD S1 BC/D S4 B/C S5
FS3 A/BCD S1 BD/C S6 B/D S7
FS4 AB/CD S8 A/B S9 C/D S3
FS5 ABC/D S10 AB/C S11 A/B S9
FS6 ABC/D S10 AC/B S12 A/C S13
FS7 ABC/D S10 BC/A S23 B/C S5
FS8 B/ACD S14 AC/D S15 A/C S13
FS9 B/ACD S14 A/CD S16 C/D S3
FS10 B/ACD S14 AD/C S24 A/D S20
FS11 C/ABD S17 AB/D S18 A/B S9
FS12 C/ABD S17 AD/B S19 A/D S20
FS13 C/ABD S17 BD/A S25 B/D S7
FS14 AC/BD S21 A/C S13 B/D S7
FS15 AD/BC S22 A/D S20 B/C S5
switching from one operation mode to the next and emptying and filling take up time,
which is not the case for continuous processing.
The advantage of continuous over batch for large scale capacities is, therefore,
clear and holds in general.
The advantage of batch versus continuous for small production rate thus holds
for processes with several process steps.
For single process steps such as a reaction step the picture is often far more com-
plicated and the single choice of batch versus continuous processing is a whole series
of choices with dependencies.
Batch reaction processes are often carried out in mechanically stirred reactors in
which cooling is carried out by the reactor wall and by evaporating a solvent. It is
then cooled in an overhead condenser and then returned to the reaction liquid. Often
no catalyst is used, or a homogeneous catalyst is used. This is often an acid which
subsequently is neutralized by a base and then discarded as a salt. Also, the reactor
must often be cleaned in between batches.
During batch processing, the reactor is closed. If a runaway reaction occurs the
pressure build up can be high and rapid, causing an explosion. Batch processes have
shown far more explosions than continuous processes. As for the latter the reactor is
open and pressure build up will, therefore, be lower. DSM and GSK have made choice
for continuous flow microreactors for economic and safety reasons [18].
Continuous reactors are often operated in vessels with no mechanical stirrers. Of-
ten the reaction is catalyzed by a heterogeneous catalyst, placed as a fixed bed in the
reactor. In most cases no solvents are used.
This means that the choice of batch versus continuous also means the choice of
solvent or no solvent, the choice of catalyst type, the choice of process design, and the
choice of equipment type. For the latter two choices the concepts of process intensifi-
cation by combining functions in a vessel and the choice of microreactors is relevant.
The combined effect of process design choices means that even for small produc-
tion capacities of less than 1 ton/year, continuous processing in combination with the
choice for no solvent and no chemicals consumption has lower overall cost. To illus-
trate this statement, Schwalbe made a comparison for an industrial case where an
existing batch mechanically stirred tank reactor for a production of 500 kg/y of a fine
chemical was compared with a new continuously operated microreactor. The latter
has a payback time of 0.4 years and overall cost reduction of 2.3 M €/y, so a cost reduc-
tion of 4600 €/kg [19].
For cases outside the described conditions, such as 0.5 ton/y < production rate <
5 ton/y the choice between batch and continuous allows for other considerations to
also be considered. A consideration can be product quality variation. In batch pro-
cessing, product quality varies more than for continuous operation, as process control
under inherently dynamic batch conditions is less tight than under continuous steady
state conditions.
Still other choices can be made. A choice can be made to operate a certain process
step batch-wise and the other process steps continuously. A hold-up vessel after the
batch step is then needed to ensure continuous feeding to the continuous section.
The fed-batch variant in which part of the feed is batch-wise and the other part is
fed (semi) continuously is considered here the same as fully batch. The choice heuris-
tics still hold.
If batch is chosen then the fed-batch option can be chosen subsequently. This
is because, for instance, from a reaction selectivity analysis, distributed feed of one
component provides a higher selectivity.
For more detailed descriptions of batch design and modeling, the reader is re-
ferred to [20–22].
7.3 Analyze interim solutions
7.3.1 Introduction to analysis
This section describes rapid analysis methods for interim design solutions. This is par-
ticularly useful in the discovery and concept stages where many potential solutions for
products and processes are generated. A rapid analysis is then needed so that only the
best solution is further developed.
Analyzing differs from evaluating. Evaluating in this book is reserved for a fin-
ished design (for a stage). The evaluation is then used to check whether the final de-
sign meets all design criteria of the design scope. Evaluating is often done in cooper-
ation with people outside the design team, for instance by the members of the stage-
gate panel.
7.3.2 Qualitative analysis of interim solutions
In the discovery stage large numbers of ideas should be rapidly analyzed with little
information on each idea. This rapid analysis is best carried out with a group of ex-
perienced people. First, they should generate a set of criteria to facilitate the rapid
analysis. Here are three steps to generate the criteria and perform the rapid analysis.
Step 1: Ask the experts to go back in their memory to similar situations in the discov-
ery stage and then to remember which ideas were successfully implemented, and
then what features of the ideas appeared to be the key success factors. From these
success factors some criteria can then be directly defined.
These criteria are then specific to the analysis problem at hand and specific to the
company where the innovation takes place. The criteria often include capabilities
of the company to develop and implement certain novel products and processes
and not others.
Step 2: Show the experts the list of generic criteria for analysis of Table 4.8 in Chap-
ter 4. Also show them the SHEETS criteria list of Chapter 4. In addition, other cri-
teria such as strategic fit and intellectual property protection should be added,
from the business perspective.
The total set of criteria is then made from the results of steps 1 and 2.
Step 3: The experts attribute to each idea for each criterion a value of 1–5. For each
idea the total score is obtained by multiplication, rather than adding up. By mul-
tiplication, ideas that score on all criteria reasonably well obtain a higher final
score then ideas that score high on a few criteria and low on others. So, using
multiplication reveals ideas that score well over the range of criteria and thereby
these ideas are more likely to have a higher chance of success in the end.
This method of developing criteria by experts and in particular, step 1 of drilling into
the source of past experiences, is based on empirical findings on experts by Kahne-
man, which he calls: “the outside view.” It appears that even experts do not naturally
use all their experiences of the past, unless this experience is specifically called upon,
by these series of question to generate the outside view. [23, p. 245–247].
The results of this rapid analysis can then be used to go back and synthesizing
additional solutions or generating modifications to the ideas so that they score higher
on criteria that so far have a low score.
7.3.3 Quantitative analysis of interim concept solutions
In the concept stage ideas are designed in some detail. Here is an important question
regarding these interim concept solutions: is the concept sufficiently better than the
reference case for all criteria? Below some simple analysis methods for each criterion
type are provided.
The analysis result can be that the design should be further developed to facilitate
the analysis. The analysis result can be that the interim design does not meet certain
criteria. In that case the interim solution should be improved for the aspects of con-
cern. The analysis can be that the design seems to be good enough for the next design
step a full evaluation.
Product and process modeling play an important role in determining whether
the concept meets certain criteria, as with modeling detailed information on prod-
uct compositions, mass and energy flows are provided and the performance effects of
changing design parameters can be easily determined. Chapter 4 discusses the roles
of modeling and Chapter 8 and 9 provide methods for modeling, simulation, and op-
timization of products and processes.
7.3.3.1 Safety and health

A rapid analysis of safety and health issues of the novel design can be carried out by
a simple comparison of the novel design with the reference case. The novel design
should be safer than the reference case. Often this can be achieved by eliminating the
most unsafe parts of the reference case. For products safety is mostly concerned with
toxicity. Using the toxicity index for each ingredient in the product will be sufficient
to draw a conclusion. Chapter 11 provides the toxicity index method.
7.3.3.2 Environmental analysis

For a rapid analysis on intermediate results and for an assessment of concept design
results for the concept stage, a semiquantitative method based on large improvements
on a single impact type using the rapid LCA method of Jonker can be used [24].
The method consists of a few steps.
Step 1: Define the functional unit of the design as the same as for the reference case.
Step 2: Select the most critical environmental impact type.
Step 3: Compare the environmental impact of the design with the reference case.
Step 4: Conclude whether the design is sufficiently better or not.
Step 5: Improve the interim design until its meets the environmental criterion.
Chapter 12 provides more information on evaluating for environmental performance.

7.3.3.3 Sustainable development analysis

A rapid analysis on contributions to sustainable development can be done by ask-
ing whether the design provides a novel solution for the needs of people, whether it
would be socially accepted, if it would be environmentally friendly and if it would
be economically attractive. If some answers are negative, then the design should be
improved. If some answers cannot be given then more work on the design should be
done. Chapter 12 provides more information on sustainable development.
7.3.3.4 Economics rapid analysis in the concept stage

A rapid economic analysis in the concept stage can be simply based on adding the cost
of all feeds for the product and comparing that with the anticipated market price. If
the total feeds cost price is higher than the market price the design should improve on
this aspect. Chapter 10 provides several other methods for analyzing interim design
solutions.
For all other criteria no formal rapid analysis methods are available. Asking the
opinion of an experienced expert for an interim solution regarding marketing or other
criteria is a rapid method.
7.3.3.5 Role of modeling and analysis

By making a model and then simulating the design behavior, an analysis can be made
of its technical feasibility. The model also generates other output parameter values,
useful for many other analysis items. Chapters 8 and 9 provide detailed information
on how to model, simulate, and analyze product and process designs.
7.4 Best selection from alternative concept solutions
In the discovery and concept stages very large numbers of interim solutions are gener-
ated from which the most promising options should be selected with little information
at hand. For efficiency reasons these ideas should first be ranked. Only the top ideas
are then further worked out.
It is recommended that ranking includes criteria development for the discovery
stage that are generated by experienced R&D people mainly from inside the company.
In particular, the criterion for technical feasibility should be defined such that it is
meaningful in the particular company’s context, because bringing ideas to successful
implementation strongly depends on the capabilities of that company. The intuition
of those innovation experts can be considered as reliable, because it has been built up
in a “regular environment” [23]; namely the R&D area of their company.
However, it is also important to have an expert from outside the company in the
evaluation panel. Preferably it is someone with experience with innovation in more
than one company. This “outside view” as defined by Kahneman brings additional
knowledge to the table based on a broader experience [23, p. 245].
An important element of activating the intuition and memory of experts is to
prompt their memory. This is done as follows. The leader of the evaluation session
asks the experts to go back in their memory to which projects were successful and what
characteristics in the early discovery or concept stage appeared to be key success fac-
tors [23, p. 245–246]. The author has several times executed this memory prompting
by asking the experts to go back in time in their memory and think of projects that in
the end were successful. Every time the experts fell silent for minutes and then came
up with stories and appropriate criteria that even surprised themselves. They did not
know that they knew them.
First, criteria for the ranking should be generated. The generic table for problem
definition and evaluation can be used as a starting point. It is not very practical to have
many criteria. Six in total is often the maximum to keep the ranking manageable.
Safety, health, and environment may be combined for one criterion. Social accep-
tance, market opportunity, strategic fit, and economic attractiveness may also be com-
bined in a criterion. Technical feasibility to develop the idea to commercial realization
is also likely to be a criterion.
They should also carry out the description of the scaling of the criteria. A scale of
1–5 is in general sufficient for each criterion. A meaningful short statement for each
score value helps later to speed up the scoring for each criterion. According to Kahne-
man, the use of weight factors will not improve the overall quality of the judgment,
because the human mind unconsciously also makes a judgment of the outcome of the
whole and will adjust the score using weight factors. In fact making the criteria list
together and then directly asking the experts to give the total score will result in very
similar conclusions to the result that adds up the scoring of individual criteria scor-
ing [23].
The valuation of each idea or concept for each criterion is in general best carried
out by scoring the experts individually for each idea. In this way the errors of the ex-
perts are independent, so that the averaged final score value is more reliable than
when the scores are obtained by agreement. In this way, one avoids the halo effect of
people tending to form a consensus [23].
However, for complex process alternatives and for complex product ideas this in-
dividual scoring may be inferior to a plenary session in which first a discussion is
started on how the idea would perform. In this discussion each member’s imagina-
tion is stimulated to envisage how the idea would perform and then plenary scoring
is, in the opinion of the author, of more value.
The overall score is best obtained by the multiplication of the score values per
criterion. In this way, alternatives that score a reasonable value for each criterion get
a higher overall score than alternatives that score highly on one or two criteria and
low on all others. This means that the top alternatives are more robust toward future
uncertainties since they score well on all criteria and not on merely one or two. This
is a better method than adding up the individual scores as advocated by Kepner and
Tregoe [25]. The latter method compensates low scores with high scores and alterna-
tives often obtain the same total sum score, although their score distribution is very
different.
The use of weight factors for criteria is not recommended for this intuitive scoring
of alternatives, as the human mind unconsciously “corrects” the score for the weight
factors [23].
More sophisticated methods such as Analytical Hierarchy Process (AHP) and Data
Envelopment Analysis (DEA) for ranking alternatives with multiple criteria take far
more effort to obtain a ranking result. Moreover the ranking result obtained is harder
to communicate to management, who must provide the budget to further develop the
top alternative.
7.4.1 Selecting best process concept option
In the discovery and sometimes also in the concept stage, many process options are
(and should be) generated. The design is often only a simple block-flow diagram. In-
puts and outputs are known to some degree from scouting experiments. Process con-
cept alternatives will be available from literature searches. For efficiency reasons these
concepts should be ranked and only the best selected for further development. Rank-
ing with a panel is, however, not always immediately available. For the individual de-
signer the following simple method has been developed to still select the best option.
The simple selection method is carried out as follows. A large number of process
alternatives are obtained from literature searches. Each alternative solution is then
first tested whether for any of the safety, health, environment and social acceptance
aspects there is a major problem foreseen. If so the alternative is filed and for the
time being not further investigated. For the remaining alternative process alternatives,
block-flow diagrams are made and the yield of product on feedstock is researched in
literature. Then an economic ranking is made based on this design information, as
provided in the next paragraphs.
7.4.2 Economic ranking of process concepts
For quickly economically evaluating process routes based on different conversion

types such as thermal, heterogeneous catalysis, homogeneous catalysis, enzymatic
catalysis or microbial a semiquantitative method has been developed [18]. It is based
on historic statistics of process cost element distributions as found in Perry’s chemical
engineers’ handbook [13].
These statistics show that the major process cost element is the feedstock cost. If
the process routes are all from the same feedstock then the product yield on feedstock
cost is the only parameter value needed as indicator for feedstock cost for the different
process routes. If the feedstock differ then the feedstock cost price should be looked
up. The feedstock cost per ton of product is then obtained from
Feedstock cost/ton product = Feedstock price ($/ton)/yield
The other process costs such as investment cost charge, maintenance, utilities, and
operations are all strongly related to the major pieces of equipment [29] and these are
strongly related to the number of process steps for conversion and separation. The
number of process steps is derived from the number of conversion steps and the mini-
mum number of separation steps involved. The latter is directly derived from the stoi-
chiometric equations showing what byproducts will minimally be formed, and hence
what minimum separations are needed. If a solvent or a homogeneous catalyst is used
then a separation step for their recovery should be added [26].
If the reactions form an equilibrium between product and feedstock then also an
additional separation will be needed. Whether an equilibrium is present can be estab-
lished from Equation (7.1) for the equilibrium constant K [28]:
K = e− RT
∆G
(7.1)
The value of the free enthalpy value, ∆G, can be looked up and K determined. If K is
much larger than 100, then single pass deep conversion is feasible. If K is around 1
then single pass conversion is not feasible and an additional separation step has to be
added to the total separation steps.
The selection method is then as shown in Figure 7.5: for each alternative feedstock
the yield figure is noted down. If there are feedstock type differences between the al-
ternatives then the feedstock cost per ton of product is determined from the feedstock
price and the product yield on that feedstock. The total number of process steps is
also determined from the stoichiometry equation based on the reaction, the equilib-
rium value K, and whether separation is needed for homogeneous catalysis or solvent
recycle.
The best process route has the lowest feedstock cost per ton of product and the
lowest number of process steps. If a best selection cannot be made, because the lowest
feedstock and the lowest number of process steps belong to different alternatives then
the lowest feedstock cost option should be chosen. In all other cases the top two or
three options should be further investigated in detail, by more detailed process design
and cost analysis, using information from Chapter 10.
An example of a process concept design with the essential information needed for
the ranking method is shown in Figure 7.6 [26, 27].
Main: Feedstock cost Yield

Second: Investment cost Process steps: Ntotal
Document
Yield overall
Analysis Nreactions + Nseparations
Info Reactions? Yes Reactions
Safety H. E. ? Stoichiometrics Rank: Highest Yield
Lowest Ntotal
No No
File Best?
Yes
Selected option
Fig. 7.5: Industrial process route selection method.
Catalyst + water
Glucose DHA lactic acid

Catalyst: Homogeneous
Nreact = 2 Nsep = 3
Glucose
solvent: water
Yield: 93 % max theory
Lactic a. Acrylic a. + H2O Acrylic

Catalyst: Heterogeneous
Acid
Solvent: Lactic acid
Water
Lactic acid
Fig. 7.6: Block flow diagram to acrylic acid: Route C. Reference: C.B. Rasendra, Platform chemicals
from biomass, PhD thesis, U. Groningen, 2012.
7.5 Finalizing a design
7.5.1 Introduction
This section is about modifying the design such that:

(a) All criteria must be met.
(b) The design is optimized using the desired criteria.
(c) The design is modified such that the risks are evenly spread over the criteria, by
which the overall risk of failure is reduced and robustness is obtained.
7.5 Finalizing a design | 155
This design action is called a balancing design, or trade-off design. This design action
is mainly situated in the feasibility stage, where for the first time a commercial scale
design is made that meets all criteria, defined as a feasible design. This feasibility is
strongly related to minimizing the risk of failure; hence the focus on minimizing the
overall risk.
The subsequent sections deal with criteria types, optimization over all criteria,
and increasing robustness toward future uncertainties by using scenarios.
The design criteria set in the scoping step of the design are often further defined
as the criteria that must be met, otherwise the design will not be accepted, and desired
criteria for which the design should be optimized.
7.5.2 Balancing designs
When a design nears its completion, a lot of knowledge will have been generated re-
vealing the relations between design decisions and design performance. Then it is
time to balance the design such that it becomes robust to uncertainty in the value of
certain parameters such as feedstock prices. This robustness can be obtained to a cer-
tain extent by choosing design parameters such that the design output or performance
is away from the most critical constraint boundaries and that the optimum is a smooth
optimum and not peaky. Polar graphs are very useful in communicating the balanced
design. For showing the effects of uncertainty ranges in design parameters on the de-
sign performance and showing that even for the uncertainty ranges the design stays
within its constraints, Tornado diagrams can be useful.
7.5.3 Increasing robustness toward future uncertainties using scenarios
Because the design will be implemented in the future and may last many years it is
important not only to take the present status of the evaluation criteria into account
but also their future development. If this is not done then the product and process
can rapidly become out of fashion and may need to be ended. This in turn then leads
to loss of capital and loss of potential revenues.
One way of looking at the design in view of future uncertainties and by doing so
making it more robust to these uncertainties, is by using a set of opposing scenarios.
An example of such a set is provided in Figure 7.7, obtained from IBM. It shows four
scenarios generated from using two axes with opposing society trends. The horizontal
axis is about possible nature and industry trends, either being fragmented or consoli-
dated and the vertical axis is about customer preferences either being individualistic
or homogeneous. With these axes, four scenarios are built. Each scenario is internally
consistent and plausible and has elements that are relevant to the design.
Customer:
Individualistic Proactive
Local Flavor World of Extremes

• Customers personalized • Super shoppers
values • Integrated global supply
• High innovation level chains
Industry: Industry:
Fragmented Global
Disconnected Integrated
Behind Walls Sameness sells
• Localized shopping • Customers price-conscious
• Fragmented competition • Megascale competitors
Customer:
Homogeneous Passive
Fig. 7.7: Scenario set example.
The design is placed in each scenario and conclusions are drawn under each scenario
how well the design would behave. If the design behaves well in each scenario it is
likely that the design is robust toward future uncertainties. If the design behaves badly
under at least one scenario then the design is not robust. If that is the case the design
team should consider changing the design, or making it adaptable, in such a way that
it will be successful under all scenarios, and hence is made more robust toward future
uncertainties.
Scenario sets may be obtained from literature sources such as the United Nations
or big enterprises such as IBM or Shell. Companies or design teams can build their
own scenarios using the method of opposing potential trends and generating plausi-
ble worlds for each quadrant.
7.5.4 Intellectual property (IP) creation and protection
Intellectual property is created by design when the novel design is drawn and written
down and the content is protected from use by others. Designs of a company can be
protected in various ways. It can be done by patents, copyright, or design rights. Chap-
ter 3, Section 3.6.5 provides more information. Most important is to have IP protection
before the design is made public in any way.
7.6 Evaluating and reporting designs | 157
7.6 Evaluating and reporting designs
7.6.1 Introduction
Evaluating a design means that the design result is checked whether it meets all de-
sign goals and design criteria. If several designs have been made than each design
is individually checked in this way. This checking will be done by people outside the
design team. The information for which this checking is carried out, however, is pro-
vided by the design team. So the quality of the evaluation is the responsibility of the
designers and of the evaluators.
The purpose of this section is then to provide an overview of evaluation items for
which the design team must provide information. A very important part of evaluat-
ing any design is comparing it with the reference case, being the existing commercial
technology for the same or similar purpose (function). Therefore, first guidelines for
selecting and defining the reference case are provided in Section 7.6.2. Then guidelines
for evaluations are provided in Section 7.6.3.
7.6.2 Reference case for evaluations
A reference case is the present commercial product with the same function as the novel
product. If the novel product fulfills a function that presently is fulfilled by more than
one product then all those products combined are taken as the reference case. The ref-
erence process is the best commercial scale process for the same product or function.
Having such a reference case to compare with the novel product and process is a
powerful communication means to the evaluation panel. For each criterion, the novel
product and process is compared with that present commercial product and process,
and the novel design must compete. All advantages and disadvantages of the novel
product and process are directly clear.
Finding all relevant information about the reference case may take considerable
effort. For reference products it may take a full chemical analysis of the competitors’
products as well as performance tests. For a reference process it may involve reading
public environmental impact reports of the competitors’ process. In some cases it may
even involve making a design of the competitors’ process from limited public informa-
tion available. However, this effort will pay off knowing the present performance of the
best competing product and process so that improvements by the novel product and
process can be determined for all evaluation criteria.
7.6.3 Evaluating designs
Evaluating designs in a commercial organization will in general be done at the gate

from one innovation stage to the other. Criteria for the evaluation in general will con-
cern:
– Business aspects, such as strategy fit and economics (foreseen benefits, foreseen
cost, market share)
– Safety, health, environment and social aspects
– Technical feasibility
Chapter 3 provides for each stage (gate) details on what information should be pro-
vided and in particular what experimental proof is needed for each stage gate to be
able to judge the technical feasibility of the innovation. Specific evaluation methods
for economics, health and safety, environmental, social, and sustainable development
are provided in Chapter 10–13.
7.6.4 Reporting
Any design consists of words, symbols, drawings, and pictures to report the design
result to others. So reporting orally and in writing is extremely important. Especially
in the discovery stage this truth is sometimes forgotten. A brilliant idea may be men-
tioned in a discussion in the corridor. If it is not written down and reported to others it
will be lost. Chapter 13 elaborates extensively on the importance of reporting designs
and provides guidelines for doing so for each part of a design in each stage.
[1] Verkerk MJ, et.al. Philosophy of Technology, London: Routledge, 2007.

[2] Tassoul M. Creative facilitation. Delft: VSSD, 2009.
[3] Korevaar G. Sustainable Chemical Processes and Products: New Design Methodology and
Design Tools, Delft: Eburon, 2004.
[4] Altshuller G. TRIZ keys to Technical Innovation, Worcester USA: Technical Innovation Center,
2005.
[5] Harmsen GJ. Chewing gum that does not foul the streets, oral presentation and abstract Pro-
ceedings ECCE 2007, Copenhagen, 2007, Conf CD ISBN 978-87-91435-57-9.
[6] Bermingham SK. A design procedure and predictive models for solution crystallisation pro-
cesses. Development and application, PhD thesis, Delft: Delft University Press, 2003.
[7] Bruin S. Phase equilibria for food product and process design, Fluid phase equilibria, Vol. 158–
160, 657–671, 1999.
[8] Wesdorp L. Good food requires engineering?! Simultaneous product and process development,
Keynote lecture on ESCAPE-12 Symposium, The Hague, on CD-ROM supplement in: Grievink J,
van Schijndel J (eds.). Computer-Aided Chemical Engineering, Vol. 10, Amsterdam: Elsevier,
2002.
[9] Cussler EL, Moggridge GD. Chemical Product Design, Cambridge: Cambridge University Press,
2001.
[10] Wesselingh JA, et.al. Design and Development of Biological, Chemical, Food and Pharmaceuti-
cal Products, Chichester, John Wiley & Sons, 2007.
[11] Seider WD, et al., Product and Process Design Principles, Hoboken NJ: John Wiley & Sons,
2009.
[12] Stankiewicz AJ, Moulijn JA (eds.). Re-engineering the Chemical Processing Plant, New York:
Dekker, 2000.
[13] Green DW, Perry RH (eds.). Perry’s Chemical Engineers’ Handbook, Eighth Edition, New York:
McGraw-Hill, 2008.
[14] Levenspiel O. Chemical Reaction engineering, 3rd edn, New York: J. Wiley& Sons, 1999.
[15] Douglas JM. Conceptual Design of Chemical Processes. New York: McGraw-Hill International
editions, 1988.
[16] Schembecker G. Process Synthesis Course, Breda, Process Design Center, 2000.
[17] Harmsen GJ. Reactive Distillation: The frontrunner of Industrial Process Intensification: A full
review of commercial applications, research, scale-up, design and operation, Chemical Engi-
neering & Processing: Process Intensification, 46(9), 2007, 774–780.
[18] Harmsen J. Economics and Environmental Impact of Process Intensification: An assessment for
the petrochemical, pharmaceutical and fine chemicals industries, Chapter 14, Process Inten-
sification for Green Chemistry, KVK Boodhoo and AP Harvey (eds.), Chichester: John Wiley &
Sons, (2013).
[19] Schwalbe T. Chemical Synthesis in Microreactors, Chimia, 56, 2002, 636–646.
[20] Diwekar U. Batch Processing Modeling and Design, New York: Taylor & Francis Inc, Abingdon,
2014.
[21] Sharratt PN. Handbook-of-batch-process-design, New York: Springer, 1997.
[22] Korovessi E, Linninger AA. Batch Processes, Boca Raton: CRC Press, 2005
[23] Kahneman D. Thinking, Fast and Slow, London: Penguin Books, 2012.
[24] Jonker G, Harmsen J. Engineering for sustainability A practical guide for sustainable design.
Amsterdam, Elsevier, 2012.
[25] Kepner CH, Tregoe BB. The rational manager, Princeton NJ: Kepner-Tregoe Inc., 1981.
[26] Harmsen J. How to choose a process route: From the lab to production, NPT Procestechnologie,
2015, 16–17.
[27] Harmsen J. Best Renewable Routes to Adipic Acid, Acrylic Acid and Propionic Acid, Oral and ab-
stract, AIChE Spring Meeting & 12th Global Congress on Process Safety, Houston, USA, 2016,
10–14.
[28] Moore WJ. Physical Chemistry, 4th edn, Englewood Cliffs N.J.: Longmans Green and Co., 1962.
[29] Perry RH, Green D, Perry’s Chemical Engineers’ Handbook, Eighth Edition, New York: McGraw-
Hill, 2008.
8 Product modeling and optimization
In this chapter, not only the role of mathematical models in the design of products is
described, but also the roles of other types of models: verbal, schematic, and physical.
The information in this chapter is applied in the design levels and design steps of Delft
Design Map (DDM) as indicated in Table 8.1.
Tab. 8.1: Delft Design Map (DDM) parts related to modeling and optimization of Chapter 8
Stage-GateTM : Concept Feasibility Development

Fram- Supply chain Process tech- Process Final
ing imbedding nology engineering
Design levels ⇒ PF CW-PQ PQ-PF I-O SP TN UN PI ED OI FO (Final)
⇓ Design steps/
activities
1 Scope
2 Knowledge
3 Synthesize
4 Analyze
5 Evaluate
6 Select
7 Report
8.1 Verbal, schematic, mathematical, and physical models
In Section 4.6 the role of (mathematical) modeling in design is described. This section
also emphasizes the use of nonmathematical types of models for both product and
process design activities. Before going into these different model types, the creation
of models by designers is the heart of the design activities, as all these models together
make up/describe the design alternatives and the final design (be it from different per-
spectives).
This has been characterized by Van Overveld [1] (lecturer in general design
methodology at TU Eindhoven) and in his very inspirational way has taught this
in the TU Delft and TU/Eindhoven post-Master PDEng designer programs (see Sec-
tion 14.2.4). Van Overveld characterized ‘designing’ and ‘a design’ as:
– “Designing is: making decisions to make stakeholders happy, according to an in-
tentional plan, leading to a new ‘artefact’, an improvement to an existing artefact
or a new or improved part of an artefact.”
– “‘A design’ is a representation of the (new/improved) artefact, in the form of one
or more models, that are mutually consistent. Each model focuses on one aspect
https://doi.org/10.1515/9783110467741-008
8.1 Verbal, schematic, mathematical, and physical models | 161
of the artefact (geometry, physics, chemistry, operation, . . . ) and communicates

the artefact with the stakeholders, documents the decisions, helps the realization
and aids in the use of the artefact.”
Different model types are created and used also in the (bio)chemical product and pro-
cess design field. These models describe, in their own way, parts of the design results
and decisions. The most used model types are also classified in general systems theory
by e.g., Skyttner [2]. Next to mathematical models, iconic/physical models, schematic
models, and verbal models are also used. Engineers get well exposed to mathemati-
cal models and schematic models like graphs and charts, but receive less education
and training in creating schematic and physical/iconic models. Reporting on design
context, scope, knowledge, analysis, evaluation, and selection decisions is not recog-
nized as constituting a model as well. All these models serve different purposes and
are used during and after the design process. For communication purposes the phys-
ical and schematic models play a paramount role.
Viewing the design as a collection of (mutually) consistent models also helps in defin-
ing the scope and deliverables for the design team (in specifying what types of models
and to what level of detail these models will be delivered).
Examples of these four model types will be given in Sections 8.2 and 8.3 and in
Tables 8.5 and 8.6, and references are given of the examples of these models used in
this book.
– Physical (or iconic) models. These models:
– Look like the real system they represent (car, building, manufacturing plant,
product prototypes, mock-ups)
– Can be small scale or full-scale
– Schematic models. These models are pictorial representations of conceptual re-
lationships of the design:
– Molecular structures drawings, (bio)chemical reaction pathways
– Product structure drawings (spatial distribution of phases with each individ-
ual size, shape, form)
– Product structure drawing in different stages of manufacturing or use
– Graphs showing relationships (suggested, measured) between product and
process variables (reaction rate versus concentrations, heat transfer rate ver-
sus temperature differences, payback time versus capital investment)
– Tables (containing symbols and data); pure component properties, stream
specification, process stream summary
– Diagrams (House of Quality, process block diagram, process flow diagram)
162 | 8 Product modeling and optimization
– Verbal models. These models contain words (organized in lines of text or tables)
that represent systems or situations that (may) exist in reality. These words pro-
vide:
– Design problem context
– Relevant knowledge to solve a problem, analyze it, and evaluate performance
– Assistance in construction of (parts) of the design (equipment CAD drawing)
– Assistance in operating, maintaining, and repairing manuals
– Justify design decisions and recommendations
– Organized in lines, bullet lists, (text) tables, paragraphs, chapters
And finally:
– Mathematical models:
– Are a reduced representation of reality
– Most abstract and precise models in a structural language
– Such models are systematic and precise indeed, but they will often be re-
stricted with respect to their domain of valid application. I.e., the underly-
ing modeling assumptions must be applicable in a particular domain and,
ideally, an experimental validation of the model or of its critical submodels
must have taken place.
These models of an artefact (verbal, iconic, schematic, mathematic) are representa-

tions of (parts of) the original artefact. A model has a correspondence to the actual
artefact and is created/synthesized to solve design issues/problems of the real arte-
fact. The design issues can be analyzed/evaluated by manipulating the model (be-
cause manipulating the actual artefact is not possible, as it is not there yet). The mod-
els should be so good a representation of (part of) the real artefact, that results of the
manipulation of the model can be mapped back to the artefact. The complete set of
models represents the artefact’s design.
8.2 Process design schematic and mathematical models useful

for product design
In (bio)chemical product and process design all four types of models are used, but
only some of them have become standard tools. Chemical or process engineers – his-
torically educated and trained in process design and mathematical process modeling
and simulation – have been exposed to the use of a limited class of schematic models.
Chemical engineers value mathematical models very much as a tool in process
design and are extensively using them in process and equipment design and have
so for a long time. The mathematical modeling of manufacturing processes (com-
posed of unit operations like chemical reactions and separations, including thermo-
8.2 Process design schematic and mathematical models useful for product design | 163
dynamic, kinetics, and transfer phenomena models) has a long history and is exten-
sively used in process simulators and mathematical optimization software in all inno-
vation phases. This will be extensively discussed in Chapter 9. In the initial innovation
phases, mathematical models start in a simple form (e.g., mass balances based on re-
action stoichiometry and conversion/yield assumptions) and develop in complexity
for those processes that show potential to succeed and warrant more detailed and
extensive modeling. Said mathematical process models are often linked to engineer-
ing diagrams or drawings like process block schemes (PBS), process flow diagrams/
schemes (PFD/PFS), and piping and instrumentation diagrams (P&ID). These draw-
ings – often produced to internal and/or external standards, may form the basis for
mathematical modeling or a vehicle to capture the mathematical models’ quantita-
tive results.
Examples of the PBS and PFD of the TAME (Tert-Amyl Methyl Ether) production
process are presented in Figures 8.1 and 8.2. 3D CAD process and equipment drawings
are also used widely by process design and engineering companies.
Process Water Bleed

N.N.F.
2,000 t/a (0.02)
Hydrogen
1,700 t/a (0.02)
Reaction Product Methanol

Raffinate 2
Raffinate 1 Section Purific. Recovery 232,200 t/a (2.83)
283,500 t/a (3.45)
7 Bara 322,100
2.5 Bara 240,000
2.5 Bara
Methanol
27,100 t/a (0.33) 70-82 oC t/a 44-120 oC t/a 46-129 oC
(3.92) (2.92)
9,800 t/a (0.12) Methanol Recycle

9,800 t/a (0.12)
TAME product
82,100 t/a (1.00)
Total IN: Total OUT:

314,300 t/a (3.83) 314,300 t/a (3.83)
Please note:
• Values between brackets ( ) are t/t values.
• Blocks and character fonts could have still smaller size.
• Values normally are round off, however without loosing consistency on balance.
• All streams ENTER at the LEFT hand side of the block scheme.
• ALL streams LEAVE at the RIGHT hand side of the block scheme.
• N.N.F. or NNF = Normally No Flow.
• Indicate temperature and pressure inside blocks and for all streams entering (LEFT) and leaving
(RIGHT) the block scheme (not done in the above scheme).
• Indicate stream numbers for each stream (cross-referencing with Process Stream Summary).
• Never put arrows at the end of the connecting lines; for better readability of the scheme put the
arrows at the middle of the line.
Fig. 8.1: Process Block Scheme (PBS) example – TAME (Tert-Amyl Methyl Ether) production pro-
cess [3].
CW Raffinate II
20
TC CW
E03
12 13 K01
LC
V01 E07
PC
14 2.4 361 23 LC
C01
P04 V02
C02
2.5
315 FC
C03 P07 P08
CW 21 FC
PC 24
PC
TC
T01 E01 R01 R02 T02 27 361 27
7
TC
E05
CW ?? CW
7 7 ?? 24 330 2.5
24 CW Bleed
10 TC FC
770 2.5 317 17
FC
77 5 4 PC ??
3 343 LC
2.5 298 12 E06 TC NNF
Hydrogen 7 TC TC TC
7 CW
FC LC 19 LC LC
303 E02 330
2
E04
7 318 6 2.5 2.5
Raff. I 7
E08
P01 FC 319
303 ?? 15 22
1 8 9 ?? 393 E09 25
25 402 2.5 11 361 11
7
Methanol P02 343 352 328 P05 2.6 P06 P09
3 TC
FR 7 7 2.5 CW
298
FC
23
11
Process P03 Methanol Recycle 361 2.1
Water
7 313 TAME
28
Process Equipment Summary Designers Process Flow Scheme - Improved Solution
C01 : Tame RD-Column E05 : Abs. Cooler P02 : Raff. I Feed P09 : Bottoms C03 Project : Tertiary Amylene Methyl Ether (TAME) from
C02 : Methanol Abs. E06 : Abs./Feed HE P03 : Proc. Water Feed R01 : Tri-Funct. React. Iso-Amylene and Methanol
C03 : Methanol Strip. E07 : Condenser C03 P04 : Reflux C01 R02 : Boiling Pnt. React. Anonymus, c.s.
E01 : Aft. Cooler T01 E08 : Reboiler C03 P05 : Bottoms C01 T01 : Exp. Hydrogen Proj. ID Number : CPDxxxx
E02 : Effl. Cooler R02 E09 : TAME Cooler P06 : Bottoms C02 T02 : Exp. R02 Effl.’s Completion Date : January 1st 1998
E03 : Condenser C01 K01 : Ovhd C01 P0 : Reflux C03 V01 : Reflux Accu C01
E04 : Reboiler C01 P01 : Methanol Feed P08 : Methanol Recycle V02 : Reflux Accu C03 Stream number Temp. (K) Pressure (Bara)
󳶣 Fig. 8.2: Process Flow Diagram/Scheme (PFD/PFS) example – TAME (Tert-Amyl Methyl Ether) pro-
duction process [4].
These engineering diagrams and drawings also play a key communication tool within
the design team, but also to other stakeholders outside the team. They form key input
documents for preliminary economic analysis (gross margin calculations based on
product production and raw materials usage, see Chapter 10), for equipment lists and
equipment data sheets (PFD) and for HAZOP studies (P&ID) (see Chapter 11) to name
a few. The more detailed 3D CAD equipment and plant drawings are used for layout
optimization and equipment and plant construction purposes. As a communication
tool often between different companies (brand owner, engineering and procurement
companies, equipment manufacturing and plant installation companies) these stan-
dardized drawings are widely spread.
The standardization of engineering drawings and their standardized use have not
yet cascaded down to the design of chemical products. Chemical products that are
process-like devices (fuel cells, lab-on-chip, sensor-on-a-chip) are also easily repre-
sented by the above tools, as the device function and performance delivery resemble
a chemical process.
For other products (like biofuels, engine lubricants, coatings, laundry detergents,
flexible packaging, etc.) product drawings are much less standardized (as they are
often only used in a limited range of stakeholders) and are, therefore, not very widely
applied.
For product design teams, drawings form a very important – but often vastly un-
derestimated – tool for new product design and development. It is mainly in the de-
sign knowledge capture and the communication thereof, that the roles of drawings/
sketches are paramount. Drawings are key as the first mental models representing the
interactions of the product (and its components) with its surroundings during its en-
tire life cycle, not only for visualizing these key product component interactions, but
also as a basis for mathematical modeling and communication, an analysis and syn-
thesis vehicle for generating the optimal product design. Therefore, the skill of making
drawings/sketches in product design is viewed as a key competence every designer
should learn to master.
8.3 Product design schematic models
8.3.1 House of Quality model for consumer function and property function
In reviewing the different product and process design phases (as presented in Sec-
tion 4.3), the use of product related drawings/sketches and their links to mathematical
models will be shown here.
Fig. 8.3: Consumer and property functions in the consumer wants and product attribute design
spaces [5].
The first design levels after the project framing level are: ‘Consumer Wants–Product
Concept’ and ‘Product Concept – Product Function.’ In the ‘Consumer Wants– Product
Concept’ level the translation of consumer needs and wants (voice of customer) into
quantifiable product attributes takes place. These needs, wants and attributes can be
vastly different for any (chemical) product. The Quality Function Deployment (QFD)
method using the House of Quality (HoQ) design tool has been applied very success-
fully for this.
Almeida et al. [5] present a nice overview of the relationship between consumer
wants/perception (great taste, scoopable directly from freezer, healthy) to the property
function for an ice cream example (see Figure 8.3).
In this diagram, these consumer preferences for ice cream are first translated by
a ‘consumer (or quality) function’ into quantifiable product attributes (e.g., smooth-
ness, freshness, softness, stickiness). The consumer liking/perceptions of a product
are mostly obtained through questionnaires and interviews. Careful preparation of
these customers’ interviews requires a good understanding of the product in use. Es-
pecially for product innovations it is good practice to visualize the various customer/
product interactions. This starts, for instance, with the purchase of the ice cream in
the supermarket, the transport and storage prior to consumption, the consumption
activities, and final waste disposal. The product attributes are scored against an arbi-
trary scale that represents the relative performance of the attribute against consumer
expectations.
In the second step these quantifiable attributes are then linked by a ‘property
function’ to measurable product properties based on the composition and product
microstructure. For the ice cream case study, product properties include: water crys-
tal size and volume fraction, fat crystal size and volume fraction, air cell sizes, and
volume fraction, etc. A schematic representation of an ice cream’s microstructure is
depicted in Figure 8.4.
The flow of design decisions runs from consumer product qualities to the process
function and the selection of raw material feeds. This flow is the reverse of the cause
and effect chain of physical events when making a product. While the logic of the flow
Fig. 8.4: Schematic representation of the

microstructure of ice cream [5].
of design decisions is in itself impeccable, a practical information gap problem arises.

This problem is due to the ever-increasing degree of freedom when going with the de-
sign decision flow, i.e., in every step of determining functions (consumer ⇒ property
⇒ process) there are relatively few specifications of what the outcomes of a function
should be, while there is an excess of inputs to a function. Consequently, there is an
abundance of degrees of freedom. The question arises how one can handle or fix this
degrees of freedom excess? This information gap problem can be solved in a mean-
ingful way when a lot of side information is added per function. That is, additional
restrictive requirements on the function inputs must be imposed, along with an opti-
mization of the performance of a function. Finding meaningful restrictions and setting
suitable optimization goals are essential parts of the design process.
The quality function deployment (QFD) method can be used to translate con-
sumer wants (consumer voice) into product attributes requirements. As presented
by Almeida et al. [5], this translation occurs at the initial matrix of the QFD, termed
House of Quality (HoQ)-1 (Figure 8.5). In Figure 8.6 the two subsequent steps (HoQ-2
and HoQ-3) translating consumer wants via product attributes and (physical) product
properties to the process function requirements are shown. House of Quality tem-
plates for Microsoft Excel are available at the QFD ONLINE website [6]. Seider et al. [7]
also provide good insight in how to apply the QFD/HoQ methodology.
8.3.2 Understanding the product application/use process: visualizing systems

splits
In support of the House of Quality QFD method, it greatly helps to visualize and draw
the product in the environment(s) where the product will be applied or used. The
mouthfeel, creaminess, and iciness (sensory) product attributes should be met and
not changed during the life cycle phases prior to consumption. More importantly, by
quantifying the different conditions across the manufacturing and in-use processes
Fig. 8.5: Schematic representation of the use of House of Quality (HoQ) 1 in the Quality Function
Deployment (QFD) methodology translating consumer wants (voice of consumers) into product at-
tribute requirements [5].
(temperature and relative humidity during storage, transport, display, shopping,

home storage, and consumption, including time of exposure, etc.) potential risks for
product quality can be identified and modeled. If prolonged exposures to higher tem-
peratures prior to the ice cream consumption are identified, product/environmental
interaction drawings and mathematical models can be used to quantify, substantiate,
and mitigate this risk. Transport phenomena models involving the ice cream material,
the ice cream wrap, carton packaging box and the heat transport models can be used.
Relevant physical transport and material properties (heat conductivity in moist air,
ice cream, packaging material, density and specific heat, heat of melting, etc.) come
into the equation.
By applying the approaches in this section and Section 8.4.1, the understanding
and quantification of the key product application process mechanism can greatly en-
hance product (re-)design.
Wesselingh et al. [8] emphasize the essentials of product/systems interaction and
promote drawing system splits to start the modeling approach. He advocates making
different splits of the product/system interactions to get an overview of the product in
use and during the different phases of its lifetime. He uses the example of a laundry
washing machine and a laundry detergent product. Different systems can be chosen to
get different viewpoints on this system and to make simple mathematical models. An
1
HOW?
Product attributes
requirements
Customer requirements
2
WHAT?
HOW?
Product properties
requirements
Product attributes
3
WHAT?
HOW?
Process requirements
Product properties
WHAT?
Fig. 8.6: Schematic representation of the use of Houses of Quality (HoQ) 1, 2 and 3 for translating
consumer wants into product attributes, product properties, and process requirements.
input/output mass and energy balance analysis of washing machine (inputs: soiled
dry textile, clean cold water, laundry detergent, electricity power; outputs: wet clean
textile, dirty water [containing soil and detergent components], heat losses).
Another way to split this system is to subdivide the laundry process into process steps,
such as: add laundry, add detergent into dispenser, add water, heat-up, wash, rinse,
centrifuge, and take out laundry. The system can be quantified by taking simple heat
and mass balance equations and using system variables like mass and temperature of
textile, water, detergent, required washing temperature, power supply, heat-up rate,
washing time, rinsing time, and centrifugation.
A closer look can be taken to examine the textile composition and structure. How
is the textile structure: solid fibers, spun into yarns, or woven into textile? What are
the volume fractions of voids in the yarns and between the yarns in the woven textile?
One can quantify the fiber density and diameter, the yarn density and the total surface
area of the fibers that needs to be wetted and covered with surfactant molecules for
the efficient removal of soil particles or stains.
The interaction of the detergent product in water and in contact with the textile
in the washing machine can also be visualized. The textile tumbles repeatedly into
the water, and water (with dissolved surface active components) (partly) refreshes the
liquid between the fibers and yarns. Through the bending and stretching of the tex-
tile the replacement of dirty liquid by fresh liquid is taking place, and through the
surface-active activity of the detergent ingredients soil is removed from the surface of
the fibers and is kept suspended in water. Antiredepositing ingredients avoid the re-
settling of the soil on the fibers. The liquid water refreshing rate can be computed/
estimated based on a simple transport phenomena model for the convective flow of
water in/out of the textile between the yarns (by bending), and the much slower trans-
port of water/dirt/chemicals into the yarns (between the fibers) by diffusion. The rate
determining step for the overall process is the diffusion in the yarns to the fibers (as
can be derived from regime analysis and characteristic times estimations (covered in
Section 8.4.1).
The soil, stains, and the detergent product and structure itself can also be sketched
and modeled. They can be split into different stain types (soluble and insoluble par-
ticles, fat/oil, anionic surfactants when not scavenged). Identification of molecular
identity, quantities, and dissolution behavior in water are all important. The sketches/
drawings can capture and visualize the key phenomena to be considered in the use
and life cycle phases of the product.
8.4 Mathematical models for consumer and property functions
In this section, different approaches to create mathematical models for consumer

function and product property function are presented.
8.4.1 Characteristic times and regime analysis
A very convenient way of quantifying the different processes/mechanisms taking

place in applying the product and finally delivering the performance is to compute
the so-called characteristic times (or time constants) for these mechanisms. In the
1960s this practice of developing time constants for rates of changes in chemical
engineering systems developed at the disciplinary interface of physical transport
phenomena and process dynamics. In the 1980s this methodology of determining
characteristic times within a process was applied to the scale-up of bioreactors, by
using the ‘scale-up by scale-down’ approach. This method was used to calculate the
various characteristic times for mechanisms (mixing, mass and heat transfer, oxy-
gen depletion, product formation, and biomass growth) in stirred aerated fermentors
(Oosterhuis [9], Sweere [10]) and to establish the rate-limiting mechanism(s). The
rate-limiting mechanism determines the regime such as the oxygen or substrate lim-
iting regime in (aerobic) fermentations. The process was researched and optimized in
a scaled-down fermentor in which the rate-limiting mechanism(s) on this laboratory
scale were the same as in the large-scale fermentor to be designed.
This approach has been further developed and taught over the years by Kossen,
Luyben, Heijnen, and Van der Lans of TU Delft’s biotechnology department, and is still
taught to MSc and PDEng students in Delft University of Technology’s biotechnology
and chemical engineering fields.
This regime analysis method using characteristic times is also very versatile for
understanding and improving product application/use processes and the design of
the microstructure and composition of the product.
Characteristic times or time constants can be defined as the ratio of a capacity
over a corresponding rate, where capacity and rate are to be used in a rather general
Tab. 8.2: Characteristic times for transport phenomena and (chemical) kinetics.
Time constant [s] Nominator Denominator

Travel time: τ = L/v L [m]: travel length v [m/s]: travel velocity
Residence time: τ = V/F V [m3 ]: volume F [m3 /s]: volume flow rate
Diffusion time: t d = L 2 /D L [m]: diffusion length D [m2 /s]: diffusion coefficient
Conversion time:
General:
– t c = (CV)/(rc V), t c = C/rc V [m3 ]: volume V [m3 ]: volume
C [kmol/m3 ]: rc [kmol/m3 /s]:
– reactant concentration – conversion rate per volume
Zero order: t c = C/k C [kmol/m3 ] k [kmol/m3 /s]:
– 0-order rate constant
1st order: t c = 1/k 1 [–] k [1/s]:
– 1st order rate constant
Mass transfer time
From gas bubble into liquid:
– t mt = 1/(kl a) 1 [–] kl a [s−1 ] or [(m/s) ⋅ (m2 /m3 )]:
Depletion of gas bubble: – kl : liquid side mass transfer
– t mt = m/kl a m [–]: ratio C g∗ /C l∗ coefficient [m/s]
C g∗ : reactant concentration in – a: specific surface area
gas corresponding to the [m2 /m3 ]
reactant saturation
concentration in liquid (C l∗ )
Heat transfer
(equilibration) time:
– t ht = Vρc p /(UA) Vρc p [J/K] UA [W/K] [(W/(m2 ⋅ K)) ⋅ (m2 )]
[(m3 ⋅ kg/m3 ) ⋅ (J/(kg/K))] (1/U = 1/α I + (ρc p )/λ + 1/α e )
Momentum transfer time (shear
stress relaxation):
– t shear = (L 2 ρ)/η L [m]: shear layer thickness η: [kg/(ms)] viscosity
ρ [kg/m3 ]: density
way. Some examples of time constants that can be derived from differential mass and
energy balances are presented in Table 8.2.
Characteristic times for the laundry detergent example in the previous section can
be calculated. For the steps involved in the washing process characteristic times can
be estimated for: heating up water, heating up textile, disintegration and dissolution
of the detergent powder into the water, diffusion time of chemicals/water into yarns,
time for convective transport of water/chemicals into and out of the woven yarns, en-
zymatic conversion of fat, proteins, bleach reactions. Based on results and equations,
the rate-limiting step(s) can be identified. By experimentation – in a properly designed
scaled-down experimental setup (operating at the regime where the same step is rate-
limiting) the ‘rate’ constants can be determined. The experimental results and regime
insight can be used to modify the product design (composition and/or structure), and
the in-use process of the product. It may prevent making noneffective changes (and
cost) in products and processes that are not limiting.
It helps greatly to use schematic ‘drawing to scale’ where possible, so that the
designers appreciate the different length scales of the product and systems it inter-
acts with. Order of magnitude estimations for the different system mechanisms’ time
constants/characteristic times are sufficient. By determining the slowest or rate deter-
mining process(es), one can focus on these to improve the product and product appli-
cation process, and assume that mechanisms with the faster order of magnitude take
place ‘instantaneously’. This improves the understanding of the system and also sim-
plifies the mathematical modeling required. In [11] some further examples of product/
system splits and modeling are presented:
– Detergent requirements (p. 38)
– Flow properties of toothpaste (p. 64–66)
– Running an oil tanker (p. 77–80)
– Injection needle (p. 95–97)
– Designing a capsule for controlled release (p. 119–120)
– Design of Rockwool (p. 170–172)
A very useful source for transport phenomena data and models historically dispersed
over many literature sources was compiled and published by Janssen [12]. It provides
a good starting point for preliminary estimation of the rate key (transport) processes
in the application of products, depending on structure and size.
A comparable approach to quickly estimating material and component properties
for component substitution in existing products is also presented by Cussler and Mog-
gridge [13]. They present examples of component substitutions based on thermody-
namics (like solubilities) and kinetics (chemical conversion rates and mass and heat
transfer rates).
8.4.1.1 Estimation skills – Fermi problems

Estimation skills like those just mentioned are key for a quick appreciation of the im-
portant and less important phenomena to be considered in the design of the product
and its application process/use.
Another very nice way for designers to train and apply estimation skills (or ‘back-
of-the-envelope’ calculations) is to solve so-called ‘Fermi problems’, named after the
physician Enrico Fermi. Fermi estimates deliver order of magnitude estimates. Infor-
mation with this accuracy is in many cases often sufficient to make a decision – for
example to neglect a phenomenon or to study it in more detail.
8.4.1.2 ‘Systemic’ time constants – eigenvalues

Estimation of time constants of the separate rate phenomena can give much local in-
sight into a particular phenomenon. Unfortunately, having a full set of time constants
as such is hardly indicative of the behavior of the overall system (product/process) in
which various phenomena interact. Suppose there are N resources in a process (e.g.,
species mass, energy) which can interact with each other and each interaction repre-
sents a physical phenomenon with a rate effect (reaction, phase transfer, temperature
change, . . . ). Then 12 ⋅ N ⋅ (N − 1) time constants can be obtained related to the rate phe-
nomena. The alternative is to order the rate parameters (from which the time constants
were derived) in a (square) matrix. An example is the Jacobian matrix (an ordered set
of linear derivatives of an equation with respect to all variables for all equations) of a
set of conservation equations for the resources in a process (unit). Such a matrix has
N × N entries and each entry can be turned into an equivalent time constant. However,
mathematical analysis shows that the overall behavior of the process/product system
is determined by the eigenvalues of such a matrix. One obtains only N eigenvalues.
The joint effects of 12 ⋅ N ⋅ (N − 1) time constants of the system phenomena are con-
densed into a generalized N or “systemic” time constants. The benefit of computing
and analyzing such eigenvalues extends the knowledge that can be inferred from the
full set of rate related time constants. For instance, eigenvalues will reveal:
– Any linear dependency between resources, meaning that a resource varies as a lin-
ear combination of some other resources (occurrence of zero-value eigenvalue[s]);
– Stiffness: some resources respond very swiftly to changes while others are lagging
(very high ratio of the largest eigenvalue over the smallest nonzero one);
– Oscillatory behavior (occurring for an even number of complex-valued eigenval-
ues with nonzero imaginary parts)
– Instability of the process/product system
(some eigenvalues have positive real parts causing exponential growth)
These patterns of behavior are hard to fathom from just seeing a set of real-valued time
constants.
8.4.2 Data driven nonlinear product modeling: artificial neural networks
Due to the absence of fundamental relationships between the consumer perception

of the product’s sensory attributes (e.g., smoothness, softness, stickiness, etc.) and
product physical attributes (composition, ice crystal size distribution, air cell sizes,
etc.), the ‘product property function’ can be formulated by means of data mining tech-
niques.
In 2008 Bongers [14] presented a method for relating ice cream consumer percep-
tion to product physical attributes. This approach was based on large sets of exper-
imental data (and Artificial Neural Networks [ANN]). The product property function
involved a reduced set of physical product properties that could predict the consumer
perception of the ice cream product. Based on this work Unilever has been able to
introduce an innovative lower caloric ice cream with superior consumer perception
by reducing fat content and controlling ice crystal and air bubble sizes by improved
processing.
A similar approach was adopted by Almeida (2007) [33] in the rationalization of
the ingredients in the manufacture of a mayonnaise-like emulsified product. By means
of data mining techniques (partial least square and neural network techniques), only
relevant product attributes were classified in sensorial and analytical attributes. These
attributes were then used to create alternatives for the product’s microstructure and
ingredient composition.
The neural networks technique was also used by Dubbelboer et al. [15] in the for-
mulation of the property function for high-internal-phase emulsions.
Another class of chemical product design tools – for studying and modeling prod-
uct property functions – are DOE (Design of Experiments) software tools MODDE 11
(Umetrics) and Design-Expert® . Next to their capability to design experiments, these
tools also perform results analysis (model building and regression) and are increas-
ingly used to build product property models.
8.4.3 Scientific models for product state and behavior
8.4.3.1 Constituents
Before embarking on a description of the possible constituents of a product, it is help-
ful to introduce a common distinction between two different types of products. Follow-
ing this distinction, the class of products to be considered for design will be further
refined. Two product types can be discerned in relation to their extensive form:
– ‘Matter’, being continuously extensive (like any fluid)
– ‘Device’, a discrete entity with predetermined dimensions (like solid objects)
Fluid phase products are continuously extensive. Single molecular gases or liquids
qualify, but blended single phase fluids and formulations do as well, in addition to
multiphase systems like emulsions, foams, and gels (see Table 8.4). This is indepen-
dent of how the product is made (batch-wise or continuous manufacturing) or how
the product is packaged. With respect to solid phase products both forms can occur.
If the size of the particulates of the solid phase is very small and there are numerous
ones that can move more or less freely, the product becomes fluid-like again, such as
a crystalline powder. It is not an individual particulate that matters in the use of the
product. The particulates are used collectively as a kind of bulk material. Therefore,
the product is classified as extensive.
However, if the solid phase product has to be handled on the basis of individ-
ual pieces rather than as a bulk, and the solid phase product has sufficiently large
(macroscopic) dimensions, say, well above one millimeter, the product will classify as
discrete. A slab of steel and an electronic chip are examples of a discrete product, even
when the size is adjustable in the manufacturing process. Also, the multiple solid state
layers on a wafer will be considered as discrete components in a device-like product.
The characterization of a structured product in this paragraph is written primarily
with the continuously extensive product in mind. At the end of the paragraph some
supplementary remarks will be made concerning device-like, discrete products. Here,
the following assumptions are made, that the product:
– Is continuously extensive
– Is made up of several chemical and/or biological species
– Has one (pseudo)continuous thermodynamic phase
– Is dispersed when more than one (fluid) phase is present
– Has an interface between each set of two phases that acts as a distinguishable
entity having other properties than those in the bulk of the adjacent phases
– Has diffusion of species and thermal energy in and between the phases
– May have chemical reactions between species
The buildup of constituents in a multiphase structured product is visualized in Fig-

ure 8.7, where the relation between phase composition, the multiphase structure of
the mixture, and performance is shown.
8.4.3.2 Product structure

Concerning the geometric aspects of the bulk structure, it is observed that in many
products the bulk is made up of a dispersion of one or more thermodynamic phases
with a multilevel distribution; see Figure 8.8.
Examples of such dispersed systems are emulsions (liquid phase in another liq-
uid phase), foams (large volume fraction of gas bubbles in a dispersing liquid), bub-
bly liquids (dilute suspension of gas bubbles in liquid), and gels (semirigid disper-
sion of solids in a fluid). Dispersion may repeat itself at increasingly smaller scales.
Usually, one level of dispersion is considered in chemical equipment modeling, but
for structured products like food, cosmetics and pharmaceuticals, the use of multiple
Chemical Chemical Phases & Mixture’s Performance in

reactions species composition structure use
S1
S2 Phase (α)
S3
(continuous)
Interface Multiphase Performance

S4
αβ structure parameters
(dispersed)
S5
Phase (β)
S6
Fig. 8.7: Product composition, structure and performance for fluid phases.
Fig. 8.8: Multilevel/repetitive phase dispersions (oil/water/oil/water).

Tab. 8.3: A matrix representation of the phase ratios.
Continuous phase Distributed phase α Distributed phase β Distributed phase γ

(α) (α)
Phase α continuous εβ = 0.3 ε γ = 0.1
(β) (β)
Phase β ε α = 0.2 continuous ε γ = 0.3
(γ) (γ)
Phase γ εα = 0.1 εβ = 0.4 continuous
levels may be relevant. For repetitive dispersions, the spatial complexity of the system
will further increase. A discrete volume entity can act as a continuous phase in which
even smaller discrete volume entities are spatially distributed. E.g., a droplet of a dis-
persed phase (β) is contained in a continuous phase (α), which is in turn contained in
a droplet of the dispersed phase (β).
However, a droplet of the dispersed phase (β) may contain even smaller droplets
of the phase (α) and of another phase (γ). This dispersion of phases might repeat it-
self at successive levels of increasing detail. The extreme case of a multilevel spatial
distribution of dispersions (or even a continuous distribution) is not considered. The
repetitive occurrence of dispersed phases has implications for the calculation of the
overall phase ratios in the product. This is explained by means of a small example.
The first column in Table 8.3 indicates the continuous phase. The matrix row per-
taining to a continuous phase shows which mass fractions of the other (two) phases
are present (as dispersed phases) in that continuous phase. E.g., when a volume ele-
ment of phase β is the continuous phase it will by example contain a 0.2 mass fraction
of phase α and a 0.3 mass fraction of phase (γ). This information allows for computing
the overall phase ratios in a simple way. For example, consider a two-level dispersion
between three phases, where at the first level of dispersion, phase α is the continuous
phase, while the other two phases (β) and (γ) are dispersed. At the second level of
dispersion droplets of phase (β) contain dispersions of phases (α) and (γ), as well as
droplets of phase (γ) containing dispersions of phases (α) and (β).
Computation of relative amounts yields:
(α) (α) (α) (β) (α) (γ)
εoverall
α = [1 − ε β − ε γ ] + ε β ⋅ ε α + ε γ ⋅ ε α = 0.67 (8.1)
(α) (β) (β) (α) (γ)
εoverall
β = ε β ⋅ [1 − εα − εγ ] + εγ ⋅ εβ = 0.19 (8.2)
(α) (β) (α) (γ) (γ)
εoverall
γ = εβ ⋅ εγ + ε γ ⋅ [1 − εα − εβ ] = 0.14 (8.3)
Note that the overall phase ratios do sum up to unity.
8.4.3.3 Interfaces between dispersed phases

Each volume body of a phase will have bulk-like features, (e.g., in the bulk of the dis-
persed entities, like droplets) as well as interfacial features for the contact surfaces be-
tween phases. When describing interfaces between two thermodynamic phases, one
can adopt two viewpoints. In the “macroscale” view the surface is considered to be an
idealized two-dimensional, smooth object of negligible thickness. In the ‘microscale’
view it is seen as a curved three-dimensional body with a finite thickness (see [16]).
A special interface is the exterior surface of a product. The exterior shape of a fluid-
like product with low or moderate viscosity takes the form of its containment vessel.
There are usually no exterior surface functions of such products expected. However,
a fixated exterior surface will occur for solid phase products or fluid products with a
very high viscosity. Then the form of the surface and surface functions to be performed
are made part of the design.
8.4.4 Product structure matrix
One can think of a multitude of combinations of thermodynamics phases that are pos-
sible for creating multiphase products. Table 8.4 provides a schematic overview of the
various options.
A representative example of the complexity of a multiphase dispersed system is
ice cream. Ice cream is an emulsion, suspension and foam at the same time. The con-
tinuous phase is water with a high concentration of sugars. The liquid water contains
a population of small ice crystals. The ice crystals form the first dispersed phase. The
second dispersed phase is air, which is foam-like and dispersed in the form of small
bubbles in the continuous liquid phase. The third dispersed phase is fat, distributed
as small droplets in the water phase, so-called fat globules. The stability of this emul-
sion is controlled by means of emulsifiers. Actually, emulsifiers are being used to par-
tially destabilize the fat emulsion and so cause partial coalescence. These partially
Tab. 8.4: Product structure matrix for multiphase chemical products (adapted version, original from
Prof. A.A. Broekhuis, RU Groningen, Netherlands).
Dispersed phase ⇒ Gas Liquid Solid

⇓ Continuous phase
Gas Mixtures Aerosols / Mist Dust
Liquid Solutions Solutions Solutions
(organics, water, met- Bubbles Emulsions Colloids / sols
als, polymers) Foams Gels Suspensions
Pastes
Solid Adsorbents Adsorbents Blends, alloys
(organics, metals, Foams Blends Powders, granules composites
ceramics, polymers Compounds
Device-like (discrete):
Solid films, coating, multiple
stacked layers
coalesced fat droplets tend to collect at the waterside of the interface with the air bub-
bles and so help to stabilize the foam of air in water. The coalesced fat can form a
spatial 3D network.
A vast amount of scientific knowledge and detailed scientific models have been
developed for describing the behavior of (chemical) products and their components.
A very nice overview of the chemical product design related theories is presented
in Bröckel et al. [17], members of the European Federation of Chemical Engineers
(EFCE), section group “Product Design and Engineering.” The following product de-
sign related theories and basics are presented:
– Interaction forces between particles (by Schubert)
– From forces between atoms and molecules, to adhesion forces between parti-
cles in both gaseous and liquid environments
– Fundamentals of crystallization (by Ulrich, Jones)
– Emulsification techniques for the formulation of emulsions and suspensions (by
Schuchmann)
– Characterization of dispersed systems (by Polke, Schäfer)
– Modeling of chemical systems to predict product properties (by Gani, Abildskov)
Wesselingh [18] presents an elegant approach for using simple, but very illustrative
and useful models to quantify various product physical structures (in the nm to mm
range). These models can be used for a first order of magnitude quantification of prod-
uct structures in terms of volume fractions, surface area of dispersed phases, number
of surfactant molecules needed to form monolayers on the interface, etc. Adsorption
mechanisms and rheology are also explained and quantified.
Bröckel et al. present in [19] a number of illustrative examples of product design,
like:
– Biodiesel (by Meier)
– Fats, oils and waxes (by Reiz and Kleinebudde)
– Starch and starch-based products (by Leeb and Schuchmann)
– Gelatin (for food, pharmaceutical and technical applications – by Babel)
– Sugars (by Häusler)
– Synthetic precipitated silicas (by Schmoll)
– Heterogeneous catalysts (by Kunz et al.)
– Peroral products (by Mäder)
– Carotenoid products (by Leuenberger)
– Coffee-based beverages (by Schuchmann)
– Laundry powder components (by Boerefijn)
– Agrochemical products (by Frank)
In 2013 Bröckel et al. [20] extended the product design and engineering series even fur-
ther with a focus on the formulation of gels and pastes. Again, many examples from
food, personal care, and ionic liquid applications are discussed, together with the fun-
damentals of the rheology of dispersed systems, and how to apply these fundamentals
for the formulation of gels and pastes products.
8.5 Relations between product and process modeling
8.5.1 Causal flow of information on process – product modeling
The purpose of any manufacturing process is to convert the given feeds into prod-
ucts with target performances. For product and process design one can make models
of the feeds, the process and the products. The product state is an output from a be-
havioral model of the product. This state is also important in predicting the perfor-
mance metrics of a product by means of a performance model of a product. This can
be done on the basis of the physical-chemical state of the product as well as the pro-
cessing conditions that prevail under a product application. It is an issue in product
and process modeling how to consistently link the behavioral models of products and
processes. The behavioral models of the feeds, the process, and of the products must
match, i.e., be mutually consistent. For simple feeds and products, the situation is sim-
ple indeed. A simple product has a single phase with a homogeneous composition.
This fully matches with the common characterization of a stream in a process flow-
sheet model by its composition and flow rate. However, a complex product (and/or
feed) has a single-phase product of a distributive nature (e.g., a crystal size distri-
bution) or it has a multiphase composite structure, possibly also with one or more
phase with distributive properties. This was covered already in the preceding para-
graph (Section 8.4). This section will focus on connecting models of the product and
its manufacturing process.
For product design and improvement, it is desirable to have clear, causal quanti-
tative links between:
(a) the structure and composition of a product at the (supra)molecular scale as re-
sulting from the product manufacturing process,
(b) the macroscopic physical properties of a product, and
(c) the qualities of a product and associated performance for users.
Therefore, one needs three models in a cascade to generate this flow of information
from product ingredients in the feeds via the product manufacturing process to prod-
uct properties and the customer qualities in product use. This is the causal flow of
information (physical cause ⇒ effect), shown in Figure 8.9.
Each of these models will be highlighted. The positioning of the three models and
the flow of information between the models is in line with Bernardo (2016) [21].
Design decision variables Specification of product functions Additional inputs
molecules, ingredients
Product synthesis methods
product
phase structure • Enumeration approach
boundary
• Mathematical Programming
conditions
composition • Metaheuristic approaches
Specification of required feeds (z)
flowsheet topology
(Y: discrete structure) Process function model
physical
Flowsheet structure: S(Y; d, z) ≥ 0
functional unit sizes parameters (p)
Process states: M(x; u, d; z, p, w) = 0
(d: continuous structure) Product states: P(s, x) = 0 disturbances (w)
Performance: eprocess = E (x, u, d)
processing conditions
(u: temporal, adjustable)
process states (x)
product states (s)
process performance (e)
Product property functions
s = { structure (σ) | composition (c) }
Top-down, physical
Constitutive model: R( r ; σ, c ; π ) = 0
causal flow of parameters (π)
information
product states (s) product properties (r)
Product quality & performance functions

Q( q ; r ; ζ ) = 0 ; eproduct = E (q, s, ζ )
external conditions
of product use (ζ)
quality factors (q) product performance (eproduct )
Fig. 8.9: Causal flow of information through function models from manufacturing process to product
qualities and performance.
8.5.1.1 The product manufacturing ⇒ product structure and state model

The structure and the physical state of a product come into existence in a manufactur-
ing process, starting from product ingredients or precursors in its raw feed streams.
A model of a product manufacturing process has to be able to describe the structure
and states of the process as well as the physical state of its product(s). Hence, a mathe-
matical representation of matter in process streams and unit operations should allow
for the formation and transport of a structured product. Making a proper representa-
tion of product structure and state(s) is to be considered an integral part of the pro-
cess modeling. However, current process modeling practice is a bit reluctant in this
respect, dealing with simple (single phase) products only. Most commercial process
flowsheet simulation programs do not offer facilities to model and simulate the for-
mation of structured products, except as an add-on. Dedicated models of product for-
mation may be interfaced with a process simulator. Equation-oriented model building
software is well suited to include such models (e.g., gPROMS [22], JACOBIAN [23]). Ded-
icated product-process modeling tools are becoming available for classes of structured
products; e.g., gPROMS-FORMULATE, suitable for dealing with particle populations

with distributive properties [24].
8.5.1.2 A product structure and state ⇒ product properties model

This model derives physical and chemical product properties of the (bulk of the)
product (rheological properties, heat capacity, conductivity, pH,. . . ) from its underly-
ing physical (micro)structures and molecular compositions. Such structures involve
two or more thermodynamic phases, their distributive properties and the interfacing
between the phases. The chemical composition per thermodynamic phase must be
known in order to derive the macroscopic properties from suitable constitutive equa-
tions. The latter links molecular information to macroscopic transport properties.
During recent years, very nice progress has been made through the developments of
several product modeling case studies by the research groups of Prof. R. Gani, Prof.
K.M. Ng, and Prof. M. Eden. This has recently led to the first ‘chemical product simu-
lation and design tool’ (ProCAPD), that is offered and supported by the new company
‘PSE for SPEED’, started by Gani in 2017 [25]. Product properties can be simulated
and optimized based on the products’ molecular composition and structure. The
tool uses very extensive databases and has been demonstrated for design-analysis
of products as: single molecular products (solvents, refrigerants), mixtures/blends
(fuels, lubricants), and liquid formulated products (cosmetics, paints). TU Delft hosts
each year a workshop by Prof. R. Gani for PhD students and PDEng trainees (see
Section 14.3.4.1).
8.5.1.3 The product property ⇒ customer qualities (attributes) model

This model links customer related qualities and performance indicators with the phys-
ical properties of a product. Customer qualities are almost always empirically observ-
able quantities, including the conditions under which a product is used or applied.
Thus, the conditions in a product application process should be taken into account in
a customer qualities model. Thus, such qualities are rooted in both product properties
and product application conditions. Rigorous models based on physical first princi-
ples of product use are hard to derive or simply too expensive. Therefore, a customer
qualities model is mostly an empirical, data driven, black-box model, either a statisti-
cal one or an Artificial Neural Network (ANN) – already discussed in Section 8.4.2. The
model creates a data-supported bridge between the product properties, the conditions
of (external) use of a product, and the product qualities and performance metrics, as
appreciated by customers.
8.5.1.4 Inter-connectivity between product and process models

The causal flow of information through these three models assumes that information
is already available on the required ingredients for the product(s) in the process from
raw feeds to manufacturing. This has to be generated by a preceding product synthe-

sis effort. It is stressed that the design of a product and its manufacturing process runs
conceptually in an opposite direction. First one identifies the desired product qualities
and the associated performance metrics (SHEETS related). Then, one works towards
a suitable set of product properties and onto product composition and structure. This
allows for the identification of suitable product ingredients. Then one can think of
a conceptual design of a manufacturing process. The flow from product qualities to
product ingredients and a process design generates a tremendous increase in infor-
mation. This increase comes from a vast range of design decisions, particularly on the
process manufacturing side:
– Process topology involves the selection and ordering of processing functions
along with their streams connections in a process flowsheet. These are discrete
decisions, initially man-made, sometimes leading to superstructures over which
one can optimize a design.
– Type and sizes (volumes, exchange surfaces, length. . . ) of processing units have
to be derived from specified functional targets for the units (e.g., degree of con-
version, of separation, heat transfer duties, ratio of pressure increase, etc.).
– Operational conditions in the process units (pressure, temperature, preferred flow
regimes, phase ratios, etc.)
8.5.1.5 Mathematical model-based product-process optimization

Many product-process design cases are supported by repetitive process – product
model simulations in which human designers vary design decisions while aiming
for the target product qualities. Such iterative approaches often rely on serial design
decision making. Such iterative, serial approaches lead to feasible designs but not
necessarily optimum ones. Therefore, more comprehensive optimization approaches
are developed. These approaches rely on a computing-driven automated search over
the joint spaces of the design decision variables for the optimum performance. The
objective is to optimize some SHEETS performance metrics of process and product.
The search for a feasible optimum should remain within the validated domains of the
process models, particularly taking care to stay within the regions of applicability of
constitutive equations.
Model-based optimization of product-process designs experiences difficulties of
four generic kinds:
– The model equations do not comply with the mathematical requirements for the
existence of an extremum (e.g., convexity), undermining a theoretical guarantee
for an optimum.
– Even if a model would satisfy all conditions for the existence of an optimum there
may be multiple ones and the true extremum may be hard to find within the set of
local extrema.
– Exhaustive computational efforts during the search for the optimum due to the
high dimensionality of the search space as well as nonlinearity of the process and
product models
– In addition to using real-valued continuous variables (e.g., P, T, x) it can be nec-
essary to use discrete variables with binary {0, 1} or integer {0, 1, 2, 3, . . . } val-
ues. Such integers may enumerate a class of chemical groups; e.g., {1 = −CH2 −;
2 = CH3 −; 3 = −CH=HC−; 4 = HO−}.
To cope with the hardships in a comprehensive optimization of product and process

designs one can apply some form of problem decomposition. Two kind of decomposi-
tions are mentioned:
– First optimize the product structure, composition, and performance. One can ar-
range this outcome as a tight set of product related constraints to be obeyed in a
process optimization. However, the outcome is likely to be an expensive process,
making the perfect product at a negative profit, if the process is even feasible. One
may work toward a compromise by relaxing the product optimization and sacri-
ficing slightly product qualities that tend to make a manufacturing process more
expensive. By means of repeated simulations one may find a feasible solution for
both product and process designs. It is hard to say if the combined solution is the
optimum one.
– A sequential approach to the product design can ease the computational burden.
In computer-aided molecular design (CAMD), discrete variables occur to repre-
sent chemical groups that can be synthesized into myriads of different molecular
structures for the product. Based on a molecular structure, the continuous, real-
valued product properties can be determined. The search for such discrete set of
variables for molecular structure creation is too demanding in combination with
process synthesis optimization. An approximation would be to work with con-
tinuous product variables only, creating a hypothetical product with optimized
product properties and qualities. Then, the first cycle is to jointly optimize the de-
signs of process and (hypothetical) product. The next cycle of optimization is to
search in the discrete space of chemical groups for a combination that forms a
(real) product with properties and qualities close to those of the optimized hypo-
thetical one. Having found the closest real product with near optimal properties
and qualities, one can re-optimize the process design for this real product. This
two-step approach is developed and applied to the design of a solvent based pro-
cess in [26], and it is also an approach that is applied in case studies by Gani [25].
First, the designs of the process and of a hypothetical solvent (for internal use
in the process) are jointly optimized. Such optimization is possible by using only
continuous variables in the process and the product models. Having once iden-
tified the ideal hypothetical solvent, a search is made for a small set of real sol-
vents closely matching the identified ideal solvent properties. From this small set
of near-optimum real solvents, the best performing candidate is selected by re-
optimizing the process for each solvent. This design approach accounts for the
interactions between process and solvent designs in an affordable optimization
procedure.
8.5.2 Notes on first-principles models of formation of a structured product
Structured products are made up of multiple thermodynamic phases. The continuous

phase of the product may contain one or more dispersed phases of gas bubbles, liq-
uid droplets, or solid particles such as crystals. The particulates of a dispersed phase
are often distributed in size (nonuniformly) and may have internally different physical
and mechanical states and chemical composition. Furthermore, a product will have
a geometric extension in up to three dimensions. The distribution of the phases and
their properties can vary with the spatial position. The structure of a product involves
the distributions of phase properties as well as the spatial distributions. Last but not
least, the structure of a product may change over time due to internal physical and
chemical events in the phases and between the phases, such as chemical reactions,
heat and mass transfer, and flow of phases. A visual abstraction of a structured prod-
uct and an outline of a generic mathematical model of a structured product are given
in Figure 8.10.
Phase domain (α) ; state s(α) Product

Φ(α,β)
Properties
External
r
influences Phase domain (β) ; state s(β)
u(t) Boundary fluxes
Φ(β,γ) between phases
Phase domain (γ) ; state s(γ)
Behavioral model per phase for composition & structure:
∂ s(*) /∂ t = M (s(*), ∇z* s(*), Φ(*, # ) , u(t); p(*) )
Fluxes of mass & energy across interfaces: [*, # = { α, β, γ } ]
B(Φ(*, # ), ∇z* s(*), ∇z# s(#) ) = 0
Product structure – properties relationships:
R( s(α), s(β) , s(γ) , r ) = 0
State of a phase: s(*) = { temp., press., compos., size distribution, mass hold-up}
Fig. 8.10: Summary of a generic model of a multiphase, structured product.

In a structured product, each phase has its own internal physical state (s) that evolves
in time due to internal events in the phase domain as well as to interactions with the
neighboring phases by fluxes (φ) and with the external world (u). Heat, mass, and mo-
mentum fluxes between adjacent phases are governed by the states of those phases.
The properties of a multiphase product (r) are defined by the integral effect of the
states (s) of the phases domains {α, β, . . . }. Each of the phases in a structured product
has to be premanufactured in an upstream part of the process as a ‘simple’ product,
involving a single phase characterized by its physical state (temperature, pressure,
and composition) and physical properties (density, heat capacity, viscosity, thermal
conductivity, and surface tension). The various simple products will go as ingredients
into a product formation unit, such as an extruder or a colloid mill, to form the desired
structured product.
Process modeling and simulation support the design and analysis of manufactur-
ing and product application processes. For simple products, there is a coherent body of
modeling knowledge for a broad class of manufacturing processes. This knowledge is
captured in the common commercial flowsheet simulators for processes dealing with
locally uniform fluid phases (gases, liquids). Such generic modeling is far less com-
mon for processes which handle structured products. This is due to the complexity
and diversity of structured products and the (labor) cost of modeling. Products leav-
ing a manufacturing process and entering a product application process will also set
the level of complexity for the associated process modeling. This complexity can be
effectively dealt with by means of first principles, equation-oriented modeling, com-
bined with the numerical power of modern equation solvers. Such modeling platforms
are under development and already offered for formulated products and their manu-
facturing processes [24, 25].
To end with a challenging example, let’s have a look at the modeling of freezing ice
cream. Ice cream has a microstructure consisting of ice crystals, air bubbles, liquid fat,
and a sugary solution (see Figure 8.4). Starting with the liquid feed and air, ice cream is
formed in a freezer. The freezer is an externally cooled extruder with a scraped surface
heat exchanger (Figure 8.11). Such units and models are topics of study in the chemical
engineering literature; e.g., [5, 27, 28].
The water in the sugary solution freezes at the cooled surface of the barrel, result-
ing in the growth of a layer of ice on this surface. This layer is periodically scraped off
by rotating blades and dispersed as a population of ice crystals of different forms and
sizes in the liquid – air mixture. The mixture of liquid/air/ice is axially transported by
the rotor. Fundamental modeling of the ice cream formation is of a staggering com-
plexity. It is staggering because in principle one has to deal with:
– Eight phases: ice crystals, ice layer, air bubbles, fat droplets, aqueous sugar solu-
tion, rotor, wall, coolant
– Three spatial coordinates (even a spiraling one) and time
– Distributions of particulate size (or mass) for the ice crystals, air bubbles, fat
droplets
– Kinetics of growth, agglomeration, breakage of crystals, bubbles, and droplets

– Mass, energy, and momentum balances for the phases in the ice cream mixture
– Energy balances for rotor, wall, and coolant as well as a momentum balance for
the rotor
– Physical transport phenomena (mass, energy, momentum) in and transfers be-
tween phases
– Thermodynamic equilibrium conditions at interfaces between phases
– Physical properties for thermodynamics and transport (with non-Newtonian flow)
In view of this overwhelming complexity, current ice cream freezing modeling efforts
are directed at pragmatically developing very much simplified, yet realistic models of
this process unit for design and control [27, 28].
Another parallel modeling effort is required for the application process of the
structured product. This is even more challenging. For instance, the application pro-
cess of ice cream is human consumption and digestion. Currently, mechanistic models
of the mouth-feel, melting behavior, flavor development and sensing, digestion of ice
cream in intestines, and of the impact of ice cream ingredients on body condition and
long term health are hard to find. Yet, new developments in food engineering are on
the way by studying food digestion in the human gastrointestinal track as a multiscale
process, in analogy to multiscale chemical processes [29].
In summary to the relation process-product modeling and computing one can
notice distinct developments towards models of (micro)structured products based on
first principles as well as data driven. First principles product models have to account
for much structural information (spatially and distributive properties) leading to
complicated sets of (partial) different equations on multiple domains with changing
boundaries. This is in contrast to much of current process modeling and computing
where large sets of nonlinear algebraic equations prevail. Combining and integrat-
Ice cream freezing in an extruder with external cooling

and scraped surface heat exchange
θ Barrel
H
Screw
wc
Db
wf
ef
Growth and scraping Sp Coolant flow
of a thin ice layer
Fig. 8.11: Ice cream freezing in an extruder [5, 28].

ing product and process modeling is needed for further advancements in chemical
engineering applications.
8.6 Product models: overview
For product design in the (bio)chemical engineering field, the different types of mod-
els that are created and used in designing and discussed in this chapter are listed in
Tables 8.5 and 8.6. The focus in these tables is on the first levels of the Delft Design Map
to show the large number of types of models created and used in a design process.
Tab. 8.5: Schematic overview of various model types created and used during the DDM’s project
framing and product concept-consumer wants (quality function) design levels (see also Sec-
tion 4.3.2).
Design Models (physical, schematic, verbal, mathematical)

levels (examples only, not complete list)
PF: – Business context graphs

Project – Development/prediction in time: market volume, sales volume, price level
Framing product, price level raw materials, energy, population growth, etc.
– Supply chain block diagram
– Value chain diagram
– Product’s life cycle diagram
– Schematic showing design driver and design type
– BOSCARD table
– Project organization chart (members and accountabilities)
– Project team Belbin roles diagram
– Project and time planning chart
– ...
CW-PC: – Product application/use process block diagram
Consumer – Product life cycle block diagram
Wants – – Brainwriting collage or table with identified stakeholders from all life cycle phases
Product – Storyboard showing various consumer interactions with product
Concept: – Storyboard showing other stakeholders interactions with product
(quality – House of Quality (1) diagram
function) – Consumer wants to product (quantitative) attributes
– Product attributes quantitative specifications table
– ...
8.6 Product models: overview | 189
Tab. 8.6: Schematic overview of various model types: created and used during the DDM’s product
concept (product) property function design levels (see also Section 4.3.2).
Design Models (physical, schematic, verbal, mathematical)

Levels (examples only, not complete list)
PC-PF: – Schematic drawing representing alternative product structures

Product – Schematic drawings of the phases (solid/liquid/gas), their size (distribution),
Concept: volume fraction, spatial distribution in the product, composition of each phase
(product) – See Figure 8.4 for ice cream product
Property – Pure component properties table (See Table A13.3)
Function – Mathematical models of alternative products and states
– Composition, structure, and physical properties (crystal size and distribution,
droplet size and distribution, volume, specific surface area, particle density,
powder bed density, viscosity); see Section 8.3.2.
– Steady state but may be extended to the change of product state over its life cycle
(manufacturing, storage, transport, application/use, end-of-use), or dynamic
models
– Block diagram showing product application/use processes
– Application/use of: transportation fuels, lubricants, paint, laundry detergents,
medication administering, ice cream consumption
– Schematics showing key mechanism of the application/use processes: systems
splits (Section 8.3.2)
– Mass transfer: mixing, dispersion, dissolution, heat transfer, (bio)chemical
reaction
– Time constant equations of the product application/use process mechanisms
– Artificial Neural Networks models
– Mathematical models of product’s (components’) behavior during product’s
application/use
– Also in storage, transport, collection, recycling
– (Bio)chemical reactions (stoichiometry, byproducts, waste products production)
– Physical changes: phase changes, dispersion, dissolution, heat and mass
transfer (energy use in product application/use)
– Simulating effects of different product alternatives
– House of Quality (2) diagram
– Product (quantitative) attributes translated into product properties
(composition/structure, application/use conditions)
– Brainwriting diagram of alternative (bio)chemical components
– Molecular formula / structure
– Reaction equations
– Stoichiometry, byproducts, waste products produced in application/use
– Supply chain block diagram update
– Preliminary ‘SHEETS’ performance models (verbal, mathematical)
– SHEETS: Safety, Health, Environment, Economy, Technology, Social
– Safety, health, environmental: effects of raw material and products transport and
logistics, during waste production during product application/use
– Economy: raw material costs, energy use in product application/use
8.7 Modeling for ‘Safety, Health, Environment, Economy,

Technology, Social (SHEETS)’
For schematic and mathematical modeling of products in relation to the ‘SHEETS’

evaluation criteria the reader is referred to:
– Safety Chapter 11
– Health Chapter 11
– Environment Chapter 12
– Economy Chapter 10
– Technology this chapter and Chapter 9 (for process modeling)
– Social Chapter 12
List of symbols and abbreviations
Symbols
A [m2 ] surface area (e.g., for heat transfer or mass transfer)
a [m2 /m3 ] specific surface area
c [*] generic composition variable in product property model
C [kmol/m3 ] reactant concentration
C∗g [kmol/m3 ] reactant concentration in gas corresponding to C∗l
C∗l [kmol/m3 ] saturation reactant concentration in the liquid
corresponding to C∗g
cp [J/(kg ⋅ K)] specific heat capacity
d [*] generic size (length, area, volume) parameter of a
process unit
D [m2 /s] diffusion coefficient
eprocess [*] generic process performance indicator in process model
eproduct [*] generic product performance indicator in product quality
model
F [m3 /s] volume flow rate
k [kmol/(m3 s)] 0-order rate constant
k [1/s] 1st order rate constant
kl [m/s] liquid side mass transfer coefficient
L [m] travel length, diffusion length
m [–] equilibrium concentration ratio (gas/liquid)
p [*] generic parameter in a process/product model
q [*] generic quality variable in a product quality model
r [*] generic product property variable
rc [kmol/(m3 s)] conversion rate
s [*] generic state variable of a product model
List of symbols and abbreviations | 191
tc [s] conversion time

td [s] diffusion time
tht [s] heat transfer time
tmt [s] mass transfer time
tshear [s] shear stress relaxation time
u [*] generic input variable to a process/product model
U [W/(m2 K)] overall heat transfer coefficient
V [m3 ] volume
v [m/s] travel velocity
w [*] generic disturbance variable of a process/product model
x [*] generic state variable in process model
Y [–] binary variable
z [*] variable of a feed stream to a process model
αi [W/(m2 K)] internal heat transfer coefficient

αe [W/(m2 K)] external heat transfer coefficient
(α)
εβ [–] phase mass ratio (of phase β dispersed in continuous
phase α)
(α)
εγ [–] phase mass ratio (phase γ dispersed in continuous phase α)
(β)
εα [–] phase mass ratio (phase α dispersed in continuous phase β)
(β)
εγ [–] phase mass ratio (phase γ dispersed in continuous phase β)
(γ)
εα [–] phase mass ratio (phase α dispersed in continuous phase γ)
(γ)
εβ [–] phase mass ratio (phase β dispersed in continuous phase γ)
εoverall
α [–] overall phase mass ratio of phase α in total system
εoverall
β [–] overall phase mass ratio of phase β in total system
εoverall
γ [–] overall phase mass ratio of phase γ in total system
ζ [*] generic external condition of use in product quality model
η [kg/(ms)] viscosity
λ [W/(mK)] heat conductivity
π [*] generic physical parameter for a product model
ρ [kg/m3 ] specific density
σ [*] generic structure variable in a product property model
Φ [*/(m2 s)] generic flux of a physical quantity
τ [s] travel time, residence time
Abbreviations
3D 3-Dimensional
ANN Artificial Neural Network
BOSCARD Background, Objectives, Scope, Constraints, Assumptions, Reporting,
Deliverables
CAMD Computer-Aided Molecular Design
CAD Computer Aided Design

CW-PC Consumer Wants – Product Concept
DOE Design Of Experiments
EFCE European Federation of Chemical Engineers
HAZOP HAZard & OPerability study
HoQ House of Quality
P&ID Piping & Instrumentation Diagram
PBS Process Block Scheme
PC-PF Product Concept – Property Function
PF Project Framing
PDEng Professional Doctorate in Engineering
PFD Process Flow Diagram
PFS Process Flow Scheme
QFD Quality Function Deployment
SHEETS Safety, Health, Environment, Economy, Technology, Social
TAME Tert-Amyl Methyl Ether
[1] Van Overveld CWAM. PowerPoint Slides guest lecture in PDEng course Advanced Principles of
Product and Process Design (ST6064) at TU Delft, 2010.
[2] Skyttner L. General systems theory – problems, perspectives, practice 2nd edn, World Scien-
tific Publishing Co. Pte. Ltd., Singapore, 2005, 97–99.
[3] De Haan AB, Swinkels PLJ, de Koning PJ. Instruction manual conceptual design project
CH3843 – from idea to design, Appendix-18, Department of Chemical Engineering, Delft Uni-
versity of Technology, Delft, Netherlands, 2016.
CH3843 – from idea to design, Appendix-17, Department of Chemical Engineering, Delft Uni-
versity of Technology, Delft, Netherlands, 2016.
synthesis in foods manufacture. In: Martin M, Eden MR, Chemmangattuvalappil NG, eds. Com-
puter Aided Chemical Engineering Vol.39 Tools for chemical product design – from consumer
products to biomedicine, Amsterdam, Netherlands, Elsevier BV, 2017, 417–441.
[6] QFD Online Quality Function Deployment tools and information for real time application, free
QFD Templates 2007–2010. Accessed October 10, 2017, at www.qfdonline.com/templates/.
[7] Seider WD, Seader JD, Lewin DR, Widagdo S. Process and Process Design Principles: Synthesis,
Analysis and Evaluation, 3rd edn, Hoboken, NJ, USA, John Wiley & Sons Inc., 2010, 45–47.
[8] Wesselingh JA, Kill S, Vigild ME. Design & development of biological, chemical, food and phar-
maceutical products, Hoboken NJ, USA, John Wiley & Sons Inc., 2007, 27–38.
[9] Oosterhuis NMG. PhD Thesis, Delft University of Technology, 1984. Accessed August 19,
2017, at https://repository.tudelft.nl/islandora/object/uuid:03a887b7-8c20-4052-8d6b-
7fe76918d7ec/datastream/OBJ.
[10] Sweere APJ, Luyben KChAM, Kossen NWF. Regime analysis and scale-down: tools to investigate
the performance of bioreactors. Enzyme Microb. Technol., 9, July, 1987, 386–398.
maceutical products, Hoboken NJ, USA, John Wiley & Sons Inc., 2007, 38, 64–66, 77–80, 95–
97, 119–120, 170–172.
[12] Janssen LPBM, Warmoeskerken MMCG. Transport phenomena data companion, VSSD, Delftse
Uitgevers Maatschappij, Delft, 1987.
[13] Cussler EL, Moggridge GD. Chemical Product Design, New York, USA, Cambridge University
Press, 2001, 75–85.
[14] Bongers PMM. Model of the product properties for process synthesis. In: Braunschweig B, Jou-
lia X (eds.), Computer Aided Chemical Engineering 25, 18th European Symposium on Computer
Aided Process Engineering – ESCAPE 18. Amsterdam, Netherlands Elsevier BV, 2008, 55–60.
[15] Dubbelboer A, Zondervan E, Meuldijk J, Hoogland H, Bongers PMM. A neural network appli-
cation in the design of emulsion-based products. In: Lockhart Bogle ID, Fairweather M (eds.),
Proceedings of the 22nd European Symposium on Computer Aided Process Engineering, 17–20
June 2012, London. Amsterdam, Netherlands, Elsevier BV, 2012, 692–696.
[16] Edwards DA, Brenner H, Wasan DT. Interfacial transport phenomena and rheology, Boston MA,
USA, Butterworth-Heinemann, 1991.
[17] Bröckel U, Meier W, Wagner G, eds. Product design and engineering – best practices, volume 1:
basics and technologies, Weinheim, Germany, Wiley VCH Verlag, 2007.
maceutical products, Hoboken NJ, USA, John Wiley & Sons Inc., 2007, 259–293.
[19] Bröckel U, Meier W, Wagner G (eds.). Product design and engineering – best practices, vol-
ume 2: raw materials, additives and applications, Weinheim, Germany, Wiley VCH Verlag, 2007.
[20] Bröckel U, Meier W, Wagner G (eds.). Product design and engineering: formulation of gels and
pastes, Weinheim, Germany, Wiley VCH Verlag, 2013.
[21] Bernardo FP. Integrated process and product design optimization, Ch. 12. In: Martin M, Eden
MR, Chemmangattuvalappil NG, eds. Computer Aided Chemical Engineering Vol.39 Tools for
chemical product design – from consumer products to biomedicine. Amsterdam, Netherlands,
Elsevier BV, 2016, 347–371.
[22] Process Systems Enterprise (UK). gPROMS: advanced process modelling platform. Accessed
October 5, 2017, at https://www.psenterprise.com/products/gproms.
[23] Process Systems Engineering Laboratory, Massachusetts Institute of Technology (USA),
Process modelling & computing software JACOBIAN. Accessed October 5, 2017, at https:
//yoric.mit.edu/Jacobian.
[24] Process Systems Enterprise (UK). gPROMS|FORMULATE: gPROMS Formulated Products, plat-
form for integrated digital design of robust formulated products and their manufacturing
processes. Accessed October 5, 2017, at https://www.psenterprise.com/products/gproms/
formulatedproducts.
[25] PSEforSPEED Company Ltd., Computer-aided application of process systems engineering (PSE)
for sustainable product-process engineering, evaluation & design (SPEED). Accessed Octo-
ber 5, 2017, at http://www.pseforspeed.com.
[26] Bardow A, Steur K, Gross J. Continuous-Molecular Targeting for Integrated Solvent and Process
Design, Industrial & Engineering Chemistry Research, 49, 2010, 2834–2840.
[27] Arellano M, Benkhelifa H, Alvarez G, Flick D. Coupling population balance and residence time
distribution for the ice crystallization modeling in a scraped surface heat exchanger, Chemical
Engineering Science, 102, 2013, 502–513.
[28] Dorneanu B. Model Reduction in Chemical Engineering. Case Studies Applied to Process Anal-
ysis, Design and Operation. PhD Thesis, Delft University of Technology, 2011. Accessed Octo-
ber 5, 2017, at https://repository.tudelft.nl/islandora/object/uuid:afecbacd-0ac1-4a01-b7ab-

db1ba0c42f50?collection=research.
[29] Bornhorst GM, Gouseti O, Wickham MSJ, Bakalis S. Engineering Digestion: Multiscale Pro-
cesses of Food Digestion, Journal of Food Science, 81, 2016, R534–543.
[30] Ng KM, Gani R, Dam-Johansen K (eds.). Chemical product design: towards a perspective
through case studies, Computer Aided Chemical Engineering – Volume 23, Amsterdam, Nether-
lands, Elsevier BV, 2007.
[31] Khalloufi S, Kharaghani A, Almeida-Rivera C, Nijsse J, Dalen G, Tsotsas E. Monitoring of initial
porosity and new pores formation during drying: A scientific debate and a technical challenge,
Trends in Food Science & Technology, 45, 2015, 179–186.
[32] Khalloufi S, Almeida-Rivera C, Janssen J, Van der Vaart M, Bongers P. Mathematical model for
simulating the springback effect of gel matrixes during drying processes and its experimental
validation, Drying Technology 29, 2011, 1972–1980.
[33] Almeida-Rivera C, Jain P, Bruin S, Bongers P. Integrated product and process design approach
for rationalization of food products. In: Pleşu V, Şerban Agachi P (eds.), Computer Aided Chem-
ical Engineering – Volume 24, Amsterdam, Netherlands, Elsevier BV, 2007, 449–454.
[34] Linke P, Papadopoulos, AI, Seferlis P. Systematic Methods for Working Fluid Selection and
the Design, Integration and Control of Organic Rankine Cycles – A Review. ENERGIES, 8, 2015,
4755–4801.
[35] Papadopoulos AI, Stijepovic M, Linke P, Seferlis P, Voutetakis S. Toward Optimum Working
Fluid Mixtures for Organic Rankine Cycles using Molecular Design and Sensitivity Analysis.
Industrial & engineering chemistry research, 52(12), 2013, 116–133.
[36] Schilling J, Tillmanns D, Lampe M, Hopp M, Gross J, Bardow A. From molecules to dollars: in-
tegrating molecular design into thermo-economic process design using consistent thermody-
namic modeling, Molecular systems design & engineering, 2, 2017, 301–320.
|
Part C: Design optimization
9 Process modeling and optimization
Synopsis
This chapter deals with modeling in support of three successive innovation stages of
product. Linear process modeling is associated with applications in the concept stage.
Nonlinear process modeling is linked with the feasibility and development stages.
Process simulations will find applications in the development stage, while the fea-
sibility stage benefits from optimization applications, which scout the entire design
space and optimize over it. The modeling approach chosen here has a practical in-
clination. Many academic textbooks on modeling cover the common modeling steps,
such as development of model equations, mathematical model analysis and numer-
ical means for computing solutions. Therefore, these steps are largely skipped here.
Instead the contextual and application sides of modeling in process engineering will
get more attention. For instance, how does one get to a relevant model for a process
engineering problem? Furthermore, how does one analyze and evaluate the outcomes
from computational model applications and how are clues derived for process innova-
tions? Aspects of sensitivity analysis, uncertainties and identification of technological
constraints from model based solutions will also be discussed.
9.1 Justification and objectives of process modeling
Modeling objectives for process synthesis, analysis and optimization

The purpose of process modeling and computing is to offer effective, quantitative
means to synthesize, analyze and evaluate process designs. The main goal of this
modeling and computing chapter is to bring out the conceptual aspects of modeling,
simulation and optimization for process design applications. In view of limited text
space this emphasis will go at the expense of the underlying mathematical and nu-
merical aspects of model building and equation solving. Here we will simply assume
that:
– Models will be properly formulated from mathematical and computational per-
spectives.
– Their domain of applicability (containing feasible physical solutions) is well un-
derstood.
– An applied numerical solution procedure is consistent with the model and suffi-
ciently accurate.
The focus will be on the context and practice of developing model applications for
the concept, feasibility and development stages in support of product and process de-
signs; see also Chapter 4. Attention will also be given on how to extract relevant infor-
https://doi.org/10.1515/9783110467741-009
198 | 9 Process modeling and optimization
mation from model applications for purpose of improvement and innovation. There is
an established industrial practice of supporting process design case studies with com-
mercial steady state process flow sheet simulators. This vast area of computer aided
process simulations has been thoroughly covered in [1], and therefore will be skipped
here. Dynamic modeling and model reductions for process control and operations are
also omitted.
Further background reading on process modeling for simulation and optimization
in process development, design and operations is offered in a number of high quality
text books; regarding modeling [2–8]; flow sheeting simulations for design [1, 9–12]
and optimization [13–15] and [16].
Conceptual representation of a process: network features

A process is an open system that internally consists of a network interconnected and
interacting unit operations. The system is in interplay with its physical, economic
and social surroundings. The system performs specified production functions; i.e.,
turning physical feeds into products with a chemical functionality. The production
is guided by process and product control information. The system may operate in dif-
ferent modes over time. Each mode is characterized by a distinct internal state of the
process.
A process is connected to its surroundings by:
– Feed, product and waste streams (the latter to waste treatment facilities).
– Control set points for its operational conditions and modes as supervised by plant
operators.
– The environmental conditions.
– The cash flow it generates for the process owner.
The behavior and performance of a process is codetermined by the interactions be-

tween a process and its surroundings. This includes the imposed modes of operation.
Each mode can be seen as the realization of an external scenario, under which the pro-
cess must make product(s) while performing well enough. Development of scenarios
is part of the design; these scenarios are used for testing candidate processes under
design. Consequently, the set of process models must be able to handle all scenarios.
This explains why next to steady state models also dynamic models are required.
Sharing a generic view on process modeling and computing

Design models carry a lot of process information on behavior and performance. The
common steps in the modeling work procedure are outlined in Table 9.1. These steps
are presented in a sequential fashion. This is not how it works in practice. There is a lot
of going forward and backward over the steps during the modeling activity. One may
also visualize these activities as placed on the perimeter of a circle with easy criss-
crossing between activities.
9.1 Justification and objectives of process modeling | 199
Tab. 9.1: Common steps in a modeling work procedure.
Step Activity
1 Select the object of modeling, and justify a modeling effort from anticipated benefits.
2 Agree on goal(s) of model and intended application domain(s) for future model use.
3 Find a suitable conceptual representation of the object to be modeled.
4 Develop a mathematical model of the object for its structure/behavior/performance.
5 Code the mathematical model in software with verification of correctness of coding.
6 Select a numerical solver with enough numerical precision for intended applications.
7 Validate the coded model by comparing computed answers with experimental data.
8 Select case studies based on interaction scenarios between object and its surroundings.
9 Apply model to case studies and analyze computed results on achieved accuracies.
10 Evaluate computed results on significance for goals and new insights.
11 Report model, case studies, numerical results and the evaluation of results.
The next sections will deal with the nature of model applications in the three design
stages: concept, feasibility and development. The successive stages of process devel-
opment and design call for models of different functionalities and capabilities. Three
classes of models are considered:
– Linear models for process input-output blocks: for use in the concept stage.
– Nonlinear models for process simulation: for use in the development stage.
– Nonlinear models for process optimization: for use in the feasibility stage.
For each stage the main modeling aspects generic to that stage will be discussed, in-
volving:
– Consistent model formulation with proper degree of freedom analysis.
– Significance of solutions and sensitivity analysis.
– Analysis of active inequality constraints in case of a process optimization.
– Identification of technological limitations in optimized solutions.
Having dealt with the generic aspects, a number of (small scale) cases are presented as
illustrations of certain modeling aspects. These examples are small scale on purpose
in order to illustrate a way of thinking and acting in modeling. There is a good reason
to keep them small. The many facilities of process modeling software with associated
computing power often tend to overwhelm novice modelers. Therefore, it is preferable
to develop basic model synthesis and analysis skills first by means of small scale prob-
lems. These problems can even be solved by hand; they serve as vehicles to transfer
ways of thinking in model synthesis, analysis and evaluation. Each case is concisely
presented in this chapter by means of motivation and objective, modeling approach
and main outcomes. Details of the cases and results are collected in the Appendix to
this Chapter A9.
Before embarking on modeling for the three design stages it is expedient to make
a remark about a restrictive meaning of some terms in this chapter. The terms behavior
and performance will be used in a more focused way than commonly done in chemical
process engineering. The term behavior of a process or product will relate to patterns
of change in its (bio) chemical and physical quantities. The term performance is in-
tentionally reserved for factors that express some kind of value and appreciation by
human society, including customers of a product, operators and owners of a process
and/or legal authority. Such performance factors are related to safety, health, econ-
omy and ecology.
9.2 Contributions to a concept stage with linear modeling
9.2.1 Modeling for a concept stage
Let us move from general modeling aspect to the specifics of modeling for process
development and design. In this chapter a distinction is made between three generic
main items of any process model:
– A sub model describing the physical/chemical/biological behavior of a process.
This sub model deals with changes in physical and chemical resources (species
mass, energy, momentum, electrical charge, etc.) as they go through a process.
– Another sub model quantifying some indicators for process performance. These
indicators reflect the degree of appreciation by process owners of what the process
(design) achieves in terms of economy and ecology under safe operation.
– Scenarios defining pertinent influences of the surroundings of a process on the
process itself. One must specify external scenarios under which a process is sup-
posed to perform its function(s) in order to compute and analyze process behavior
and performance.
A process interacts with its environment and this feature must be represented in a
process model. The common, causal flow of information to and from a process model
is shown in Figure 9.1.
It represents the flow of physical quantities from process inputs to its outputs. The
inputs cover the physical feed streams, the selected operating conditions (u), physical
parameters (p) for the process units, as well as other known external influences (w)
acting on the process. The fourth class of inputs involves the geometric design vari-
ables (d) of the process units (e.g., lengths, areas, volumes). Their numerical values
will have been established in the preceding design phase. The outputs (x) are the col-
lective set of physical states of units and streams in the process. In this Figure, it is
assumed that the types of unit operation and the ways in which such units can be
connected to an internal network structure have been decided upon by preceding con-
ceptual design decisions. Figure 9.1 mentions the three main sub models: behavioral,
External influences (w) (p) Physical parameters

(beyond control) (known)
Process model
System states
Fixed inputs
Flowsheet structure:
(x)
(z) fixed network in process
Behavioral (physical) part:
Adjustable inputs M (x; u, d; z, p, w) = 0

Performance metrics
Performance part:
(u)*
E = E (x, u, d, z) (E)
(d)* Continuous design variables

(e.g., equipment area, volume)
Model based analysis Given a scenario: ==> Compute:

of a process design: {z, p, w} and {u*, d*} {x, E} and sensitivities
{u*, d*} are given design decision variables
Fig. 9.1: Information flow through a process model for analysis of a process design.
performance and the external scenario(s). Here, the structure part is assumed fixed.
This assumption needs to be relaxed when modeling for process synthesis studies that
specifically deal with alternative process structures. Consequently, the model setup in
Figure 9.1 is suitable for the analysis of a design only. The inputs must all be given by
the designer/modeler; there is no automatic adjustment of the inputs as to target for
process states or to optimize process performance metrics.
Design leads to a reversal of information flow to a model

In actual design situations the geometric design variables will be unknowns rather
than given quantities. These unknowns add to the degrees of freedom in a design sit-
uation. So, to take away the additional degrees of freedom, the model needs to be
augmented by as many design specifications as there are unknown geometric design
variables. These new design specifications are often given as specified targets for some
key physical outputs of the process, e.g., nominal production rate and product quali-
ties as well as constraints on waste streams (sizes and waste content). There can also
be specifications of duties of process units, such as degree of conversion and selectiv-
ity in a reactor and a desired degree of separation or purities in a separation column.
The general rule is that targets on physical outputs are set, while having the required
geometry of the equipment as an outcome to meet such targets. The flow of design
information from targets to geometric features can be reverse to the causal (physical)
flow of information.
Modeling for the concept stage

A concept stage is focused on creating alternative, competing product concepts. In the
background, though, one must also keep in mind the basic processing principles for
eventually making such a product from the available feeds. It is prudent to think al-
ready about a process block diagram which outlines the key processing steps to make
a product. The viability of a product concept will also be determined by its manufac-
turing complexity and costs. At this concept stage, process models and computations
should be kept simple, addressing the main features of processing functions only. For
this reason, it suffices to use linear models of the processing steps in a process block
diagram for a first exploration. A model is linear when each equation in the model is
linear in all variables. There are five common forms of deterministic models, summa-
rized in Appendix A9. The application cases in this section deal with linear algebraic
and differential equations with real variables. The exception is the modeling of the
structure of a block diagram/flow sheet involving a linear algebraic model with binary
variables. Six cases will be given to illustrate some aspects of modeling.
9.2.2 Cases for concept stage modeling
Case 9.1: Exploring the stoichiometric space for an innovative chemical conversion
Motivation and objective

Chemical processes have a conversion section in which the main chemical species in the feed
streams are converted into target species for the product(s). A conversion is characterized at the
chemical level by (a) the reaction stoichiometry; (b) associated thermodynamic reaction equilib-
rium; (c) reaction kinetics; and (d) level of catalytic activity and catalyst stability over time. The
considerations of the associated physical transport phenomena, phase contacting patterns and
hydrodynamic flow regimes are part of the subsequent physical reactor selection and design. The
reaction kinetics can be seen as an option for optimization of product yield over the reaction con-
ditions and over the ratios for reactant feed concentrations. Stoichiometry is often assumed to
be given and fixed. Yet stoichiometry can also be a degree of freedom in the design of a process
concept. A simple example will be given of a search over a stoichiometric space to find the best
reaction serving a particular conversion goal. In this case the objective is the chemical conversion
of carbon-dioxide as a greenhouse gas into carbon-monoxide as a chemical reactant in syngas.
The latter is, among others, a feed for methanol synthesis and Fischer–Tropsch synthesis for liquid
hydrocarbon fuels. The reuse of carbon in chemical products helps in closing the carbon cycle.
Approach
Linear algebra analysis of the stoichiometric matrix is applied to extend the range of dry reforming
reactions.
Outcome
The outcome of this reaction modeling example shows how a systematic exploration of the space of
reaction stoichiometry can lead to an optimized yield of a target product. The derived stoichiomet-
ric model enables the exploration of potentially interesting and useful reactions in a given species
space. The super dry reforming reaction is the highest performing reaction for CO2 to CO conver-
sion. An experimental study of this reaction and its inventive underlying catalytic scheme was re-
ported in [17]. The stoichiometric analysis also revealed the high dry and medium dry reforming
reactions as interesting side reaction. In principle, these reactions may occur simultaneously with
super dry reforming. In a reflection on uncertainties and limitations playing a role in this modeling,
one should consider that the outcome space is determined by the chosen set of chemical species.
One could argue that solid carbon (Csolid ) should be added to the chemical species set so as to
allow for more reactions, such as the Boudouard reaction (2 CO ⇔ CO2 + Csolid ).
Case 9.2: A conversion – separation – recycle structure using lumped units
Motivation and objective of modeling

When moving from concept to feasibility stage, one must have some ideas about the use of species
mass resources in a future manufacturing process for the product under consideration. Quite a few
product options may have to be explored, so simple process modeling and fast computing are
preferred. The simplest mathematical models in chemical engineering are linear ones, which can
be formulated and solved easily. As an example, a linear algebraic model of a common reactor-
separator-recycle processing structure is derived and solved. This case may serve as a leading
example on how to tackle such cases in general.
Approach
A process block diagram is generated with a unique labelling of the process units, the streams
connecting the units as well as the chemical species in the streams. The units in the process are
treated in a lumped, input-output way. The functional task(s) of a unit are simply characterized by
means of model parameters, such as degree of conversion, selectivity and separation factors. Lin-
ear component mass balances will be generated over all process units and integrated to balances
over the entire process. A degree of freedom analysis reveals how many external conditions need
to be specified as part of a scenario under which a process operates.
Outcome
The main outcome is a process block diagram (Figure 9.2) represented by a square linear algebraic
model A ⋅ x = b. The vector, x, represents the species mass flows in the process. Hints are given
on performing sensitivity analyses regarding how these mass flows depend on the chosen values
for the process unit parameters. An energy balance is also of importance but it is bilinear (mass
flow multiplied with temperature). However, by knowing the species mass flows and by adding
some reasonable physical approximations, an energy balance is reduced to a linear form and easily
solved.
Case 9.3: Specification of connectivity in a block diagram by use of a structure model

The units in the process block diagram of Case 9.2 are connected by streams (see Figure 9.2). The
models of the process units include the species flows in the connecting streams. If units A and B are
directly connected by stream (l), the models for unit A and B make use of the same flow variables.
The behavioral models for the units account for the connections as well. There is no separate model
to describe the connectivity of units by means of streams. The behavioral and structural models
have been lumped together. However, suppose one would want to rearrange a process block dia-
gram with different connections between process units using alternative streams. In such a case,
the model equations of the units involved in a reshuffle needs to be rewritten because ingoing and
outgoing streams are relabeled. Such rewriting is error prone and may involve quite some book-
keeping effort. Furthermore, it is conceptually unpleasant that a change in the connectivity of units
must result in changes of the model equations describing unit behavior. It would be better to see
instead a change in the connections by streams. In other words, a separation between the behav-
ioral models of the units and a process flow sheet structure model is preferred.
Approach
Each process unit is given slots for the incoming and outgoing flows, where each slot gets its own
set of flow variables. This explicit definition of slot related input and output flow variables allows for
a formulation of the process unit model that is independent of the actual connections occurring in
a process block diagram. The flow sheet structure is used to link stream slots of the process units.
Each process stream must connect an output slot of a unit with an input slot of another unit.
Outcome
A decoupling arises between the behavioral models and structural connectivity of units by streams
in the flow sheet. This preferred way of modeling can be realized at a slight expense of having more
linear connectivity equations in the model. The full set of the connectivity equations forms the
structure model of the block diagram. To cover optional alternative connections between process
units, a formal extension is made to include binary variables to switch such connections on and
off. The price of the additional computing efforts due to the solving of additional linear connectivity
equations is very small; it pales in comparison to the benefit of transparent flow sheet models.
Case 9.4: Optimization of process performance by linear programming (LP)

The species balances over the process units in a process block diagram give rise to a linear model
with some residual degrees of freedom. The latter are taken away by specifying an external sce-
nario under which the process must function. Such a scenario can set some intake rates of the
feeds or targets for the product output rates. When the design problem is fully specified the model
can be solved. One often adds a performance metric for the design; e.g., an economic potential
function and/or chemical element efficiency can be computed. The latter indicates how much of a
key element (say carbon) in the feed streams ends up in the products; it is often expressed as a
fraction.
A full specification of a design problem gives a ‘point’ solution only. Practically it is more
relevant to explore the design decision space for the best performance of a design. To stay in the
linear domain, the performance metric must also be linear in the (unknown) process variables.
Approach
For linear process design optimization one has to do five things.
(a) To open up some degrees of freedom in the design model for optimization.
(b) To select a linear performance metric.
(c) To introduce linear inequality constraints on process variables to demarcate the physical do-
main in which a process design remains physically feasible.
(d) To select a reference production scenario and compute its performance.
(e) To compute the optimum solution and assess its relevance over the reference solution.
Outcome
The example of the conversion-separation-recycle process is used to illustrate an optimization
study. An economic profit function is added to the design model as well as a target output for the
main product with an upper limit on the maximum amount of side product that can be sold. The
optimized outcome is compared with the reference (nominal) situation. The gain in profit can be
explained and visualized in a feed intake diagram. Furthermore, sensitivities of the profit with re-
spect to key processing parameters (selectivity and a separation factor) and a sales parameter are
determined. This sensitivity analysis shows that a future profit enhancement would benefit most
from an improvement in the selectivity factor among these three parameters. While this numeri-
cal example is fictitious, it is the analysis method that is practically relevant to the optimization
outcome.
Case 9.5: Inverse use of linear models for process targeting and estimation

The use of a square linear process model, Ax = b, has been covered in the preceding modeling
cases. The process features are parametric captured in matrix A and vector b, while the unknown
process variables in vector x can be computed. That is, if matrix A is nonsingular. However, there
are more complicated situations of practical relevance in process design and operations where
linear process models play a role.
Let u represent a vector of inputs to a process and y represent the process outputs. These are
related by a linear process model Au = y. The number of inputs and outputs will be unequal in
general. Such a nonsquare linear model can be used in three different ways:
(1) Forward: given A and u, compute y. This case is straightforward and will be skipped.
(2) Targeting: given A and targets for y, compute u.
(3) Estimation: given a sequence of experimental observations of inputs {u1 , u2 , . . . , uK } and of
outputs {y , y , . . . , y } find the connecting process parameters which are the elements of
1 2 K
matrix A.
Targeting and estimation problems will become complicated if the numbers of inputs and outputs
are different. The objective is to offer approximate solutions to the targeting and estimation prob-
lems.
The practical question is when can these methods be applied in chemical product and pro-
cess engineering? In terms of product engineering, one can think of the outputs y representing the
empirical variables for product performance testing, and u representing the internal chemical and
phase composition variables (e.g., for fuels, lubricants, food and healthcare products). In terms
of process engineering, one can think of the relations between the inputs to a process unit and
its outputs. One may also think of an identification of a process control matrix as linking steady
state relationships between inputs and outputs around an operational working point. The practical
restriction is in the assumption of linear relationships between inputs and outputs.
Approach
One wants to find solutions to match in the best possible way, with the linearized model. However,
the ‘best possible way’ has to be defined more specifically. This can be done by means of an ad-
ditional function that aims for minimization of deviations from model linearity. For this purpose
one often uses quadratic functions over the model residuals: ε = A ⋅ u − y. The benefit of using
quadratic expressions as fitting functions in both targeting and estimation is that one can derive
closed from linear analytical expressions as answers.
Outcome
Linear analytic expressions are derived for the following cases:
(a) Targeting when there are more adjustable inputs uN than output targets y , (M < N).
M
(b) Estimation, assuming measurement errors in outputs y only.
(c) Estimation accounting for random measurement errors in both inputs and outputs
The full expressions are given in the Appendix to Chapter 9. These expressions are analytic in form,
yet computationally demanding due to repeated matrix and vector operations. These are better
done numerically; e.g., by use of Matlab™.
Regarding the resulting uncertainties in the computed model variables, it is also possible
to derive estimates for their covariance matrices. Although, the expressions for such covariance
matrices become quite involved.
Statistical targeting and estimation are complicated subjects and it is not hard to get into con-
ceptual traps. It is recommended to consult a book on engineering statistics, e.g., [18], or expert
statisticians. They can advise on suitable error distributions, fitting functions and on the interpre-
tation of the computed results.
Case 9.6: Notes on the use of other types of linear models

To highlight the existence of some other types of linear process models beyond the algebraic ones.
Approach
Offering an overview of the analytical expressions for each type of model
Outcome
Linear models are given for:
(a) Dynamic evolution of process state variables.
(b) Dynamic process models with spatially distributed states in one dimension.
(c) Discrete time process dynamics: development of a FMCG warehouse inventory.
(d) Linear equations with integer coefficients and binary variables.
This overview completes the set of applications of deterministic linear models in the concept stage.
These models are supposed to always contain a fair degree of knowledge of what is supposed to
happen inside a process (gray box model). If it is relatively cheap and easy to collect experimental
data there is an alternative of using fully data driven linear black box models. Such a model can be
generated and used in an experimental optimization procedure. Data driven modeling is covered
as an optimization case study in Section 9.4.
9.3 Nonlinear process model simulations:

contributions to a development stage
9.3.1 Nonlinear modeling and simulation for a development stage
There are many introductory academic text books available on the principles of pro-
cess modeling, associated mathematical analysis of the solutions and the numerical
9.3 Nonlinear process model simulations: contributions to a development stage | 207
simulations in the context of process design and flow sheeting. For background read-
ing one is referred to text books [1–16]. Much of this material is looking inside the
model and how to structure and solve the model equations. This section aims at look-
ing more outwardly at process modeling from a somewhat higher vantage point of
model use. First some generic structural aspects of process models and their use will
be reviewed. This review is supplemented with a few cases highlighting aspects of use
of nonlinear process models.
9.3.1.1 Process representation

Any process model is based on a conceptual representation of a process (to be). A
process is a network consisting of multiple nodes interacting with each other by ex-
changes of physical and information resources. Each node represents a unit operation.
Within a unit operation, several physical and chemical tasks can be carried out in par-
allel; e.g., reactions and separations in a reactive distillation column. The connections
between the units are specified as physical streams and as signals (for measurement
and control) in a process flow sheet. Also, utility streams servicing process units will
be considered as part of the process system.
Looking on a more detailed scale, the inside each unit operation is made up of a
subnetwork consisting of interconnected compartments. Within these compartments,
the states of the resources change because of physical and (bio) chemical phenomena
(e.g., reactions, phase changes, fluid and particles flows, heat and mass transfers).
There can be active agents facilitating the changes of states of physical entities, such
as catalysts and mass separation agents such as membranes. The active agents for
signals in process control and automation are sensors (physical quantities ⇒ infor-
mation signal) and actuators, e.g., valves (signal ⇒ physical quantity). The process
control software using dedicated control laws act as ‘unit operations’ for processing
signals.
Obviously, a process is an open system having connections with the outer world:
– Physical streams entering the process as feeds and leaving as products and
wastes.
– Physical disturbances acting on the process units, such as heat losses to the envi-
ronment.
– Information flows from the process sensors to the plant operators.
– Process control set points are imposed on the process, collectively defining an op-
erating mode.
The interactions between process and its outside world are partially defined by the
operational scenario imposed on a process as well as by the external disturbances,
and partly by the resulting state of the process.
9.3.1.2 Product modeling

In this subsection the focus will be on the modeling and simulation of a process with-
out particular attention to the structural complexities of the product(s). The process
stream delivering the product has no special status; i.e., in the sense of having some
different modeling attributes. The product stream, such as any process stream, is made
up of one or two coexisting fluidic thermodynamic phases. No provision is made for
distributive properties of the product, such as a size distribution of the particulates in
a phase or a chain length distribution of polymers. Also, it is assumed that the proper-
ties (e.g., chemical composition, geometric shape, porosity, etc.) of the active agents
in a process (e.g., catalyst, solvents and membranes) are known and can be used in
the process model. One may want to account for distributive properties of thermody-
namic phases in the process units and in the formation of products. This capability
requires a high level of integrated product and process modeling sophistication (see
Chapter 8). Such a level is still beyond the capabilities of the current generations of
process flow sheeting software.
9.3.1.3 Process equipment modeling

A process plant is made up of process equipment items in which processing opera-
tions take place. In other words, a process equipment piece hosts one or more of the
nodes of the processing network. Regarding the function of process equipment, there
is much more to it than just the process events inside. The complementary aspects are
the integrities of its mechanical condition and its operational software for safety and
control and its availability for the production function the equipment it was designed
for. The latter aspects (its state of availability for normal production conditions, state
of mechanical integrity and safety system aspects) are not part of the conventional
process modeling in chemical process engineering. These aspects are modeled sepa-
rately under the common assumption that actual physical state of the matter inside
a process equipment item has rarely an immediate influence on its mechanical state
and availability. Having outlined these aspects of process representation, the focus
will narrow to the modeling of the physical and chemical events and states inside a
process, along with the flow of resources.
9.3.1.4 Scope of a process model

A process model is an amalgamation of models of the various units in the process,
as well as of the streams connecting these units including utility streams. In contrast,
the unit operations models are an ensemble of compartment models, with sub models
of the behavior of the active agents and of the transfer of the physical and chemical
resources between compartments. It is practical to introduce a formal compact repre-
sentation of the process model. Let process model [M] represent the process system
[PS] at a suitable scale of detail for design purposes:
PS ⇒ [M] M(x; u, d; p, w) = 0
M: Set of all equations describing physical and chemical behavior inside the process.
x: State variables (physical, information) of the process units and internal and exit
streams.
u: Input variables associated with incoming streams that can be manipulated.
d: Geometric design variables (fixed structures and sizes).
p: Physical parameters, having some uncertainty characterized by a covariance ma-
trix.
w: Disturbances, acting more or less randomly on the process from outside or from
inside.
The process model [M] is deterministic while most of its model equations are nonlin-
ear in the process state variables. The model equations are continuous in the variables
and (at least once) differentiable in its variables.
The above model formulation [M] may seem to suggest that this model structure is
restricted in its validity to continuous processes with lumped process units, resulting
in a set of nonlinear algebraic equations. What about distributed process units having
an internal geometric coordinate, such as a tubular reactor? Or a batch process unit
where time is imminent and such a unit is charged and discharged at specific times?
Rather than having a continuous, steady flow in and out, it will be a time varying flow
or even a big chunk of mass transferred at once from one process operation to the
next. Fortunately, the same model format and reasoning applies. The only change is
in a broader interpretation of a state variable x. In a distributed process unit, a state
variable x may have a spatial derivative in direction z: ∇z x = ∂x(z, t)/∂z. Similarly
there can be a time derivative of a state variable: x󸀠 = ∂x(z, t)/∂t. In these cases, one
can extend a single state x to its triple {x, ∇z x, x󸀠 } and insert this triple in the over-
all state vector. Then of course, the process model must include the corresponding
boundary and initial conditions for these derivatives.
The process model [M] describes the internal physical chemical behavior of the
process. However, the modeling problem is still open ended, leaving some degrees of
freedom, because the process model itself does not yet account for the external condi-
tions in its surroundings for which it has to be solved to find the process states x. This
open ended nature is evident from the number of unspecified variables in the model
{u, d; p, w}. A set of specifications for these variables is called a scenario. A scenario
reflects the design and external conditions under which one wants to determine the
physical and (bio) chemical process behavior. It is practical to make a distinction be-
tween the relatively simple process analysis cases and the more complicated process
synthesis cases.
9.3.1.5 Process analysis scenarios

In such cases, a design of the process is available and the design variables d are
known. For instance, one may want to analyze how state variables x (of internal
process units and streams) will vary under variations of inputs (u) and design vari-
ables (d). For that purpose, the process model [M] must be complemented with an
analysis scenario. A typical base case analysis scenario contains the following numer-
ical specifications:
[SDesign ] : d − dref = 0
[SInput ] : u − uref = 0
[Sparam ] : p − pref = 0
[Sdisturb ] : w − waverage = 0
Here, each specific quantity represents a particular numerical value, applicable in the
base case. So, given specifications for all variables {u; d; p, w} except the states x,
which can be solved from M. The combination {x; M and S} is often a very large set of
algebraic variables and equations in equal numbers that need to be solved numeri-
cally. In case of a dynamic process system, one must specify the trajectories of inputs
and disturbances as functions of time.
9.3.1.6 Process performance evaluation metrics

The process model enables computation of its internal states and this model has an
inward focus on what is happening in the process. There is also a complementary out-
ward look on the performance of a process from a societal perspective. The process
under design can be evaluated by means of different, complementary metrics. Here,
three kinds of metrics are distinguished, representing different values to the process
operator, owner and society:
– technical metrics: Etech = Pte (x, u, d)
– economic metrics: Eecon = Pen (x, u, d; cI , cO , pP )
– ecological metrics: Eecol = Pec (x, u, d; eI , eO )
The technical metrics can cover product yields over feed intake(s); energy efficiency;
element efficiency and risks (DF&EI – Dow Fire & Explosion Index). The economic
ones account for size related investment costs (cI ), throughput related operating costs
(cO ) and the proceeds from product sales (pp ). The time value of money can be ac-
counted for by deriving net present value profiles. Ecological metrics will address sus-
tainability issues. For instance, a process design can be assessed by second law anal-
ysis to determine its (irreversible) exergy losses. Once having solved a process model
under a specified scenario with a process simulator, it is relatively a minor effort to
compute the performance metrics as well and perform sensitivity analyses.
9.3.1.7 Sensitivity analyses

One can compute how much small changes in inputs u and in design variables d
will influence the performance metrics (E) through the state variables x. This can be
done by determining the scaled sensitivities of the performance metrics with respect
to the states: SEx = {∂ln(E)/∂ln(x)} and of the process states with respect to the pro-
cess inputs (u): Sxu = {∂ln(x)/∂ln(u)} and with respect to the design variables (d):
Sxd = {∂ln(x)/∂ln(d)}. Process simulators offer these facilities for such local sensitiv-
ity analyses. One may rank these sensitivities in order of decreasing magnitude (i.e.,
using absolute values of computed sensitivities) and see which inputs and design vari-
ables affect the process behavior most strongly. Similarly, one may compute and rank
order the sensitivities with respect to the physical model parameters. The latter is par-
ticularly relevant if some of the model parameters have a substantial uncertainty mar-
gin in their determination from experimental data. An overview of sensitivity analyses
for chemical systems is offered in [19].
Simulations for process analysis are conceptually simple, as a process model com-
putes the states directly from the inputs and the design variables. It is a forward flow
of information in the process model [M]. The situation becomes more complicated
when a few of the key process states are fixed as design targets for process perfor-
mance. Then, one must search for values of the design variables (d) which match the
specified process performance. This is a so called ‘inverse’ problem as the flow of infor-
mation from targets on a few states goes ‘backwards’ to the required design variables
that generate these targets. For process models consisting only of algebraic equations,
this reverse flow of information can usually be handled well without excessive com-
putational efforts. For models containing (partial) differential equations, an inverse
problem is much tougher to solve than the forward problem. That is because the in-
tegration of the equations follows the causal chain of events along the geometric and
time coordinate(s) and there is no way of reversing it. Consequently, the models have
to be repeatedly solved from inputs to outputs, searching for an input that hits a target
output. This iterative procedure can be computationally very demanding.
9.3.1.8 Process design and synthesis cases with targets

In a design situation, one needs to find values for the unknown process design vari-
ables d. Hence, there are more unknown variables: both x and d. The degree of free-
dom in model [M] is now equal to the number of design variables, assuming there is
an external scenario for setting the remaining variables {u; p; w}. This degree of free-
dom must be taken away by setting some targets {t} for the performance of a design.
The number of targets must be equal to the number of process design variables. Oth-
erwise, the design is under, or over, specified. These targets can be imposed both on
important performance indicators as well as on the geometric design variables. For
instance, the yield of a product relative to its raw material, its production rate and
its product quality can be design targets. Rather than giving direct specifications of
geometric design variables certain heuristics for their proportions are applied; e.g.,
fixing the aspect ratio (diameter/length) of a vessel. The specifications can be set up
in three categories: technical process performance, operational conditions of process
units and constraints on geometric design variables. However, the total number of
specs cannot exceed the number of design variables in process model; it must match
exactly.
[Sperf ] : Tperf (x) − tperf = 0
[Soper ] : Toper (x) − toper = 0
[Sdesign ] : Tdesign (d) − tdesign = 0
T*** : Performance target expressed as a function of process states or design

variables
t*** : Numerical value for a performance target
The technical process performance variables express the effectiveness of the process to
turn inputs (feeds and utilities) into outputs (product qualities, yields and amount of
waste). These performance variables must consistently relate to the Key Performance
Indicators (KPIs) used in the higher management levels in a manufacturing organiza-
tion.
Special attention is warranted to the occurrence of different modes of operations.
(a) Normal continuous production over a time period to make the main product
grade(s).
(b) Normal continuous production over a time period to make another product
grade(s).
(c) Dynamic transitions between two successive product grades.
(d) Transitions between different production output rates.
(e) Repetitive cleaning (e.g., for food production) and regeneration operations of pro-
cess units.
(f) Startup from standby (with low holdup and internal circulation) to nominal pro-
duction.
(g) Shutdown from high production to standby level.
Apart from the steady state production modes, all other modes of operations will re-
quires dynamic modeling and simulation.
9.3.1.9 Additional specification of an external scenario for a design

Having set the targets, one must impose additional specifications for the other vari-
ables {u; p; w} in the process model [M]. A process design scenario consists of speci-
fications for the external inputs that can be manipulated, the model parameters and
the external influences (beyond manipulation):
[Sinput ] : u − uref = 0
[Sparam ] : p − pref = 0
[Sdisturb ] : w − waverage = 0
uref : reference values for the process model input variables.

pref : reference values for the process model parameters.
waverage : reference value for the disturbances.
The input specifications will address the flow rates, compositions, temperatures and
pressures of the raw material streams, as well as the conditions of the available utility
streams. The amounts of utility consumed or produced can be made part of the process
state vector as utility amounts depend on the inner workings of the process.
It may happen that a designer faces a situation where more targets need to be set
than there are design degrees of freedom. The excess number of targets going beyond
the number of design specifications can be handled by removing as many specifica-
tions on the inputs. The inputs that are set free will be manipulated in the model com-
putations to meet the excess number of targets.
In a case of a dynamic mode of operation (e.g., a switching between production
modes) the input specifications will be dynamic ones. These dynamic scenarios may
challenge the design performance and call for readjustment of geometric design de-
cision variables, such as lowering volumes and holdups as to increase the rates of
transition (within safe bounds, of course).
The parameters in the process model will, in general, be the statistical averages re-
sulting from a parameter estimation procedure using experimental data. In favorable
circumstances one may even have a covariance matrix (Vparam ) describing the extents
and patterns of uncertainties in the parameters. Specifying the external random dis-
turbances can be conceptually hard. It often involves random (and noisy) phenomena
with poorly specified or unknown probabilistic distributions. The disturbances can
be, for instance, climate related quantities, such as the environmental temperature
and humidity, affecting the processing states. It may also be a random variation or a
gradual decline in the activity of a process agent, such as a catalyst. In design situa-
tions one often takes the expectation value (averaged value of a disturbance variable
taken over a long time interval). It is already a boon to know the spread in the varia-
tion per disturbance variable. Getting a full covariance matrix (Vdisturb ) is practically
ruled out.
9.3.1.10 Consistency in model formulation

When formulating a process model [M] there is a challenge to be consistent in the se-
lection of variables and equations within a physically feasible domain. In other words,
one must avoid redundancies and contradictions in the model. For instance, the de-
sign specification imposed on a process design model should not violate thermody-
namic feasibility and stability. To elaborate, it may happen that a process unit model
in a flow sheet has two thermodynamic phases and the evolution of the states of the
two phases in the model causes one of the two phases to lose its stability. In physical
reality such phases would merge into a single phase. In the model, such spontaneous
merging will not happen. Instead the physical properties of the two thermodynamic
models will start to coincide, possibly leading to linear dependencies between equa-
tions with a resulting loss of model solvability.
The computational solving of a model [M] requires it to be well posed:
– All equations are differentiable with respect to all variables {x; u, d; p, w}.
– The finite derivatives of equations are required to avoid numerical singularity of
the Jacobian matrix J M (= ∂M/∂x).
– Model [M] has as many linear independent equations as state variables x. Lin-
ear dependent equations are combinations of other independent equations and
therefore redundant.
It is assumed that all model equations can be solved simultaneously by means of a

suitable numerical solver. In process simulation practice for novice modelers it hap-
pens that complaints arise about slow convergence to a weird solution or even total
failure of a numerical solver to converge to a solution. The first resort is to attribute
the responsibility for ill performance to the numerical solver. Upon closer inspection
of the model it often appears there is a mathematical or physical inconsistency in the
model. Such an inconsistency will challenge any solver to converge to an unphysical
or nonexistent solution. The first case in this section (case 9.7) is a simple illustration
of a locally ill conditioned model, violating the rule of bounded derivatives and caus-
ing trouble in model simulations. The next case (9.8) deals with the consequences of
a design specification for a reactor violating thermodynamic reaction equilibrium.
9.3.1.11 Process model computations

Having set the design targets {t} as well as the complementary design scenario spec-
ifications for {u, p, w}, one can jointly solve the process model [M] and the target
specifications [S] equations for x and d. The solution that one obtains for a chosen
reference scenario is called the base case solution. One can compute solutions to vari-
ations on this base case by varying the specified reference values for the inputs (u),
physical parameters (p) and the external disturbances (w). Using the solutions in the
perturbed cases in comparison with the reference solution, one can compute the sensi-
tivities. These sensitivities can be ranked in order of (absolute) magnitude to find out
which of the variables going into the process model determine the computed states
the strongest. Figure 9.2 offers a visual overview of a simulation problem for process
design, using design targets for process states.

Design
specifications on
Process model target states (x t )
Fixed inputs System states
(z) Fixed structure (x)
Physical behavior:
M (x; u, d; z, p, w) = 0 Performance
Adjustable inputs
Performance metrics:
(u) metrics (E)
E = E (x, u, d, z)
Adjustable design
parameters (d) Computational engine
Meeting design specifications
by numerical iteration on:
• adjustable inputs (u)
Scenario for simulating a process design: • design parameters (d)
Given: ===> Compute:
{xt , z} & {p, w} {u, d, x, E } & sensitivities
Fig. 9.2: Overview of process simulation for design.
9.3.1.12 Analysis of obtained model solutions

What kind of intricacies of model solutions can one expect when having solved a pro-
cess model?
Though rather rare, one may occasionally find computationally multiple solu-
tions. Theoretically, this can never be ruled out for nonlinear models. Remember that
every polynomial of degree N has N mathematical solutions some of which are in the
complex domain. Fortunately, in many chemical engineering systems there is often
only one real solution among the mathematical set of (complex) solutions. Even if
there would be a few real solutions, only one of these has the better physical stability
characteristics; i.e., being more robust against external disturbances. Usually, it is this
solution that is preferred as the design solution.
Having found more than one solution may also turn out to be a boon. The un-
expected new solution may have a better performance than the anticipated solution.
Moreover, it will make the designers aware that the multiplicity of solutions has impli-
cations for the dynamics and stability of the future process. E.g., there is a possibility
of disturbance driven transitions between steady states.
9.3.1.13 Process performance evaluation metrics

Three kinds of metrics are considered, representing different values to the process op-
erator, owner and society:
– technical metrics: Etech = Pte (x, u, d)
– economic metrics: Eecon = Pen (x, u, d; cI , cO , pP )
– ecological metrics: Eecol = Pec (x, u, d; eI , eO )
Examples of technical metrics are: product yields over feed intake(s), energy effi-
ciency, element efficiency and risks (e.g., by DF&EI). The technical metrics will be
closely linked with the earlier introduced design performance functions (Tperf ) and
targets (tperf ) for the process design. The quantities {cI , cO , pP , eI , eO } in the economic
and ecological metrics are model parameters, related to investments (I), cost of oper-
ation (O) and prices of products (P).
9.3.1.14 Using uncertainty information in a sensitivity analysis

Having obtained a base case design, one has still to consider the impact of the un-
certainties in a design situation on the design outcome and the evaluation metrics.
There can be variability in the specifications for operational and input variables, such
as when there are different production modes. Usually these uncertainties are of a
deterministic nature. A designer may know there can be multiple production modes
with different feed intakes. These variations in the operational and input variables
are often handled by defining a number of likely production modes, as part of speci-
fied external scenarios. The resulting designs for these modes can be compared and
possibly reconciled to arrive at a robust master design that can handle all production
modes. Apart from having distinct production modes, there can be uncertainty about
the proper targets for some of the operational variables as well as about the input
variables and streams. One can analyze the impacts of these uncertainties by means
of model based sensitivity analysis. If necessary, a design can be adapted to attenuate
the sensitivities. Fortunately, there is also a way of coping with remaining sensitivities
to disturbances after completion of a design and after having put the process plant into
operation. Then, one can still compensate for many arising additional uncertainties
by adjusting the operational and input variables through process control to meet set
performance targets.
In case of a rigorous model development with experimental validation and model
parameter estimation, the uncertainties in the system parameters (p) can be coesti-
mated. These uncertainties are expressed by a covariance matrix of the parameter es-
timates. Computationally, this covariance matrix comes for free along with the mean
values of the parameter estimates as obtained in a statistical maximum likelihood pa-
rameter estimation procedure from experimental data. For further model uncertainty
analysis, one may sample parameter values from the space spanned by the covariance
matrix. Given such a set of sampled parameter values, one can compute and analyze
the resulting variances in process design variables (E), states (x) and performance met-
rics (E).
Uncertainties in external disturbances are often harder to cope with. There is often
hardly any information on the distributions of their magnitudes, if anything at all.
These uncertainties are often handled after completing the base case design by means
of:
– Application of experience based overdesign factors to process equipment. There
is a risk of this approach by applying overly conservative factors. Historical factors
may not account yet for the increased accuracy in equipment design possible by
model based design simulations.
– Some form of a mathematical uncertainty analysis, based on probabilistic distri-
butions of the uncertainties and reducing the chance that a design cannot cope
with disturbances to a low acceptable level. Such global system analysis facilities
are offered by advanced simulators, e.g., gPROMS process modeling platform [20].
The above outline completes the story on process modeling for simulation purposes.
The application of process models for optimization of a process design is part of the
next subsection.
9.3.2 Cases for development stage modeling
Three modeling cases are presented. The first two cases will highlight two common
pitfalls in modeling while the third case (Case 9.9) offers a plea for model supported
operability analysis in the design phase.
Case 9.7: Note on ill conditioned constitutive equations in design models

To highlight the detrimental effect of an ill conceived constitutive equation on the solvability of a
process (unit) model.
Approach
To use a continuous flow well mixed tank reactor (CSTR) model with a commonly used reaction rate
expression that has a local physical anomaly.
Outcome
The model has a region in which the solution will numerically fail due to an unbounded first deriva-
tive of the model equation. The implication of this example for modeling practice is that it pays to
carefully check all constitutive equations in a design model on their asymptotic behavior before
embarking on any process engineering computations.
Case 9.8: Consistency of design specifications with thermodynamics

In modeling for process design, it is rather easy to overlook the limits imposed by thermodynamics
on setting feasible design performance targets. If such design targets are set beyond thermody-
namic equilibrium, the computations with the process model will fail. The failure can be due to a
lack of convergence in the computations or finding unphysical design outcomes after ‘successful
convergence’. The purpose of this case is showing such a situation through a reactor design with
an analytical solution.
Approach
The design case is the computation of the required residence time to achieve a target conversion
in a well mixed with a single, reversible first order reaction A ⇒ B at isothermal conditions.
Outcome
Having specified a degree of conversion above the reaction equilibrium, the reactor design equa-
tions generate a numerical solution with a negative residence time as a displeasing feature.
The resulting design heuristic is obvious enough. Physical design targets should be located
within a thermodynamically feasible domain. The preferred distance between the design target
and the thermodynamic equilibrium boundary is negotiable in design.
Case 9.9 Model based analysis of operability aspects in process design

The operational success of a process plant depends, to a large extent, on the quality of its design.
The operational characteristics of process should be taken into consideration in process design
and modeling. The conventional model based process analysis is suitable to study the effect of the
dominant phenomena on the behavior and performance of a process at steady states under normal
operational modes. Operability analysis is needed to assess the effects of major deterministic and
random changes outside and inside a process. Operability analysis of a process has become an
extensive subfield of process engineering, linking process design, control and operations. It relies
much on dynamic process modeling away from the usual steady states. Here, the aim is to create
awareness of the main aspects of operability, while giving some references to pertaining literature
for further study.
Approach and outcome

Following Marlin [21], eight topics of operability are briefly covered from the perspective of teaching
such topics in process design. Dynamic process models are applicable in all of the aforementioned
operability cases, except for reliability and safety. Reliability is an exception as it requires proba-
bilistic models for equipment failure rates. Safety is another exception because it often involves
process events that go far beyond the normal operating window. The conventional process models
are not tuned to such exceptional conditions beyond the normal operating window. Safety requires
dedicated models to study the physical behavior of a process unit under abnormal conditions, such
as runaway or flow reversal effects.
9.4 Nonlinear process model optimization: contributions to a feasibility stage | 219
9.4 Nonlinear process model optimization:

contributions to a feasibility stage
9.4.1 Nonlinear modeling & optimization for a feasibility stage
An explanation is in order of why modeling in the feasibility stage is covered only after
modeling for the development stage. Commonly, in process design, it is the feasibil-
ity stage that precedes the development stage. The reason for switching the order is
rooted in the additional complexity of model use for optimization and its associated
computational procedure. In a feasibility stage one may perform an optimization over
a range of process design options using relatively simple process models to search
for an optimized performance over a wide space of design options. A few of the more
promising process options that are identified in the optimization procedure will be
transferred to the subsequent process development stage. There, such options will be
analyzed in more detail, requiring a more extensive process model along with strong
process simulation capabilities. A process model in the feasibility stage may contain
less detail and be simpler in general than the model for development. The paradox is
that the computational framing of the models in the feasibility stage is seen as being
more complicated due to its optimization features, relative to simulation only in the
development stage.
The main benefit of optimization over simulation in design is that optimization
offers a search over the entire solution space for the best outcome. A simulation re-
sults in a single point in the solution space. When, additionally, a simulation based
global system analysis is applied [22], some more light is shed on the behavior in the
vincinity of that single solution point. In system optimization, one can explore a pre-
defined space for a point of optimum performance. It is because of this capability to
explore a wider space that one wants to apply optimization in the feasibility phase
to identify the more promising options within the entire solution space (for process/
products). Furthermore, advanced optimization methods can also be used for the syn-
thesis of process and/or product structures from a predefined set of building blocks
(e.g., unit operations for a process phase building blocks for a product). In process
simulation, the process structure is fixed a priori. Therefore, it is the combination of
(i) creation of structure, (ii) the computing and simulation of physical behavior and
(iii) optimization of system performance that makes optimization applications attrac-
tive in the feasibility phase. However, there are two restrictions to be accounted for.
The search over a solution space is always a heavy computational task. Therefore the
models of the building blocks for product/process synthesis must remain relatively
simple and should focus on core features only. A similar restriction on models may
emerge from a relative lack of process and product knowledge in the feasibility stage.
The details of the mechanistic functioning of the building blocks may still be lacking
and thus approximate model descriptions are used.
The price to pay for the enhanced capabilities by optimization is a much more ex-
tensive description of the design problem relative to a simulation case. It is necessary
to describe the boundaries of the search space in an optimization as well as the do-
main of validity of the process model. In addition to the usual model equations, one
has the following attributes in an optimization problem:
– Structure related: discrete (binary) variables (Y) to characterize the presence of
building blocks, the types of building blocks and the connectivity of blocks.
– Structure related: equations (equalities and inequalities) to model flow sheet
structure.
– Domain related: valid domain of model application in the optimization.
– Feasible design targets: ranges for variation of process, design and performance
variables.
– Objective function(s): one performance metric or two metrics for a Pareto tradeoff.
9.4.1.1 Overview of common elements in an optimization frame

The elements of a design optimization problem are given:
Process variables: {x, u, d, Y; p, w}

Structure model: [S]: S(Y, x) ≥0
Process model: [M]: M(x; u, d, Y; p, w) =0
Model application domain: [AM ]: AM (x, u, d) ≥0
Design cotargets: [Sdesign ]: Tdesign (d) – tdesign ≥0
Process design scenario: [Sinput ]: u fixed – u ref =0
[Sparam ]: p – pref =0
[Sdisturb ]: w – waverage = 0
Process performance metrics for optimization:
– Technical metrics: max . Etech = Pte (x, u, d)
– EconomicEconomic metrics: max . Eecon = Pen (x, u, d; cI , cO , pP )
– Ecological metrics: max . Eecol = Pec (x, u, d; eI , eO )
Degrees of freedom: dim{d} + dim{Y} + dim{u adjust }
One may wonder about the role of the design cotargets. Why not optimize all design
variables at once? It may happen that a design optimization is part of a retrofit of an
existing process, where the design of certain parts of the process should be retained.
The use of design cotargets allows keeping a portion of the design variables fixed. In a
similar fashion, if one does not want to reinvest in new equipment, one can keep the
existing equipment, fix the design variables and still optimize the process by vary-
ing the operating conditions such as pressure, temperature and internal recirculation
flows. There is often some tradeoff possible between making changes in the geomet-
ric design variables or in the temporal ones, such as operating conditions (pressure,
temperature and/or flow rates).
9.4.1.2 Mixed integer nonlinear programming format

When the physical meaning of the variables in a process model is ignored, one can
write the design optimization in its usual compact mathematical form as a mixed in-
teger, nonlinear optimization problem (MINLP):
min F(x, Y)
{x,Y}
Subject to: g(x, Y) = 0

h(x, Y) ≥ 0
Y ∈ {0, 1} and x ∈ ℜN
The mathematical conditions for solving such a MINLP problem and the various nu-
merical algorithms are covered in textbooks on optimization in chemical engineer-
ing [11–16].
9.4.1.3 Flow of information in the computational process

In a process model the causal flow of physical information runs from physical and in-
formation inputs (feed streams, preset operating conditions and design parameters)
to physical process outputs. When targets are set on the outputs, being some states
and performance metrics, one must start adjusting some inputs and design parame-
ters as to meet these targets. In view of the nonlinearity of a process model as well
as (numerical) stability considerations it is not possible to invert the model as done
easily with linear algebra equations: A ⋅ x = b ⇒ x = A−1 ⋅ b. Stability considerations
play another important role. For instance, if one solves a model of a tubular reactor
by integrating the differential equations in the model from entrance to exit, both the
analytical and numerical solutions are stable. Going the reverse way from exit back to
entrance it is observed that the same solutions exhibit instable behavior, effectively
prohibiting such a route. Consequently, process models must be solved in a forward
manner. When aiming for targets on process states or searching for the optimum per-
formance, the process model must be very many times solved in iterative moves to-
wards the targets or the optimum performance. The causal flow of information in a
process design optimization procedure is shown in Figure 9.3.
For a proper evaluation and interpretation of the solution of an optimization prob-
lem it is strongly recommended to obtain an basic understanding of the role of the La-
grangian function, of the necessary conditions for the existence of a local extremum
(be it a minimum or a maximum) and of the sufficiency conditions for a local mini-
mum. The necessary theory is presented in optimization text books focused on chem-
ical processes, [11, 13–16].
What are practical issues arising in the application of MINLP to process optimiza-
tion?

Design
Process model specifications (xt)
Fixed inputs System states
(z) S (x, Y) ≥ 0 (x)
Optimization
Physical behavior: goals (Et,e,e )
M (x; u, d; z, p, w) = 0
Adjustable inputs AM (x; u, d; z, p, w) ≥ 0 Performance
(u) Performance: metrics (E)

E = E (x, Y, u, d, z)
Computational engine
design (d) structural (Y) Performance optimization

(continuous) (discrete) & meeting design specifications
by numerical iteration on:
• adjustable inputs (u)
• continuous design (d)
• discrete structural (Y)
Scenario for optimizing a process design:
Given: ===> Compute:
{xt, z} & {p, w} {u, d, Y, x, Et,max} & sensitivities
Fig. 9.3: Design optimization by solving an inverse problem.
(a) Establishing the domain of model validity

The derivation of the process model equations is similar to that of process sim-
ulation. The hardship is in getting a good set of inequalities to bound the solu-
tion and so prevent fictitious results outside the domain of model applicability.
In practice it is often harder than expected to collect relevant information on the
bounds of the domain of model applicability. The latter is the domain in which
the model and, particularly, its constitutive equations can be applied with suf-
ficient confidence. Commonly, one is well able to retrieve in the original docu-
mentation of a constitutive equation from which theoretical considerations and
experimental data it was derived. However it is far less common to find, along
with this equation, an explicit specification of the domain in which it can be ap-
plied with a particular degree of confidence. The delineation of such a domain is
rarely given by, for example, a set of inequality constraints. Without such a spec-
ification, the numerical optimizer may move into regions where the performance
seems to ‘improve’, while one or more constitutive equations in the process model
have lost their significance. Such a loss yields a fictitious optimization result. Of-
ten the model developer has to go back to the original literature and look into the
original sets of experimental data and infer over which ranges the experimental
conditions were varied with the resulting ranges of the response variables.
(b) Finding a representation of model uncertainty

Similarly, it can be hard to find proper documentation of the statistical properties
of the estimated parameters in the constitutive equations beyond their mean val-
ues, such as covariance matrices. Having such an uncertainty description would
be helpful in assessing the significance of the computed optimization results in
relation to underlying uncertainties in the process model.
(c) Numerical conditioning of the model
Scaling of process variables and equations is advised due to a wide range in mag-
nitudes. Enthalpy flows and terms in an enthalpy balance may be of the order
of 106 while molar fractions of trace components in mass balances can be of the
order 10−6 . Applying some scaling will help and stabilize the numerical compu-
tations by the optimizer. It will also facilitate any internal local scaling by the op-
timizer.
(d) Initialization of model variables to enhance convergence of iterations on model
equations
The assignments of starting values to the process model variables needs due at-
tention in order to facilitate a smooth convergence to a solution of the model equa-
tions. Inconsiderate assignments of initial values may play havoc with the magni-
tudes of the nonlinear terms in the model equations. Such nonlinearity is often
found in rate expressions (e.g., power-law reaction kinetics) and in thermody-
namic constitutive equations (e.g., for activity coefficients). To get started with
model initialization, it is found helpful, in practice, to artificially tone down the
impact of the nonlinear rate terms in the balance equations. It is convenient to
introduce into the model a common artificial scalar ‘inhibition factor’ as a multi-
plier to all strongly nonlinear terms in the model equations. The purpose of such
an inhibition factor is to slow down the rates of change. Assigning initially a small
value to this inhibition factor, a nonlinear model tends to behave as close to linear.
Then it is easier to achieve initial convergence. Having obtained such a converged
in between solution, one can use it as a starting point for a next round with a grad-
ual ramping up of this inhibition factor. Stepping up the inhibition factor to unity
will restore the original nonlinear features of the model.
(e) Use of suitable modeling platforms for optimization of processes
Modern process model building and computing platforms, based on equation ori-
ented computations, offer compatible links to advanced MINLP optimization fa-
cilities. Examples are gPROMS [20] and AIMMS [23]. The conventional sequential,
block oriented process flow sheet simulators are more restricted and offer some
optimization facilities for process operating conditions (e.g., control set points).
They do not offer optimization of flow sheet (super) structures.
These remarks complete the review of modeling for process (design) optimization. The
next section will deal with four cases to illustrate some aspects of optimization appli-
cations.
9.4.2 Cases for feasibility stage modeling
Four cases are covered that highlight aspects of:

– constraint analysis to identify options for technological innovations
– Pareto tradeoffs in process design between two performance criteria
– synthesis of gas to liquid (GTL) processes
– experimental process optimization with data driven black box models
Case 9.10: Constraints analysis to identify options for technological innovations

Realistic process models should reflect current limitations the technologies underlying pro-
cess unit operations in a process (conversion, separations and compression). In process design
optimization studies such limitations should be expressed as inequality constraints in the pro-
cess model. These inequalities delineate the operational domain in which a particular technology
(e.g., catalyst, membrane) can trustfully be applied (e.g., validity range of reaction kinetics, maxi-
mum allowable temperature, maximum allowable pressure differences). The purpose of constraint
analysis after an optimization of a design is to identify those active inequality constraints that
predominantly limit process performance. Knowing how inequality constraints relate to underly-
ing technologies one finds directions for improving process performance by means of innovations
in the limiting processing technologies.
Approach
Model based process optimization delivers optimized process conditions and design variables.
In addition it will also yield active set of inequality constraints constraining the process perfor-
mance. The relative importance of these inequality constraints can be identified by means of their
associated Kuhn–Tucker multipliers, coproduced as a result of the optimization. One can put the
Kuhn–Tucker multipliers in order of decreasing magnitude. In that way one obtains the order by
which active inequality constraints are restraining the objective function. This approach of finding
the relative impact of active inequalities is explained by means of small reactor design example
Outcome
The result of the reactor design optimization shows that one inequality constraint is strongly dom-
inating the optimum and could be a candidate for relaxation, while another inequality constraint
has only a marginal influence on the performance in the optimum. An analysis of the active in-
equality constraints and their Kuhn–Tucker multipliers in an optimization outcome can help to find
incentives and clues for further technology development and to enhance process performance.
In the same vein, an extensive, deep modeling study was reported in [24] for the conceptual
design of a steam cracking reactor. This design was reoptimized with respect to olefin yields by
lifting the current material induced temperature constraint by 100 K to a new anticipated level of
1300 K. The result is a very significant potential improvement in olefin yields (55 ⇒ 66% mass yield
of ethylene).
Case 9.11: A simple example of a Pareto tradeoff between two performance metrics

Virtually every process design has to consider multiple performance criteria. The ranges of all de-
sign decision variables together form a high dimensional process design space. Each point in this
design space represents the complete array of design decision variables. In these points one can
compute the process performance metrics. This is a low dimensional space of two or three dimen-
sions. The overall objective of a design is to improve the performance criteria as much as possible.
In certain regions of the design space one can move a design from one point to another point and all
performance criteria improve indeed. However, often there appears a transition to another region
where opposition between criteria arises. Making changes in a design can improve one perfor-
mance metric while making anther metric worse. There is no way of improving both. The set of all
designs where one cannot improve a performance metric without making at least one other metric
worse is called a (noninferior) Pareto set of design solutions. If one takes all performance criteria
equally serious, it is impossible to say which design within this Pareto set is the better one. All are
equally qualified as ‘best’ designs.
Working with multiple performance functions in a design gives conceptually rise to Multi Ob-
jective Nonlinear Programming (MONLP). In process engineering practice one wants to keep using
a normal single objective Mathematical Programming format and associated nonlinear program-
ming (NLP) solver. Then there are three options to deal with multiple performance criteria. The
assumption is that all criteria have to be minimized.
(1) To set a threshold level for the performance variable of each criterion, except one. The latter
one is called the dominant criterion. The purpose of having thresholds is to relax from the min-
imization and stay well above the attainable minimum level of performance for each criterion.
Each threshold will appear as an additional inequality constraint in the optimization. Next, the
design is optimized with respect to the dominant criterion, while making sure that all other
criteria meet their threshold levels. When a criterion is at its threshold level (i.e., becoming
an active inequality), its Kuhn–Tucker multiplier indicates the sensitivity of the dominant cri-
terion with respect to the criterion at threshold. The outcome may give reason to adapt the
thresholds and repeat optimization till the results are “satisfactory” to the problem owners
of a design. In other words, “satisfactory” depends on a reflective judgement by process en-
gineering management, depending on societal and company cultures as well as on local site
conditions where the process will be erected.
(2) To define a new master criterion as a linear weighted sum over all separate criteria. Often
economic performance is used as the dominant criterion and all other criteria must be trans-
lated into monetary values. The weight factors in the linear sum expression represent the
relative economic weights of the various criteria. The approach under 1) can help to establish
reasonable estimates for such weight factors (≈ using Kuhn–Tucker multiplier). One can then
minimize the new master criterion. If the results are not satisfactory to the problem owner of
a design one may start iterative improvements in adjusting the relative weights in the master
criterion.
(3) To create a Pareto set of design solutions, where each design in the set will have different
performance values. Such a Pareto set can be generated by means of approach 1) where the
selected threshold levels (initially high and relaxed) are systematically stepped down is sever-
ity. Till the level for that particular criterion is so low that optimization with respect to the
dominant criterion becomes infeasible.
Approach
A Pareto optimization is demonstrated by means of a small scale reactor design problem with two
performance criteria: profit and mass yield.
Outcome
The results indicate there is a region of design in which both performance criteria simultaneously
improve. Adjacent is another region with a Pareto front: improvement of one criterion makes the
other criterion worse. A similar result is obtained in an optimization of the design of a reactive – dis-
tillation column for MTBE synthesis [25]. An economic performance indicator is placed in a tradeoff
with an ecological indicator, the exergy efficiency of this column.
Case 9.12: Optimization of designs of gas to liquid processes

The main purpose of GTL processes is to turn natural gas (NG) into liquid fuels. GTL processes tend
to serve huge gas fields and are truly large scale. The capital investment is very high (up to several
billions US$). There are many design decisions with interacting outcomes on process performance.
Also, tradeoffs emerge between capital investment, carbon and energy efficiencies. Therefore, GTL
processes are prime candidates for model based optimization process synthesis.
Approach
The feasibility of such an approach is shown in an industrial case study on model based, optimiza-
tion driven synthesis of a large scale GTL process [26], using profitability and carbon efficiency as
performance measures. The process flow sheet is represented by means of a block diagram with
reactors and unit operations. The flow sheet is generalized to cover many possible synthesis op-
tions. The reactor models and unit operations are reduced to their core features to keep the model
relatively small. An extensive set of inequality constraints is implemented to make sure the condi-
tions in the reactors and the unit operations will stay in validated domain during optimization.
Outcome
An industrial case study [26] shows a design region in which both performance metrics can be
improved simultaneously. However, after reaching the top in profitability there is a Pareto frontier
with a tradeoff: an improvement in carbon efficiency leads to a decrease in profitability.
Few other industrial case studies of MINLP optimizations for process synthesis are found in
the scientific literature. This is compensated for by a growing number of academic publications.
Floudas and coworkers have applied optimization comprehensively to synthesize XTL processes.
In XTL, the symbol X stands generically for either Biomass or Coal or Natural Gas or even all feeds in
parallel. Also, in these studies the product slate has been extended from liquid hydrocarbon fuels
to chemicals, including methanol, DME and olefins, [27]. It is even possible to extend optimizations
to the synthesis of a national supply chain (for the USA) for the best locations for new GTL processes
given local sources of NG and main customer markets for liquid fuels [28].
Case 9.13: Process performance optimization by model and experiments

Some new products, as well as the processes to make these products, are conceived experimen-
tally in the concept stage. Some further optimization is warranted before moving to the next stages.
Such an optimization is part of the feasibility stage. If a product or a process is complicated and
only poorly understood, an effort to mechanistic modeling of its behavior and performance could
be daunting and costly. In such situations an experimental optimization of the product/process
performance with the help of data driven models can be more effective. Data driven process model
development is sometimes cheaper and quicker than developing a (semi)mechanistic model. How-
ever, it is required that one can indeed measure all input and output variables of a product/process,
relevant to the characterization of its behavior and performance.
9.5 Concluding remarks | 227
Approach
A data driven modeling and optimization procedure is presented that applies equally well to the
optimization of process and product performance. This story will focus on the conceptual outline
of experimental process optimization, using a response surface approach.
Outcome
The outcome is a procedure leading to a parametric process model and its experimental validation.
The model is repetitively improved through the following sequence:
(a) Making a computational prediction of the conditions of optimum performance.
(b) Experimental validation of this point.
(c) Adaptation: if there is still a significant deviation between the model predictions and the ex-
perimental outcomes, the most recent experimental data are used to refine model parameters.
In the end, the model has parameter estimates accurate enough to closely predict the optimum
performance point. The strength of this successive approach of gradual refinement of parameter
estimates is that model outputs and experimental data should become consistent in the region of
optimum performance.
Another similar response surface approach is reported in [18, (Chapter 20.5)], on the modeling
and optimization of a coating process. A design of experiments as well as the successive response
surface modeling (regression) are extensively discussed and analyzed for statistical significance
of the results. This application is an adaptation from a training exercise in E.I. du Pont de Nemours
and Company (USA). It is possible to extend the response surface approach to dynamic process
systems [29], such as (semi-)batch reactors. One applies a dynamic design of experiments. Upon
collection of the corresponding experimental data, a black box or gray box model can be fitted to
the data and applied for process operations (finding optimal trajectories) and control.
9.5 Concluding remarks
This chapter has highlighted approaches to model development and applications re-
lated to the concept, feasibility and development stages of product and process devel-
opment and design. The emphasis here has been on making effective and inventive use
of models for process innovation and optimization, illustrated by means of small scale
examples. This amounts to making useful interpretations of modeling and computing
results. The balance in the use of models has been in favor of steady state (algebraic)
models, at the expense of modeling of process dynamics for control and (dynamic) op-
timization. However, the core thinking patterns in analyzing models and their results
remain the same.
List of symbols, subscripts and abbreviations:
Symbols
A real (non)square matrix (real values as entries)
b real vector
c cost related parameters in economic model expressions
d symbol for design parameters (continuous) in a process model
E set of quantitative performance metrics for a process design
N dimension of a vector
p generic symbol for a set of physical parameters in a process model
pP proceeds from product sales, in economic model equation
P performance function
t target value for a design quantity or performance function in process
design
T function expressing a metric (function) related to a target
u generic symbol for inputs to a process model
w disturbance variables in a process model
x generic symbol for physical states in a process model
y set of measured outputs of a process (model)
Y generic symbol for structural design parameters (discrete) in a
process model
z generic symbol for a set of adjustable inputs to a process model
ε deviation between computed and experimental value of a process
output variable
ℜ mathematical space of real variables
∇ spatial derivative
Subscripts
Average averaged value of a quantity
design design related
ecol ecology related
ecom economy related
oper operations related
perf performance related
ref reference value of a quantity when making a comparison
tech (process) technology related
Abbreviations
CSTR Continuous flow, well stirred tank reactor
DF&EI Dow Fire & Explosion Index
DME Di-Methyl-Ether
GTL Gas to liquid
KPI Key performance indicator
LP Linear programming
MINLP Mixed integer nonlinear programming
MONLP Multi objective nonlinear programming
MTBE Methyl tertiary butyl ether
NG Natural gas
NLP Nonlinear programming
XTL X = {biomass, coal, natural gas} to liquid fuels
[1] Dimian AC, Bildea CS, Kiss AA. Integrated design and simulation of chemical processes, Sec-
ond edition, Elsevier, Computer-Aided Chemical Engineering, 35, 2014.
[2] Aris R. Mathematical modeling techniques, Research Notes in Mathematics, 24, Pitman Ad-
vanced Publishing Program, 1979.
[3] Bequette BW. Process Dynamics, Modeling, Analysis & Simulation, Prentice Hall, 1998.
[4] Aris R. Mathematical modeling, PSE Series Vol. 1, Academic Press, 1999.
[5] Hangos KM, Cameron IT. Process modeling & analysis, PSE Series Vol. 4, Academic Press,
2002.
[6] Cameron IT, Gani R. Product & process modeling: A case study approach, Elsevier, 2011.
[7] Rice RG, Do DD. Applied Mathematics & Modeling for Chemical Engineers, Second Edition,
Wiley, 2012.
[8] Zondervan E. A Numerical Primer for the Chemical Engineer, Taylor & Francis (CRC Press), 2014.
[9] Rudd DF and Watson CC. Strategy of process engineering, John Wiley &Sons, Inc., 1968.
[10] Westerberg AW, Hutchison HP, Motard RL, Winter P. Process flowsheeting, Cambridge Univer-
sity Press, 1979.
[11] Biegler LT, Grossmann IE, Westerberg AW. Systematic methods of chemical process design,
McGraw Hill, 1997.
[12] Seider WD, Seader JD, Lewin DR, Widagdo S. Product and Process Design Principles. Synthesis,
Analysis, and Evaluation, 3rd edn, John Wiley & Sons, Inc., 2010.
[13] Floudas CA. Nonlinear and mixed-integer optimization. Fundamentals and Applications, Oxford
University Press, 1995.
[14] Floudas CA. Deterministic Global Optimization: Theory, Methods and Applications, Kluwer
Academic Publishers, 2000.
[15] Edgar TF, Himmelblau DM, Lasdon LS. Optimization of chemical processes, 2nd edn, McGraw-
Hill, 2001.
[16] Biegler LT. Nonlinear Programming: Concepts, Algorithms and Applications to Chemical Pro-
cesses, SIAM, 2010.
[17] Buelens LC, Galvita VV, Poelman H, Detavernier C, Marin GB. Super-dry reforming of methane
intensifies CO2 utilization via Le Chatelier’s principle, Science 354(6311), 2016, 449–452.
[18] Ogunnaike BA. Random Phenomena. Fundamentals and Engineering Applications of Probabil-
ity and Statistics, CRC Press, Section 20.5, 879, 2010.
[19] Varma A, Morbidelli M, Wu H. Parametric sensitivity in chemical systems, Cambridge University
Press, 1999.
[20] gPROMS process modeling platform of PSEnterprise (London, UK), https://www.psenterprise.
com/products/gproms/platform.
[21] Marlin TE. Teaching operability in undergraduate chemical engineering design education.
Computers & Chemical Engineering, 34, 2010, 1421–1431.
[22] Global system analysis in gPROMS process modeling platform of PS Enterprise, London, UK,
https://www.psenterprise.com/products/gproms/technologies/global-system-analysis.
[23] AIMMS (Advanced Interactive Multidimensional Modeling System) of paragon decision
technology, Haarlem, NL, https://aimms.com/english/developers/resources/manuals/
optimization-modeling/.
[24] Van Goethem MWM, Barendregt S, Grievink J, Moulijn JA, Verheijen PJT. Model-based, thermo-
physical optimization for high olefin yield in steam cracking reactors, Chemical Engineering
Research and Design, 88, 2010, 1305–1319.
[25] Almeida-Rivera CP, Grievink J. Process Design Approach for Reactive Distillation Based on Eco-
nomics, Exergy and Responsiveness Optimization, Industrial & Engineering Chemistry Re-
search, 47, 2008, 51–65.
[26] Ellepola JE, Thijssen N, Grievink J, Baak G, Avhale A, van Schijndel J. Development of a synthe-
sis tool for Gas-To-Liquid complexes, Computers and Chemical Engineering, 42, 2012, 2–14.
[27] Baliban RC, Elia JA, Weekman Vern, Floudas CA. Process synthesis of hybrid coal, biomass,
and natural gas to liquids via Fischer–Tropsch synthesis, ZSM-5 catalytic conversion, methanol
synthesis, methanol-to-gasoline, and methanol-to-olefins/distillate technologies, Computers
and Chemical Engineering 47, 2012, 29–56.
[28] Elia JA, Baliban RC, Floudas CA. Nationwide, Regional, and Statewide Energy Supply Chain Op-
timization for Natural Gas to Liquid Transportation Fuel (GTL) Systems, Industrial & Engineering
Chemistry Research, 53, 2014, 5366–5397.
[29] Klebanov N; Georgakis C. Dynamic response surface models: A Data driven approach for the
analysis of time-varying process outputs, Industrial & Engineering Chemistry Research, 55,
2016, 4022–4034.
10 Evaluating economic performance
10.1 Introduction
Before companies agree to invest large amounts of capital on proposed projects, their
management must be convinced that the selected project(s) will provide a sound re-
turn on investment (ROI) compared to alternatives. Therefore, it is crucial that eco-
nomic evaluations are made before initiating a project, at various stages in its devel-
opment, and before attempting the design of a process and plant. These evaluations
determine whether a project should be initiated, abandoned, continued or taken to
the scale up and implementation stages. Furthermore, economic evaluations pinpoint
those parts of the project (product formulation, manufacturing and process) that re-
quire additional investigation. Even if the available technical information is insuffi-
cient for complete product/process design, economical evaluations must still be made
to determine whether the project is economically (more profitable than competing
projects) and financially (capital required for implementation can be raised) feasible.
Although cost estimation is a specialized subject and a profession in its own, design
engineers must be able to make rough cost estimates to support the decisions between
project alternatives and optimize the design.
The purpose of most chemical engineering design projects is to provide the input
information from which estimates of capital and manufacturing costs can be made
and subsequently the profitability of a project can be assessed. Therefore, this chapter
first introduces the principal methods used for comparing the economic performance
and profitability of different projects. Subsequently the components comprising the
manufacturing costs and the methods used for capital costs estimating are introduced
and discussed.
10.2 Economic project evaluation
10.2.1 Project cash flow
During any project, cash initially flows out of the company to cover the engineering
costs, equipment procurement, plant construction, and plant startup. After construc-
tion is completed the plant is put into operation and revenues from product sales start
to flow into the company. The net cash flow (CF) is the difference between earnings
and expenditures. A cash flow diagram, such as that shown in Figure 10.1, illustrates
a typical cumulative net CF forecast over the life of a project. The CFs are based on
estimates of investment, operating costs, sales volume and sales price for the project.
The diagram can be divided into the following characteristic regions:
https://doi.org/10.1515/9783110467741-010
232 | 10 Evaluating economic performance
Fig. 10.1: Project cash flow (CF) diagram.
– Investment required to design, construct and provide funds for startup of the
plant.
– Upturning cash flow curve due to a positive net cash flow as income is generated
from sales. The cumulative amount remains negative until the investment is paid
off, which is known as the break even point.
– Positive cumulative cash flow (CCF) where the project is earning a return on the
investment.
The CF summarizes all the incoming and outgoing CFs through the operating year.
Investments are considered a negative CF, while profits (after tax) plus depreciation
are positive CFs. During the years of plant construction the CF for a particular year n
is:
CFn = −φTDC − WC − CLand (10.1)
The fraction of total depreciable capital (TDC) expended that year is φ, WC is the work-
ing capital and CLand is the cost of land expended during that year. WC comprises the
additional funds, on top of the costs to build the plant, to startup the plant and keep
it in operation. It flows in and out of an existing operation and is usually assumed to
be completely recoverable at the end of a project without loss. As illustrated in Fig-
ure 10.2, WC is divided into two main categories:
(1) Current liabilities that consist of bank loans and accounts payable (money owed
to vendors for various purchases).
Fig. 10.2: Flow of working capital (adapted from [2]).
(2) Current assets that comprise

– available cash (salaries, raw material purchases, maintenance supplies and
taxes).
– accounts receivable (products shipped but not yet paid and extended credit
to customers).
– product inventory (material in storage tanks and bins).
– in process inventory (material contained in pipe lines and vessels).
– raw material inventory (material in storage tanks and bins).
– spare parts inventory.
WC can vary from as low as 5% of the fixed capital for a simple single product, process,
with little or no finished product storage, to as high as 30% for a process producing a
diverse range of product grades for a sophisticated market, such as synthetic fibers.
A typical figure for general chemicals and petrochemical plants is 15% of the total
permanent investment (TPI) as defined in Figure 10.7.
The CF for a particular year of plant operation is calculated from the operating income
(OI) resulting from subtracting the operating expenses (variable and fixed costs) from
the sales or revenue for that specific year:
OIn = S − VC − FC (10.2)
Here, S is the annual sales, VC the annual variable costs and FC the annual fixed costs.
The total annual product sales revenue is the sum of the unit price of each product
multiplied by its sales quantity. When the depreciation (capital costs) are subtracted
from the OI, the annual gross profit (GP) is obtained:
GPn = S − VC − FC − D (10.3)
Where D is the annual depreciation calculated by dividing the total depreciable cap-
ital, CTDC in Figure 10.7, by the depreciation period of N years, often chosen equal to
10 years. The annual net profit (NP) represents the net result in a specific year and is
calculated by subtracting income taxes from the GP, where t is the taxation rate:
NPn = GPn × (1 − t) (10.4)
As depreciation is considered as an internal expense that is retained in the company,

the CF from plant and product manufacturing operations for year n equals the after
tax earnings plus the depreciation:
CFn = (1 − t)(S − VC − FC − D) + D (10.5)
The cumulative cash flow (CCF) sums calculated CF for each year up to the point of
the calculation:
N
CCFN = ∑ CFn (10.6)
n=0
Where CCFn is the cumulative CF for a period of N operational years, CFn the CF for
year n, and n a positive integer for the operational year. The discounted cash flow (DCF)
calculates the value of current money in the future considering that the value of money
earned in the future is less than money in the present. This calculation is dependent
on the discount rate, which further decreases the value of money when it increases.
The discount rate used is generally the appropriate weighted average cost of capital
(WACC) that reflects the risk of the CFs by incorporating the time value of money and a
risk premium. Through the time value of money (risk-free rate) investors are compen-
sated for having to wait until their investment is paid back. The risk premium reflects
the extra return investors demand because they want to be compensated for the risk
that the CF might not materialize after all. As a result the discount rate typically ranges
from 10% for low risk projects up to 20% for high risk projects.
CFn
DCFn = (10.7)
(1 + r)n
Where DCFn is the discounted cash flow for year n and r the discount rate. The cumula-
tive discounted cash flow (CDCF) calculation uses the calculated discounted cash flow
for each year and sums them up to the point of the calculation. This number is indica-
tive for the net result of a business case. When the CDCF equals to zero, the internal
rate of return (IRR) is equal to the discount rate.
N
CDCFN = ∑ DCFn (10.8)
n=0
CDCFn is the cumulative discounted cash flow for a period of N years.

10.2.2 Economic potential project evaluation method
Before even starting a design, it is crucial to verify that the target product(s) at
least provides a value that is higher than the minimal consumption of raw ma-
terials. Such evaluation requires nothing more than first indications of possible
sales prices, raw material prices and targeted production volume, which are typi-
cally obtained from market data. For important chemicals, market data and prices
are recorded and published by various organizations such as ICIS (http://www.
icispricing.com), IHS (www.ihs.com/industry/chemical) which also publishes the
Chemical Economics Handbook containing market overviews for >300 compounds,
Nexant (www.nexant.com/industries/chemicals), Orbichem (www.orbichem.com)
and Intratec (www.intratec.us) or can be obtained via online brokers and suppliers
(www.business.com/directory/chemicals, www.alibaba.com). For specialty chemi-
cals and new products it is often more difficult to obtain reliable price data. In such
cases consultants can be involved to provide economic and marketing information,
or estimates of attainable price levels need to be made based on envisioned product
performance and related/competing products in the targeted market(s).
Based on the obtained price data indications and calculated raw material con-
sumptions, a first estimate of the economic potential (EP) of a project can be made by
simply substracting the annual cost of the raw materials from the expected annual
sales revenues:
EP = S − RMC (10.9)
S is the annual sales and RMC is the the annual raw material costs. The sales can con-
stitute only the main product but also the known byproducts. The main benefit of this
first high level evaluation is that nothing more than a black-box input-output diagram
and overall mass balance is required to assess the difference in the EP of the various
design alternatives. For most projects, this is a valuable approach as raw material costs
are the major part (50–90%) of the total manufacturing costs. When required, this EP
estimation can be refined to include utility costs, but then more detailed information
on the energy balances is also required.
10.2.3 Simple project evaluation methods
This section introduces some simple economic indicators that can be calculated
quickly once the project investment and CFs are estimated. These indicators are widely
used for preliminary screening of project attractiveness. By dividing the total capital
investment (TCI) (fixed capital plus WC; see Figure 10.7) by the average annual CF, the
pay back time (PBT) is obtained:
CTCI
PBT = (10.10)
CF
Here, CTCI is the total capital investment and CF the average annual CF. This is not the
same as the breakeven point indicated by the cash flow diagram, as it assumes that all
the investment is made in year zero and revenues begin immediately. For most chemi-
cal plant projects, this is not realistic as investments are typically spread out over one
to three years and revenues may not reach 100% of design basis until the second year
of operation. The simple PBT is strictly based on a CF but, for simplicity, taxes and
depreciation are often neglected and the average annual income is used instead of
CF. Another simple indicator of economic performance is return on investment (ROI),
which is defined as:
NPn
ROI = × 100% (10.11)
CTCI
This is generally stated as a percentage per year. The required ROI will typically de-
pend on the degree of risk associated with a project. For low risk projects, a ROI of
10% might be sufficient while for high risk ventures (new products, new markets and
new technology, for example) the required ROI might be as high as 50%.
10.2.4 Present value project evaluation methods
The above simple economic indicators do not capture the time dependency of CFs
during the project. CF timing is crucial to investors because not all capital must be
financed immediately and money earned can be reinvested and start to earn a return
as soon as it is available. For that reason money earned in the early project years is
more valuable than that earned later. This time value of money can be accounted for
by bringing the net CF in each year of the project to its ‘present value’ at the start of the
project through discounting it at some internally agreed discount rate, which is set at
the WACC in most companies.
The net present value (NPV) is one way to value a project in terms of CFs and the
value of money. It states how much the business is worth at a certain point of time;
typically a 10 year period is taken. It includes future revenues, expenses and converts
those CFs into a positive or negative value.
N
CFn
NPVN = ∑ (10.12)
n=1
(1 + r)n
The internal rate of return (IRR) is directly linked to both the discount rate and the NPV.
It says something about the rate of growth, which a project is expected to generate. By
calculating the NPV at various interest rates, it is possible to find an interest rate at
which the cumulative NPV at the end of the project is zero:
N
CFn
NPVN = ∑ (10.13)
n=1
(1 + IRR)n
This particular rate is called the IRR and is a measure of the maximum interest rate
that the project could pay while still breaking even by the end of the project life. The
more positive the IRR is, the more favorable the project is in terms of ROI. Often, this
value is used for the comparison of different projects, as it is relatively easy to calculate
and compare. Companies usually demand IRR levels higher than the internally agreed
discount rate before they invest in a project.
10.3 Manufacturing costs
To determine the financial attractiveness of a project, the manufacturing (production)

costs of a product are needed. Figure 10.3 separates the total production costs into
three main categories: direct costs, indirect costs, and general costs. Direct (variable)
costs are proportional to the production rate, whereas the indirect and general costs
(fixed and plant overhead costs) tend to be independent of the production rate. Gen-
eral costs cover management of the firm, marketing the product and research and de-
velopment.
Fig. 10.3: Breakdown of total production costs. (Adapted from [2]).
10.3.1 Direct costs
In calculating the direct costs, the annual cost of each raw material is found by mul-
tiplying the annual consumption by the price. Raw material costs often dominate the
overall manufacturing costs. Consumables such as acids, bases, absorbents, mem-
branes, solvents and catalysts are depleted or degraded over time and require replace-
ment. Utilities comprise of steam, electricity, fuel, cooling water, process water, com-
pressed air, refrigeration, waste treatment and landfill. Utility equipment is usually
located outside the process area (OSBL) and may supply several processes. Analogous
to raw materials the annual cost of each utility is determined by multiplying its annual
consumption by its price.
Chemical plants require operating labor to produce a chemical, and maintenance
labor to maintain the process. There is also indirect labor, needed to operate and main-
tain facilities and services. When developing a new process the quantity of operating
labor can often be estimated in preliminary cost analysis from company experience
with similar processes or published information on comparable processes. If a flow-
sheet or process flow diagram (PFD) is available, the operating labor can be estimated
from the number of operations based on previous experience:
Annual Labor Cost

Operator/Shift cost/year
= ∑ (Equipment type × ) × (number shifts) × ( ) (10.14)
Unit operator
Table 10.1 indicates some typical labor requirements for various types of process
equipment. The number of shifts varies between countries but is generally 5 for 24/7
operating continuous plants and 4 for 24/5 operating batch plants. The required
amount of operating supervision is closely related to the total amount of operating
labor and typically averages about 15% of the operating labor costs.
Tab. 10.1: Typical labor requirements for process equipment [9].
Equipment type Operators/unit/shift

Compressors and blowers 0.15
Centrifuge 0.35
Crystallizer 0.20
Dryer 0.50
Evaporator 0.25
Filter (continuous/batch) 0.15/1.0
Heat exchanger 0.10
Vessel, tower 0.2–0.5
Reactor (continuous/batch) 0.5/1.0
Maintenance costs consist of materials, labor, and supervision and may sometimes
be higher than the cost of operating labor. Although maintenance costs increase as a
plant ages, for economical estimates, an average value over the plant life is assumed.
The maintenance supplies (materials and services) are typically estimated as a freac-
ton of the TDC investment and range from 3.5% for a fluids processing plant to 5%
for a solids processing plant and 4.5% for solids-fluids processing. Maintenance labor
Tab. 10.2: Breakdown of direct operational and indirect costs [4].
Operational costs
Operating labor See equation (10.14)
Operating supervision 15% of operating labor
Operating supplies 15% of maintenance supplies
Maintenance supplies
– Fluid handling 3.5% of total depreciable capital (C TDC )
– Fluid-solids handling 4.5% of total depreciable capital (C TDC )
– Solids handling 5.0% of TDC (C TDC )
Maintenance labor, 120% of maintenance supplies
supervision
Laboratory charges 15% of operating labor
Royalties 4% of total production costs without depreciation
Indirect costs
Overall plant overhead 22% of operating + maintenance labor and supervision

Fixed costs
– Depreciation 10% of total depreciable capital (C TDC , chosen lifetime dependent)
– Property taxes 2% of TPI (C TPI )
– Insurance 1% of total depreciable capital (C TDC )
– Interest Excluded from CF when profitability calculation is the main objective
– Rent 0% of total permanent investment (C TPI )
General costs
– Administration 2% of sales
– Marketing & sales 3% of sales (transfer 1% of sales)
– Research & development 5% of sales (2–10% depending on business)
wages and benefits are typically of the same magninitude as the mainenance sup-
plies costs. The costs for engineers, supervision and overhead are estimated at 30% of
the direct maintenance labor. Custodial supplies, safety items, tools, column packing,
and uniforms, which are not raw materials or maintenance supplies, are considered
as operating supplies. A typical cost value for these types of supplies is 15% of the
maintenance supplies.
In order to meet the specifications of the products, regular analysis of process
streams is mandatory to determine and control their quality. As initial estimates, these
laboratory charges for quality control typically range from 10 to 20% of the operating
labor. The amount of royalties (licensing fees) depends heavily on the uniqueness of
the process and can range from 0 to 6% of the total production cost, or alternatively,
0 to 5% of product sales. All these direct costs components are summarized in Ta-
ble 10.2.
10.3.2 Indirect and general costs
As depicted in Figure 10.3, indirect costs consisting of fixed and plant overhead costs,
change only little or not at all with the amount of production (plant capacity) because
they are indirectly related to the production rate.
Fixed costs include depreciation, property taxes, insurance, interest and rent.
These expenses are a direct function of the capital investment and financing struc-
ture. Rent is usually not taken into account in a preliminary estimate. The initial
investment is paid back by charging depreciation as a manufacturing expense. Al-
though there are several depreciation methods, using a constant yearly depreciation
rate for a fixed period in preliminary economic evaluations is commonly accepted.
A manufacturing plant or individual equipment has an economic, a physical, and a
tax life. The economic life is the period until a plant becomes obsolete, the physical
life when maintenance of a plant becomes too costly, and the tax life is fixed by the
government. The plant life is usually set between ten and twenty years. Deprecia-
ble capital includes all incurred costs for building a plant up to the point where the
plant is ready to produce, except land and site development. The magnitude of local
property taxes depends on the particular location of the plant and may be initially
be estimated at 2% of the TDC investment. Property insurance rates, typically 1% of
the TDC per year, depend on the actual operation carried out in the manufacturing
plant and the extent of protective measures taken. Interest is only a cost when it is
necessary to borrow the capital needed to invest in the plant. However, in most cases
the capital originates from the company and/or investors making project profitability
calculation the main evaluation objective. Therefore, interest is normally excluded
from the cash flow. Plant overhead are the costs involved for services and facilities to
make the complete plant function as an efficient unit. These costs are closely related
to the costs for all labor directly connected with the production operation and amount
to about 22% of the total direct labor costs (operating, maintenance and supervision).
In addition to manufacturing costs, other general costs are associated with man-
agement of a plant. Main categories to be included are administrative, marketing,
sales, research and development. Administrative costs are those expenses connected
to executive and administrative activities. For a preliminary estimate they may be ap-
proximated by 2% of the sales. Marketing and sales costs vary widely for different types
of manufacturing plants. For larger volume products 3% of the product sales is con-
sidered a reasonable estimate. When the product is only transferred ‘over the fence’,
1% of sales is a suitable approximation.
Finally, the process and product need to be continuously improved. Therefore,
the cost of research and development should be added to the production cost. In the
chemical industry these costs range from 1 to 10 percent of the total sales depending
on the business. Table 10.2 summarizes the indirect and general cost components.
Bill of Materials Annual producon xxxxx ton/year

Raw materials Consumpon Price Cost Annual Anual costs
(unit/kg product) (€/unit raw (€/kg product) consumpon (€/year)
material) (unit/year)
Reactant A
Reactant B
Catalyst
Solvent
Carbon
...
BOM Total raw materials
(Bill of
Materials) Ulies
Steam (ton) 25
Electricity (kWh) 0,07
Natural gas (GJ) 1,9
Process water (m3) 0,35
Cooling water (m3) 0,05
Waste water (m3) 1
Landfill (ton) 20
Other waste (ton) 300
...
Total ulies
Total variable costs
+ Operaonal cost
Operang cost
Labor
Supervision
Supplies
Direct Maintenance cost
cost Supplies
Labor, supervision
Laboratory charges
Royales
Total operaonal cost
+ Indirect cost
Plant overhead
Indirect Depreciaon
cost Other
Total indirect cost
Manufacturing cost
+ General cost
Administraon
General Markeng & sales
cost R&D
Total general cost
Total producon cost
Fig. 10.4: Cost sheet outline.
10.3.3 Cost sheet
For economic project evaluation, accuracy is crucial in estimating total production

cost. The most important factor is to make sure all costs associated with making and
selling the product are included. These are included in the cost sheet depicted in Fig-
ure 10.4. In this cost sheet the raw materials and utilities comprise of the so-called bill
of materials (BOM), indicating exactly what is needed as input to produce one kilo-
gram or metric ton of product and therefore a main means of communication between
the technical and financial people within a company.
Fig. 10.5: Components of the total depreciable capitalTDC for a chemical manufacturing plant.
(Adapted from [2]).
10.4 Estimation of capital costs
10.4.1 Capital cost components
The total depreciable captital (TDC) involves the total cost for the design, construction
and installation of a new plant or the revamp of an existing plant. It also includes the
associated modifications required for the preparation of the plant site. A new man-
ufacturing plant may be an addition to an existing site or a grassroots plant without
nearby auxiliary facilities. It is customary to separate the directly associated process-
ing equipment from the auxiliary facilities by an imaginary border named battery lim-
its. Utilizing this division, the total capital investment (TCI) is made up of direct costs:
(1) Inside battery limits (ISBL) investment for the plant itself.
(2) Outside battery limits (OSBL) investment auxiliary facility/site modifications and
improvements.
The TCI is also made up of indirect costs:

(3) Engineering and construction.
(4) Contingency charges.
As summarized in Figure 10.5, the ISBL investment includes the purchasing and instal-
lation cost of all the equipment items that make up the new plant. These direct plant
costs (DPC) consist of:
– Purchased equipment costs (PEC) for all major process equipment including field
fabrication and testing if necessary, such as:
– vessels, reactors, columns
– furnaces, heat exchangers, coolers
– pumps, compressors, motors, fans, turbines
– filters, centrifuges, driers
– Installation costs to integrate all equipment items into an operating plant, such as:
– foundations, structures, roads, piling, process buildings, sewers
– piping, valves, insulation, paint
– wiring, instrumentation
– solvents, catalysts
– Installation labor and supervision, including construction costs such as con-
struction equipment rental, temporary construction (rigging, trailers, etc.),
temporary water and power, construction workshops
It is extremely important to define the ISBL scope of a project carefully in the early
stages of a project, as other project costs are often estimated from the ISBL cost. With
a poorly defined ISBL scope, the overall project economics can be badly miscalculated.
After discussing the other capital cost components the following sections of this chap-
ter provide several methods, from very global to highly detailed, for estimating ISBL
costs.
OSBL investment includes the costs of all additions, improvements and modifica-
tions that must be made to the site infrastructure to accommodate adding a new plant
or revamp an existing plant. Figure 10.6 illustrates typical offsite auxiliary facilities
that are associated to the construction of a grassroots plant. Depending on the extent
of needed/available offsite facilities, they can be a significant contribution to the TDC.
Typical offsite investments include:
– Power generation plants, turbine engines, standby generators, electric main sub-
stations, transformers, switchgear and powerlines.
– Boilers, steam lines, condensate lines, boiler feed water treatment plant and sup-
ply pumps.
– Cooling towers, circulation pumps, cooling water lines and cooling water treat-
ment.
– Water supply pipes, demineralization plant, waste water treatment, site drainage
and sewers.
– Air separation plants (to provide site nitrogen for inert gas) and nitrogen lines.
– Dryers and blowers for instrument air and instrument air lines.
– Pipe bridges, feed and product pipelines, tanker farms, loading facilities, silos,
conveyors, docks, warehouses, railroads and lift trucks.
– Laboratories, offices, canteens, changing rooms, central control rooms and main-
tenance workshops.
– Emergency services such as firefighting equipment, fire hydrants and medical fa-
cilities.
– Site security, fencing, gatehouses and landscaping.
OSBL investment is typically estimated as a proportion of the ISBL investment in the

early stages of design. Offsite costs are usually in the range of 10% to 100% of ISBL
costs, depending on the project scope and its impact on site infrastructure. For typical
chemicals projects, offsite costs are usually between 20% and 50% of ISBL cost, and
40% is usually used as an initial estimate if no details of the site are known. For an es-
tablished site with well developed infrastructure (‘brownfield’ sites), offsite costs will
generally be lower. On the other hand, if the site infrastructure is in need of repair or
upgrading to meet new regulations, or if the plant is built on a completely new site (a
‘greenfield’ site), then offsite costs will be higher. Table 10.3 provides some guidelines
to make approximate estimates of OSBL investment as a function of process complex-
ity and site conditions.
The engineering and construction costs comprise the costs of detailed design of the
plant and other engineering services required to carry out the project:
– Detailed design engineering of process equipment, piping systems, control sys-
tems and off sites, plant layout, drafting, cost engineering, scale models, and civil
engineering.
Fig. 10.6: Layout including auxiliary facilities (OSBL) of a chemical manufacturing plant.
Tab. 10.3: Approximate OSBL investment as percentage of ISBL investment [1].
Process complexity Existing site New site

Typical bulk chemical 30% 40%
Specialty chemical 30% 50%
Solids handling process 40% 100%
– Procurement of main plant items and bulks.

– Construction supervision and services.
– Administrative charges, including engineering supervision, project management,
expediting, inspection, travel and living expenses and home office overheads.
– Contractor’s profit.
In most cases, operating companies bring in one or more of the major engineering
contracting firms to carry out all these activities. Initially engineering and construc-
tion costs can be estimated as 30% of the total plant costs (ISBL plus OSBL) for smaller
projects and 10% of the total plant costs for larger projects.
As all cost estimates are uncertain, contingency charges are added into the project
budget to allow for variation from the cost estimate. In addition, contingency costs
also help to cover minor project scope changes, price changes (e.g., steel, copper,
catalyst, etc.), currency fluctuations and other unexpected problems. A contingency
charge can be thought of as an additional fee charged by the contractor to reduce the
likelihood of losing money on a fixed price bid. A minimum contingency charge of
10% of ISBL plus OSBL cost should be used on all projects. For uncertain technologies
higher contingency charges (up to 50%) can be used.
As illustrated by Figure 10.7, the DPC, OSBL investment, engineering and construc-
tion and contingency make up the TDC. The cost of land, typically about 1–3% of the
Fig. 10.7: Connection between capital investment components.

TDC, is not included because its value usually does not decrease with time. Addition-
ally, capital for plant startup/modification expenditures (materials, equipment, main-
tenance and repairs) is commonly required before the constructed plant can operate
at maximum design conditions. It should be part of any capital estimate because it is
essential to the success of the venture. Normally an allowance of 8–10% of the TDC
is satisfactory, which is commonly included as a one time only expenditure during
the first year of plant operation and therefore also not included. Adding the cost of
land and plant startup yields the TPI. As already discussed before, the owner needs
to invest some capital in maintaining plant operations, in addition to the fixed capital
investment that was used to design and construct the plant. The capital that is tied
up in maintaining inventories of feeds, products, and spare parts, together with cash
on hand and the difference between money owed by customers (accounts receivable)
and money owed to suppliers (accounts payable), is termed the WC of the plant. By
adding the WC, the final TCI of a project is obtained.
10.4.2 Evolution and purpose of capital cost estimates
The accuracy of an estimate depends on the available amount of design detail, the
cost data accuracy, and the time spent on estimate preparation. As illustrated by Fig-
ure 10.8, the available information only justifies approximate estimates in the early
project stages and the accuracy of the estimate improves as the project progresses.
Fig. 10.8: Evolution of investment estimates.

Commonly, according to their accuracy and purpose, the following capital cost esti-
mate classification is used:
– Order of magnitude estimates, ±50%, usually applied for initial feasibility/screen-
ing studies, based on the costs of similar processes or functional units and requir-
ing minimal to no design information.
– Study estimates, ±35%, used to make coarse choices between design alternatives
based on limited cost data and design detail.
– Preliminary estimates, ±20%, used to make detailed choices between design alter-
natives and first optimizations based on more detailed design and more accurate
cost data, including vendor quotes for critical equipment.
– Definitive estimates, ±10%, used to authorize funds to bring the design to the point
where a more detailed and thus accurate estimate can be made. With experience
and cost data from similar projects, acceptable accuracy estimates can be based
on the project flowsheet, a rough P&I diagram and the approximate sizes of the
major equipment items.
– Detailed estimates, accuracy ±5%, which are used for project cost control and es-
timates for fixed price contracts. This quality estimates require a completed front-
end engineering design (FEED), including a (near) complete process design, firm
equipment quotes, and a detailed construction cost breakdown and estimation.
To go beyond the preliminary estimate accuracy requires a fairly detailed design of

the plant. The cost of estimate preparation thus becomes the cost of process design
and sizing the main equipment items. To improve the accuracy further the contractor
needs to establish the plot plan and plant layout to make accurate estimates of actual
installation costs (piping, wiring, structural steel) that will be needed. The cost of esti-
mate preparation can thus range from 0.1% of the total project cost for ±35% accuracy
up to approximately 5% for a detailed ±5% estimate.
10.4.3 Order of magnitude estimates
During early design stages or preliminary marketing studies it is often desired to make
a quick (order of magnitude) capital cost estimate without having to complete a plant
design. Several short cut methods using existing plant data or step counting have been
developed to enable total plant investment estimates within ±50% accuracy for pre-
liminary studies. These methods are also suitable to perform a rough check on more
detailed estimates based on process equipment costs later in the process design.
10.4.3.1 Existing plant data

The quickest way to make an order of magnitude estimate of plant costs is to scale it
from the known cost of an earlier plant that used the same technology or from pub-
Tab. 10.4: Plant capacity and capital investment for some chemicals (year=2000) [4].
Product Process Reference capacity C REF n

(103 tons/yr) (106 $)
Acetic acid Methanol + CO 10 8 0.68
Acetone Propylene 100 33 0.45
Ammonia Natural gas 100 29 0.53
Ethanol Ethylene hydration
Ethylene Refinery gas 50 16 0.83
Ethylene oxide Ethylene oxidation 50 59 0.78
Methanol Natural gas 60 15 0.60
Nitric acid Ammonia 100 8 0.60
Sulfuric acid Contact, sulfur 100 4 0.65
Urea Ammonia and CO2 60 10 0.70
lished data. This requires no design information other than the production rate, year
the plant was built and plant location. The capital cost of a plant can be related to
capacity by the following equation:
Desired Capacity n
CTDC = CREF ( ) (10.15)
Reference Capacity
Here, CTDC = total depreciable capital of the plant with the desired capacity and
CREF = total depreciable capital for the reference capacity. As illustrated by Table 10.4,
the capacity exponent n can range from 0.4 for small scale processes (pharmaceuticals
and specialty chemicals, for example) up to 0.9 for processes that use a lot of mechan-
ical work or gas compression (solids or gas processing). As the exponent n is always
less than 1.0, constructing larger plants tends to cost less per unit production capacity,
which is known as economy of scale. Averaged across the whole chemical industry, n
is about 0.6, and hence Equation 10.15 is commonly referred to as the ‘six-tenths rule’
and can be used to get a rough estimate of the TDC.
As all cost estimating methods use historical data, and material/labor costs are
subject to inflation, the old cost data have to be updated for estimating at the de-
sign stage, and to forecast the future construction cost of the plant. This is commonly
done by using published composite cost indices published for various industries in the
trade journals to relate present to past costs. These are weighted average indices com-
bining data for equipment categories, construction labor, buildings plus engineering
and supervision in proportions considered typical for the particular industry. For the
chemical process industry, the Chemical Engineering Plant Cost Index (CEPCI) is the
most widely used composite index. It is based on price developments in the United
States process industry, and published monthly in the journal, Chemical Engineer-
ing. The estimated TDC can now be updated to the desired year by multiplying with
the CEPCI index ratio between the desired and reference year:
Desired capacity n CEPCI

CTDC = CREF ( ) ( ) (10.16)
Reference capacity CEPCIREF
Figure 10.9 depicts the development of the annual CEPCI indices since 1957–1959, when
the plant costs were relatively stable and the index defined as 100.
Overall index Construction labor Engineering supervision

Equipment Buildings
Fig. 10.9: Chemical Engineering Plant Cost IndexCEPCI indices between 1957–1959.
Most plant and equipment cost data are available on a U.S. Gulf Coast (USGC) or North
West Europe (NWE) basis, as these have historically been the main centers of the chem-
ical industry. However, for other locations the cost of building a plant will depend on
local costs for labor, fabrication, construction, shipping, transporting and importing
(taxes) as well as currency exchange rates. In early stage cost estimates these differ-
ences are typically captured by using a location factor (LF) relative to the USGC or NWE
basis to correct the TDC:
Desired Capacity n CEPCI
CTDC (location) = LF × CTDC = LF × CREF ( ) ( ) (10.17)
Reference Capacity CEPCIREF
LF for international locations are a strong function of currency exchange rates, and
hence fluctuate in time. Table 10.5 provides example LF in U.S. dollars based on 2003
data.
Tab. 10.5: Location factors (LF) in U.S. dollars based on 2003 data [1].
Country Region LF
U.S. Gulf Coast 1.00
Canada 1.00
Western Europe France 1.13
Germany 1.11
Netherlands 1.19
United Kingdom 1.02
Mexico 1.03
Brazil 1.14
China imported 1.12
indigenous 0.61
S.E. Asia 1.12
Japan 1.26
India 1.02
Middle East 1.07
Russia 1.53
Australia 1.21
10.4.3.2 Step counting methods

In case no process cost data are available, an order of magnitude estimate can be
made by adding the contributions of the different plant sections, which are called func-
tional units. This approach has been developed by Bridgewater, correlating plant costs
against the number of processing steps for liquids and solids processing plants [1.7]:
Q 0.675
Q ≥ 60,000 : CDPC = 4320 N ( ) (10.18)
s
Q 0.30
Q < 60,000 : CDPC = 380,000 N ( ) (10.19)
s
CDPC represents the direct (ISBL) plant costs in U.S. dollars (U.S.Gulf Coast, Jan.2010
basis, CEPCI = 533), Q = plant capacity (metric tons/year), s = reactor ‘conversion’
(kg desired reactor product/kg total reactor input) and N = number of functional units.
A functional unit is defined as:
– A significant process step, such as reaction section, separation section, or major
unit operation such as main stream compressors.
– A multistream operation that is taken as one functional unit.
– Individual heat exchange and pumping are normally ignored as they are part of
a functional unit unless substantial cost are involved (i.e., main stream compres-
sors, refrigeration systems or process furnaces).
– Standard ‘in process’ and raw material storage are ignored because they tend to be
a constant process function, unless mechanical handling for solids are involved.
Tab. 10.6: Hill method factors [1].
FPI Process type

1.85 Solids handling
2.00 Solids-fluids handling
2.15 Fluids handling
FC Construction factor
0.15 Outdoor construction
0.40 Mixed indoor and outdoor construction
0.80 Indoor construction
FOSBL OSBL factor
0.10 Minor addition existing facilities
0.30 Major addition existing facilities
0.80 Grass roots plant
– Large raw material, intermediate or product storages are normally treated sepa-
rately from ‘the process’ in the estimate.
– Simple ‘mechanical’ separation without moving parts is also ignored (cyclone,
gravity settler) as the cost is usually relatively insignificant.
The direct permanent investment, CDPI , is now obtained by adding the OSBL factor,
FOSBL , as given in Table 10.6, to the Bridgewater direct plant cost that already includes
engineering and construction for the ISBL part of the plant:
CDPI = (1 + FOSBL ) × CDPC (10.20)
Finally the total depreciable capital, CTDC , total permanent investment, CTPI , and the
total capital investment, CTCI , (see Figure 10.7) are obtained by including a contin-
gency of 40%, additional 10% for costs of land, royalties and plant startup and taking
15% of the TPI as WC:
CTDC = 1.40 × CDPI (10.21)

CTPI = 1.10 × CTDC (10.22)
CTCI = 1.15 × CTPI (10.23)
An alternative as presented by Seider, Seader and Lewin [3], based on the method
of Hill, which is based on the year 2006 (CEPCI = 500), a base production rate of
4536 metric tons/year, carbon steel construction and a design pressure less than seven
bar located at the U.S. Gulf Coast. As a first step, the production rate factor, FPR , is
calculated using the six-tenths rule:
main product flow rate (ton/year) 0.6

FPR = ( ) (10.24)
4536
Tab. 10.7: Material factors (FM) [1].
Material FM
Carbon steel 1.0
Aluminum 1.07
Cast steel 1.1
Stainless steel 304/316 1.3
Stainless steel 321 1.5
Hastelloy C 1.55
Monel 1.65
Inconel and nickel 1.7
Titanium clad 2.0
The FPR is used to calculate the module cost, MC, for purchasing, delivering and in-
stalling each piece of equipment needed to perform/operate a significant process step,
similar to the previously described functional units.
Pdesign (bar, if > 7 bar) 0.25
MCi = FPR FM ( ) × $160,000 (10.25)
7
Here, FM is a material factor accounting for the differences in construction materials
as given in Table 10.7 and $160,000 is the reference price for a base functional unit
with reference capacity of 4536 metric tons/year, carbon steel construction and design
pressure <7 bar located at the U.S. Gulf Coast in the year 2006.
The individual MCs are summed, updated to the current CEPCI cost index, and
multiplied by the process type factor, FPI , given in Table 10.6 to account for piping,
instrumentation, controls and indirect costs to obtain the DPC, CDPC :
CEPCI N
CDPC = FPI ( ) ∑ MCi (10.26)
500 i=1
By multiplying the DPC with the engineering and construction factor FC and OSBL fac-
tor FOSBL , both given in Table 10.6, the direct permanent investment, CDPI , is obtained:
CDPI = (1 + F C + FOSBL ) × CDPC (10.27)
Finally the total capital investment is obtained from equation (10.23), through equa-
tions (10.21) and (10.22).
10.4.4 Study and preliminary estimates (factorial methods)
The more accurate study (±35%) and preliminary (±20%) estimates are based on es-
timating the purchase cost for all the major equipment items and adding the other
costs as factor(s) of these purchased equipment costs. Clearly the accuracy of this ap-
proach will highly depend on the reached stage of design when the estimate is made,
Tab. 10.8: Examples of typical size exponents for equipment cost [4].
Equipment n
Crystallizer 0.37
Compressor 0.69
Evaporator 0.54
Heat exchanger 0.44
Reactor 0.55
Tower 0.62
Tank 0.57
and on the reliability of the used equipment costs data. When detailed equipment
specifications are available and firm vendor quotes have been obtained, more accu-
rate preliminary estimates can be obtained, but in general this approach leads to the
study estimate accuracy of ±35%.
10.4.4.1 Estimation of purchased equipment cost

The major effort in estimating the depreciable capital cost is estimating the cost of
equipment. There are three main sources of equipment cost data: current vendor quo-
tations, past vendor quotations and literature estimates. Vendor quotations are the
most accurate, but the effort required to prepare detailed specifications and quota-
tions are not usually warranted in the early stages of a project. Thus, for quick esti-
mates, one has to rely on literature estimates and past quotations. Equipment costs
reported in the literature are either free on board (FOB), delivered or installed cost,
all usually given at some time in the past. Shipping and delivery cost will typically
amount about 5–15% of the purchased cost. Before adding equipment costs they must
all be on the same basis; either FOB, delivered or installed.
Usually, the literature contains equipment costs for other capacities than re-
quired. To scale the purchased equipment cost to the required capacity, we usually
assume that its cost varies to some power of its capacity:
Desired capacity n
PECBase = PECREF ( ) (10.28)
Reference capacity
In this equation, n is the size-exponent that can vary from less than 0.3 to larger than 1,
depending on the type of equipment. Some typical values are given in Table 10.8.
Often, the cost literature contains equipment cost at base conditions, being at low
temperature and pressure, carbon steel construction and a specific design. To update
for the actual equipment cost at other conditions, the base is multiplied by correction
factors:
PEC = PECBase F T FP FM (10.29)
The material factor, FM, has already been given in Table 10.7. Table 10.9 contains values
for the temperature, FT , and pressure, FP , correction factors. These factors depend on
Tab. 10.9: Pressure and temperature factors (FT) [2].
Design pressure (atm) FP Design temperature (°C) FT

0.005 1.3 −80 1.3
0.014 1.2 0 1.0
0.048 1.1 100 1.05
0.54 to 6.8 1.0 600 1.1
48 1.1 5,000 1.2
204 1.2 10,000 1.4
408 1.3
the type of equipment. Therefore, using the same correction factors for all equipment
is an approximation. For shell and tube heat exchangers these factors should be used
with care as the shell material may be different than the tube material. Also, it is good
practice to place the high pressure on the tube side to reduce material cost.
Finally the PEC need to be updated to the desired year by multiplying with the
CEPCI index ratio between the desired and reference year:
year Desired Capacity n CEPCI

PECi = PECREF FT FP FM ( ) ( ) (10.30)
Reference Capacity CEPCIREF
Equipment costs information can be found in the professional cost engineering litera-
ture, engineering textbooks and on various websites. Useful textbooks are Towler and
Sinnot [1], Seider, Seader and Lewin [3], Peters, Timmerhaus, and West [4], Woods [6],
Ulrich and Vasudevan [9], Cooper, Penney, Fair and Walas [11]. In these textbooks,
correlations given for preliminary estimates are typically of the form:
PEC = c1 + c2 S n (10.31)
Here, c1 and c2 are cost constants, S a size/capacity parameter and n the scale ex-
ponent for that type of equipment. A useful web based costing tool is available at
www.matche.com. Other sources to obtain first estimates are resale websites such as
www.equipnet.com and www.ippe.com, or websites such as www.alibaba.com. Alter-
natively, when available, cost estimating programs like Aspen Process Economic An-
alyzer can be used. It is highly recommended to thoroughly crosscheck the estimates
obtained from multiple sources to arrive at a reasonable reliability.
10.4.4.2 Overall installation (Lang) factors

To achieve study estimate accuracy (±35%), it is typically sufficient to consider overall
installation factors that relate the TDC (without WC) and TCI (including WC) of a plant
Tab. 10.10: Breakdown (percent of purchased equipment cost) of Lang factors [4].
Process type
Item Fluid Fluid-solid Solid
Purchase cost major equipment 100 100 100

– Installation 47 39 45
– Instrumentation and control 36 26 18
– Piping 68 31 16
– Electrical 11 10 10
– Buildings and structures 18 29 25
– Yard improvements 10 12 15
– Service facilities 70 55 40
Total direct plant cost 360 302 269
– Engineering and supervision 33 32 33
– Construction expenses 41 34 39
Total and indirect plant cost 434 368 341
– Contractor’s fee and legal expenses 26 23 21
– Contingency 44 37 35
TDC 504 428 397
FL(TDC) 5.0 4.3 4.0
– WC and land 89 75 70
TCI 593 503 467
FL(TCI) 6.0 5.0 4.7
to the total purchased equipment cost by the equations:
CTDC = FL(TDC) ∑ PECi (10.32)

i
CTCI = FL(TCI) ∑ PECi (10.33)

i
In these equations, ∑ PECi is the total delivered cost of all the major equipment items
and FL the installation factor, widely known as the Lang factor. These factors depend
on the type of plant being considered and include process equipment, installation, in-
strumentation and control, piping, electrical, engineering, etc. The values and break-
down of the Lang factors are given in Table 10.10.
10.4.4.3 Individual factors

Further improvement in accuracy to achieve a preliminary estimate (±20%) can be
achieved by using individual installation factors for different types of equipment. This
requires the availability of much more detailed design information and is therefore
only justified if reliable cost data are available and the design has been taken to the
point where all the cost items can be identified and included. In Guthrie’s [8] detailed
cost estimating method, the costs for installation, piping, and instrumentation are
taken separately for each piece of equipment. As an illustration, some of his factors are
Tab. 10.11: Guthrie’s bare-module factors [8].
Equipment Bare module factor

Shell and tube heat exchanger 2.19
Vertical pressure vessel 4.16
Horizontal pressure vessel 3.05
Gas compressor 2.15
Centrifuges 2.03
Crystallizers 2.06
Evaporators 2.45
given in Table 10.11. For more information about these individual factor methods, the
reader is referred to Cooper [5], Gerrard [7], Guthrie [8] and Ulrich and Vasudevan [9],
as this is beyond the scope of this book.
Nomenclature
BOM Bill of materials
CF Cash flow valuta/year
CCF Cumulative cash flow valuta
CDCF Cumulative discounted cash flow valuta
CEPCI Chemical engineering plant cost index –
CLand Cost of land valuta
CDPC Direct plant cost valuta
CDPI Direct permanent investment valuta
CREF Total depreciable capital reference capacity valuta
CTCI Total capital investment valuta
CTDC Total depreciable capital valuta
CTPC Total plant cost valuta
CTPI Total permanent investment valuta
c1 Cost constant valuta
c2 Cost constant valuta
D Depreciation valuta/year
DCF Discounted cash flow valuta/year
DPC Direct plant costs valuta
EP Economic potential valuta/year
FEED Front-end engineering design
FC Annual fixed costs valuta/year
FC Construction factor –
FM Material factor –
FL Lang factor –
FOSBL OSBL factor –
FP Pressure factor –
FPI Process type factor –
FPR Production rate factor –
FT Temperature factor –
FOB Free on board –
GP Gross profit valuta/year
IRR Internal rate of return –
ISBL Inside battery limits –
LF Location factor –
MC Module cost valuta
n Year –
n Capacity exponent, size exponent –
N Number of years –
N Number of functional units –
NP Net profit valuta/year
NPV Net present value valuta
OI Operating income valuta/year
OSBL Outside battery limits –
P Pressure bar
PBT Pay back time years
PEC Purchases equipment costs valuta
PFD Process flow diagram –
Q Plant capacity metric tons/year
r Discount rate –
RMC Annual raw material costs valuta/year
ROI Return on investment –
s reactor ‘conversion’ –
S Sales valuta/year
S Size/capacity parameter m3 , m2 , ton/hr
t Taxaction rate –
TCI Total capital investment valuta
TDC Total deprecial capital valuta
TPI Total permanent investment valuta
VC Annual variable costs valuta/year
WC Working capital valuta
WACC Weighted average cos of capital –
Φ Fraction of total depreciable capital (TDC) –

[1] G. Towler, R. Sinnot. Chemical Engineering Design: Principles, Practice and Economics of Plant
and Process Design, 2nd edn, Elsevier, Amsterdam, 2013.
[2] H. Silla. Chemical Process Engineering, Design and Economics, Marcel Dekker, New York,
2003.
[3] W.D. Seider, J.D. Seader, D.R. Lewin, S. Widagdo. Product and Process Design Principles, 3rd
edn, John Wiley & Sons, New York, 2010.
[4] M.S. Peters, K.D. Timmerhaus, R.E. West. Plant Design and Economics for Chemical Engineers,
5th edn, McGraw-Hill, New York, 2004.
[5] J.R. Cooper. Process Engineering Economics, Marcel Dekker, New York, 2003.
[6] D.R. Woods. Rules of Thumb in Engineering Practice, Wiley-VCH, Weinheim, 2007.
[7] A.M. Gerrard. Guide to Capital Cost Estimation, 4th edn, IChemE, London, 2000.
[8] K.M. Guthrie. Process Plant Estimating, Evaluation and Control, Craftsman, Solano Beach,
California, 1974.
[9] G.D. Ulrich, P.T. Vasudevan. Chemical Engineering Process Design and Economics: A Practical
Guide, Process Publishing, 2004.
[10] Dutch Association Cost Engineers. Price Booklet, 32nd edn, DACE, Nijkerk, 2017.
[11] J.R. Cooper, W.R. Penney, J.R. Fair, S.M. Walas. Chemical Process Equipment, 3rd edn, Butter-
worth-Heinemann, Oxford, 2012.
11 Evaluating for safety and health
11.1 Introduction
Safety concerns human safety for events of short durations, such as explosions. Health
concern lifelong human health. The latter is concerned with workers exposed for work-
ing hours per day for their working life and for people exposed to emissions and radi-
ations their life long. This chapter treats both types of evaluation and often the time
of exposure duration is mentioned but the distinction between safety and health is in
most cases not explicitly mentioned.
Assessing Health and Safety aspects during the design of an industrial facility or
product is something that nowadays seems common practice for most major compa-
nies and engineering contractors. However, this has not always been the case. The
wisdom of systematically assessing hazards and risks and the associated methodolo-
gies that have been developed to serve this purpose only came about after various
major accidents had occurred. Some of the major accidents revealed blind spots and
gaps in certain aspects of industry safety awareness at the time. The practical lessons
learned and regulatory fallout from these accidents resulted in improving safety and
risk management in the design and operation of oil, gas and chemical installations
and the design of products [1]. Hazards and risk assessments should be performed in
a systematic way and at the appropriate time during a design project. The goal of this
chapter is to provide readers an introduction to basic safety concepts, and provide
them with the tools to get started on performing risk assessments during early design
stages. References are provided at the end of this chapter for further reading on this
subject.
11.2 Risk reduction during design
Before going into details about the risk assessment and evaluation process, one needs
to understand what is meant by ‘risk’ and ‘risk management’.
Risk is a measure of the potential for human injury, environmental damage or
economic loss in terms of both the probability of the loss, damage or injury occurring,
and the magnitude of the damage, loss or injury if it does occur [2].
risk = f ((probability of occurance) ⋅ (consequences))
Risk management is the systematic application of management policies, procedures

and practices to the tasks of analyzing, assessing, and controlling risk to protect em-
ployees, the public and the environment, as well as company assets while avoiding
business interruptions [3].
https://doi.org/10.1515/9783110467741-011
260 | 11 Evaluating for safety and health
1. Hazard
identification
Risk tolerability and ALARP 2. Consequence

assessment
3. Risk analysis and

evaluation
4. Risk reduction if
required
5. Implement in
design
ALARP: risks As Low as Reasonably Practicable
Fig. 11.1: Risk reduction steps during design.
There are various models available that describe how risks are managed during design
but the basic principle is that it is an iterative process in which hazards are identified
and the risks evaluated until deemed acceptable. The following steps, illustrated in
Figure 11.1, are general for risk reduction strategies:
(1) identify the hazards
(2) assess the consequences
(3) analyze and evaluate if the risks are acceptable
(4) decide on (more) mitigating measures based on risk evaluation
(5) implement in design
The principles applied to the risk management of process engineering can also be
applicable to product engineering. The difference is more in where the focus of the
assessment lies. When assessing the hazards and risks during product engineering,
there is a shift of focus towards the end user. The production of toys is such an exam-
ple. During the risk assessments of the process, the focus is on risk exposure to the
workers present on the facility, surrounding population, local environment, etc.
The starting principle of a design should be to approach it from an inherent safety
principle. This will be discussed in the next section.
11.3 Inherent safer design principle
The concept of inherent safer design was introduced by Trevor Klotz after the Flixbor-
ough disaster and his expression sums up the basic principle: “What you don’t have
cannot leak or burn” [4]. There are four main methods for achieving inherent safer
designs:
(1) Avoidance – take measures in the design to remove hazardous materials, proce-
dures or actions completely. For example, avoiding the use of toxic, flammable or
corroding chemicals.
(2) Prevention – take measures to reduce the chance or likelihood of the hazard aris-
ing. For example, reducing the number of flanges to lower the number of potential
leak sources in a design.
(3) Control – take measures to reduce the consequences (severity) before the hazard
is realized. For example, segmenting the process in smaller sections with the help
of emergency shutdown valves to limit the amount released during a loss of hy-
drocarbon containment.
(4) Mitigation – take measures to limit the effect or spread of the incident once the
hazard has been realized such as emergency and evacuation arrangements. For
example, fire walls, deluge systems and evacuation arrangements.
The principle of inherent safety in design is that first measures should be taken to
avoid the hazard altogether and that mitigation measures should only be taken when
the other options are not reasonably practical. It is for this reason that identification of
hazards and the assessment of mitigating measures should be done early during de-
sign stages when possible. The true inherent safer design approach focusses on avoid-
ance and prevention, thus choosing different technologies or materials to avoid haz-
ards all together. This is as opposed to the safety in design approach, where additional
safety systems are introduced to control and mitigate hazards.
There are some tools and techniques available that can help compare design op-
tions from an inherent safety perspective. An example of such a tool is the Dow in-
dexes [5, 6] which, based on the properties of the chemicals intended for use as part of
the design, can rank design options based on their fire, explosion and toxic hazards.
Processes can also be compared based on the consequences during incidences. The
principle has been adopted in many codes, standards, design engineering practices
and regulations and is an important one to keep in mind during all design stages.
Having adopted the safety in design principle, and keeping this in mind during
the risk management process, the first step is to start identifying the hazards of the
proposed design.
11.4 Identification studies
The first step of systematic evaluation requires that the hazards be identified. But what
is meant by a hazard? A hazard, in the context of this chapter, is defined as an inherent
chemical or physical characteristic that has the potential to cause damage to people,
property or the environment.
This section provides an overview of some common hazard identification method-

ologies used during engineering projects. It is not intended as a complete overview of
all identification techniques or detailed treatise about the methodologies, but more
as an introduction into this field. For a detailed description of the methodologies the
reader should refer to books such as Lees’ Loss Prevention in the Process Industries [7]
and the international standard IEC/ISO 31010 Risk management – risk assessment
techniques.
Chemical reactivity hazards

The first item to evaluate in a product risk assessment is the physical and chemical
properties of the product. Following the inherent safer design principle, one should
strive to avoid or minimize the use of chemicals and/or materials that have hazardous
properties. Substitution or phasing out of chemicals that are known to be hazardous is
now becoming more apart of international legislation. EU Regulation No. 1907/2006,
Registration, Evaluation, Authorization and Restriction of Chemicals (REACH), came
into force in 2007. The goal of REACH is to improve the protection of human health
and the environment by identifying the intrinsic properties of chemical substances
and phasing out chemicals of very high concern. When developing a product in the
EU or for use in the EU, this regulation is important to be aware of and its requirements
should be taken into consideration from the feasibility stage.
But how does one discover what the hazardous properties of a product are? There
are various sources of information available that contain data with respect to toxic,
fire and explosion hazards of chemicals such as but not limited to:
– material safety datasheets (MSDS)
– information contained in standards such as NFPA 49, 325, 432, 491
– Bretherick’s Handbook of Reactive Chemical Hazards [8]
– incident reports
Having obtained information from a literature survey, the next step is to look at the
chemical and thermodynamic properties since they can provide information on po-
tential hazards. Properties that can provide information are:
– thermodynamic properties
– chemical composition, structure and bonds
Thermodynamic properties
The heat of reaction gives an indication of the potential temperature rise in a mixture.
The higher the heat of reaction, the higher the potential for energetic, uncontrolled
reactions. Using the heat of reaction and the heat capacity of the reaction mixture,
one can calculate the adiabatic temperature rise, the maximum temperature rise that
can theoretically occur without heat loss:

∆Hr
∆Tad =
Cp
Where:
∆Tad = adiabatic temperature rise (K)
∆Hr = heat of reaction (kJ/kg)
Cp = heat capacity of reaction mixture (kJ/(kg K))
The sum of the maximum process temperature and adiabatic temperature rise can
give important information about potential hazards. When looking at the result, one
should wonder if this temperature is below a temperature where:
– additional chemistry (such as decomposition) can occur
– phase transition (such as boiling or gas generation) can occur
– overpressurization from increased vaporization can occur
Another important thermodynamic property is the heat of formation. Compounds hav-

ing a high endothermic heat of formation (positive heat of formation) contain a lot of
energy and tend to be self-reactive. Correspondingly, they have a high exothermic heat
of reaction.
Chemical composition, structure and bonds

Chemical structure and bonds provide a second clue. Be suspicious of [9]:
– compounds that have high bond strains
– carbon–carbon double bonds not in benzene rings (e.g., ethylene, styrene)
– carbon–carbon triple bonds (e.g., acetylene)
– nitrogen containing compounds (NO2 groups, adjacent N atoms, etc.)
– oxygen–oxygen bonds (peroxides, hydroperoxides, ozonizes)
– ring compounds with only three or four atoms (e.g., ethylene oxide)
– metal- and halogen-containing complexes (metal fulminates; halites, halates;
etc.)
Some chemicals have the tendency to produce strong oxidation reactions which can
be explosive in nature. The oxygen required for this oxidation can be stored in the com-
pound itself or be supplied from outside during the reaction. One estimating parame-
ter for the potential hazard and instability of substances or reaction mixtures contain-
ing oxygen is the oxygen balance (OB). The oxygen balance is the amount of oxygen,
expressed as weight percent, liberated because of complete conversion of the mate-
rial to CO2 , H2 O, SO2 , Al2 O3 , N2 , and other relatively simple oxidized molecules [10].
If the oxygen balance (OB) is positive, there is an excess of amount of oxygen avail-
able in the compound for a conversion to simple components. If the OB is zero, there
is just enough oxygen available for the conversion to simple components. If the OB is
negative, there is insufficient oxygen for a complete oxidation of the compound.
The following decomposition reaction forms the basis of the OB:

Cx Hy Oz Nn → xCO2 + 2y H2 O + 2n N2 + (z − 2x − 2y ) 12 O2
In case of perfect oxidation, no excess oxygen is left after the reaction and the carbon
and hydrogen atoms have fully oxidized to carbon dioxide and water. The heat of re-
action tends to reach a maximum in this case. The parameter z − 2x − (y/2) is then
equal to zero under such ideal conditions. The oxygen balance is defined as:
1600 [z − 2x − 2y ]
OB =
MW
MW is the molecular weight of the compound. The closer the OB is to zero, the higher
the hazards since the explosion sensitivity and severity will tend to reach a maximum
at an OB of zero. The OB value can thus be used to assess the degree of hazard [10] as
follows:
Hazard potential Value of OB
Low −160 < OB < −240

Medium −240 < OB < −120
80 < OB < 160
High −120 < OB < 80
It should be stressed that the equation above is not applicable to molecules containing
other atoms and that the OB only serves as an indicator. The way oxygen is bonded
should be reviewed in relation to the OB. For an explosive such as nitroglycerine, the
oxygen balance is +3.5, which corresponds well to the high hazard potential of the
table above. Acetic acid on the other hand has an OB of 106.7 but is not an explosion
hazard since the oxygen present cannot be used efficiently for the oxidation of carbon.
When chemical and/or thermodynamic properties are not known, then the next
step in the hazards assessment is to perform tests. Guidance for performing such tests
is provided in [10, 11] and [12].
HAZard IDentification (HAZID) and ENVironmental Identification (ENVID)

Plot plans preliminary layouts are developed early during the design, before process
and instrumentation diagrams have fully matured. A relatively simple method to iden-
tify hazards during this stage is by using qualitative techniques that use key words to
think of causes, consequences and safeguards in a design. One of the more common
techniques that can be used during the feasibility stage up to detailed stage design is
the HAZID study. The HAZID is a study technique for early identification of potential
hazards and threats.
The major benefit of a HAZID is that the early identification and assessment of the
critical HSE hazards provides essential input to project development decisions. This
will lead to safer and more cost-effective design options being adopted with a mini-
mum cost of change penalty. The deliverable of the HAZID study is a set of detailed
minutes, showing the summary of the discussion with action recommendations di-
rected to the separate involved parties who were present in the meeting.
The HAZID study has been developed specifically to reflect the importance of HSE
issues on the fundamental decisions that are made at the inception of all development
projects. The HAZID study is usually the first opportunity to assemble experienced
operational, engineering and HSE staff together to address, in a short time frame, the
issues surrounding the project.
During a HAZID study review session, a guideword list is used and the session
itself facilitated by a chairman and supported by a scribe when required. An example
guideword list and part of a HAZID table are given in Table 11.1 and 11.2 for a HAZID
performed for a conceptual design of a new offshore facility. The guidewords should
be adapted to reflect the specific circumstances or features of the design if possible.
Tab. 11.1: Example guidewords for a HAZID for an offshore installation.
HAZID category Risk element, system or event

Process event Unignited gas release
Jet fire
Flash fire
Explosion
Blowout
Unignited liquid release
Pool fire
Chemical/toxic hazard
Nonprocess fire
Vent/exhaust/air intake
Hazardous area classification
External event Ship collision
Dropped load
Helicopter impact
Environmental hazards
Safety system Escape routes
Muster station
Lifeboat
Life raft
Active firefighting system
Firewalls/blast wall/shielding/PFP
Fire and gas detection
Area alarm systems
ESD/EBD
HVAC
HAZID category Risk element, system or event

Working environment, Access to maintainable equipment
operations & maintenance Limited workspace
Interaction with rig
LSA/Radiation
Noise
Handling of chemicals/explosives
Mechanical handling
Control room operations
Well intervention operations
Helicopter operations
Crane operations
Supply vessel operations
Other, e.g., Drawing comments
Environmental aspects (air, soil, landscape,
noise, energy, radiation, water, waste, light)
Tab. 11.2: Example table for a HAZID.
Risk element/ Cause/ Conse- Evaluation/ No. Action Action Due date
system or event guideword quences safeguards required party
Process event
Unignited HC
release
Jet fire
Etc.
For each guideword, expanders are provided to help the brainstorm session on identi-
fying possible hazards. The HAZID team brainstorms to find all potential hazards and
its causes.
Each hazard is examined to see what:
– consequences may arise from the hazard
– controls are in place
– what action recommendations, if the team decides what controls are not sufficient
or can be improved, action recommendations are formulated and assigned to the
responsible party.
An ENVID uses the same principle as the HAZID, except that the purpose is to identify
environmental hazards and thus the keywords are different. Below is an example list
of keywords that can be used during an ENVID:
ENVID category Guideword
Impact – air quality Greenhouse gas emissions

VOCs and fugitive emissions
NOx and SO
Ozone depleting substances
Particulate matter
Toxic metals
Odorous compounds
Impact – waste Waste production
Hazardous waste
Domestic waste
Impact – resource use Land use
Land quality
Loss of habitat
Degradation of land quality
Alteration of ecological/landscape processes
Impact – oil and chemical spills Chemical use and storage
Hydrocarbon use and storage
Bunkering and equipment failure
Human error and procedures
Pipework, valves and seals
Bunding and drainage
Impact – chemical use and discharge Chemical use and storage
Hydrocarbon use and storage
Bunkering and equipment failure
Human error and procedures
Pipework, valves and seals
Bunding and drainage
Impact – aqueous discharges Drains
Hydrocarbon and chemical contamination
Sewage discharges
Storm water
Hydro test water
Slurry
Impact – wildlife disturbance Noise/vibration and light pollution
Air pollution
Water pollution
Conservation of protected areas
Conservation of protected species
Ecosystem effects/biodiversity
Displacement/loss of access
It is important that the actions are monitored after the session. A common pitfall is that
action closeout is delayed due to demanding engineering schedules and not having
assigned a person with the proper authority to follow up and expedite the actions
with engineers. Agreeing at the start of the project on whom this responsibility will
fall will help prevent costly rework during later stages of the design.
The HAZID & ENVID methodology is a tool that can be applied to process and
product engineering assessments and is easy to understand.
HAZOP
H.G. Lawley of Imperial Chemical Industries (ICI) developed the HAZOP methodology
in the seventies, to identify causes and consequences of disturbances in process instal-
lations. HAZOP stands for Hazard and operability analysis and is a structured hazard
identification tool using a multidisciplined team. The methodology differs from the
HAZID in the sense that a HAZOP focusses on hazards that originate from the process
itself (inside out approach), while a HAZID assesses a broader range of potential haz-
ards from within but also from outside the process (outside in approach). The HAZOP
is accepted as the main technique for the identification of process hazards in the de-
sign and operation of a facility. A detailed description of the HAZOP methodology is
available in the IEC code 61882 [13].
The HAZOP process is the systematic application of combinations of parameters
(e.g., flow, pressure, temperature) and guide words (e.g., no, more, less) to produce de-
viations (no flow, less pressure) from the design intent or intended operational mode
of the installation. The purpose of the HAZOP study is to:
– Review the design and consider whether any of the conditions which may occur,
from operation, malfunction or maloperation, cause a hazard to people or cause
damage to plant and equipment.
– Review whether the precautions incorporated in the design are sufficient to either
prevent the hazard occurring or reduce any consequence to an acceptable level.
– Consider whether the plant remains safely operable, goes to a safe hold state, par-
tially shuts down, or shuts down under the circumstances in the above given ex-
amples.
The deliverable of the HAZOP study is a set of detailed minutes, presented on work-
sheets for each deviation. These show the summary of the discussion with action rec-
ommendations directed to the separate involved parties which were present in the
meeting.
A chairman will guide the meeting through the HAZOP and can be assisted by a
scribe if required. Possible deviations/deficiencies from design will be addressed and
consequences are identified during the session.
The installation being studied is divided in logical items (equipment and piping)
that can be studied according the structured brainstorm method of the HAZOP study.
Tab. 11.3: Example parameters and guidewords for a HAZOP.
Parameters Guidewords
Flow Maintenance No –
Pressure Utilities More/high
Temperature Startup Less/low
Level Shutdown Reverse
Composition/phase General As well as
Operations Other
These items are called nodes and are selected by the HAZOP chairman. HAZOP nodes
are usually divided by phase flow (liquid, gas) and isolatable sections (e.g., between
emergency shut down valves) or installation parts. The node or installation part is
reviewed by the application of:
– parameters (e.g., flow, pressure, temperature, level), in combination with
– guidewords (e.g., no, more, less, reverse, other than)
The guideword/parameter combination is known as a deviation.

Example parameters and guidewords are available in Table 11.3.
The HAZOP team brainstorms to find all potential causes of the deviation. All po-
tential causes are discussed. Each cause is separately examined to see what:
– consequences may arise with a certain deviation
– safeguards are in place (i.e., design criteria, indications, alarms, procedures, etc.)
– the team decides about whether the safeguards are not sufficient or can be im-
proved, action recommendations are formulated and assigned to the responsible
party
An example of a HAZOP minute is provided in Table 11.4.

HAZOPs are used primarily in process risk assessment and are less commonly
used in product hazard identifications. The failure mode effect and analysis (FMEA) is
a technique that resembles the HAZOP in some aspects and is much more commonly
used for product hazards assessments.
Tab. 11.4: Example of a HAZOP minute.

Action
Devia- Guide- Cause Conse- Safeguard Action Action re- Due

Node
tion no. word quence sponsible date

no.
2 50 Elevated Failure of Exothermic Bursting 12 Ensure Jane Doe 1st

tempera- heater runaway, disc on reac- venting of (Process of De-
ture control over pres- tor prevents bursting disc engineer) cember
causing too surization over pres- is to safe 2017
elevated of reactor surization location
temperature
11.5 FMEA
FMEA is a well-known hazard identification technique. This method reviews failure

on a component level and assesses the effect on the system as a whole. The review
results in recommendations to improve the product or system design with respect to
reliability and safety. Originally developed by the US military after the Second World
War, the first formal application of the FMEA was in the aerospace industry in the
1960s for the NASA Apollo missions. It is a technique that is, nowadays, used in the
product manufacturing and services industries. Like the other hazard identification
tools discussed previously in this Chapter, the FMEA is performed as a team session
facilitated by a chairman and sometimes supported by a scribe.
The objectives of a FMEA include [7]:
(1) Identification of each failure mode, the sequence of events associated with it and
of its causes and effects.
(2) Classification of each failure mode by relevant characteristics, including de-
tectability, diagonality, testability, item replaceability, compensating and operat-
ing provisions.
(3) Formulate recommendation for improvement if required.
One of the first steps during the preparation of a FMEA is to identify the systems
boundaries. A FMEA can be performed on a system level, down to individual com-
ponent level. For example, one could look at the failure of an ESD valve as a whole or
break this system down to individual components (e.g., valve body, actuator, solenoid)
and review failure of the individual components. The second step is choosing the
appropriate failure or error modes. Failure modes for safety systems are provided in
sources such as ISO 14224 [14]. During the team session, the failure causes, local and
systems effects, detection, etc., will be identified by the team members.
The advantage of the FMEA is that it is a relatively easy method to apply and that
it is suitable for hazard assessment on an equipment level. The FMEA approach also
has a well deserved reputation for efficiently analyzing the hazards associated with
electronic and computer systems, whereas the HAZOP approach may not work as well
for these types of systems [15]. The disadvantage of the method is that, depending
on the system boundaries and level of review, the study can become a lot of work to
perform. The methodology is also less suitable for systems where complex logic is
required to describe system failure [7]. Table 11.5 provides an example of an FMEA
table.
Tab. 11.5: Example list of FMEA table.
System Component Failure Failure cause Effect on Effect on Detection Compensating Actions Action Remarks
(sub system) or error other entire system method provision party
mode components
Hydraulic External Leakage of Hydrocarbon Loss pf Large leakage Shutdown None None None
actuated ESD leakage valve body release, production detected by upstream
Valve XXX process potential for Inability to pressure drop production
Production
export
medium fire and fully isolate PZTXXX Gas facilities to
explosions system during detection at stop flow from
ESD ESD deck ESD valve
External Leakage of Hydraulic spill Loss of Hydraulic Platform will None None None
leakage valve body on ESD deck, production pressure shut down on
utility loss of detection loss of
medium hydraulic PZTXXX hydraulic
pressure, pressure
valve closes
Failure Jammed gate None Failure to Limit Shutdown None None None
to close valve identified close export, switches, upstream and
on de- in case of timeout downstream
mand ESD, possible alarms facilities
escalation of Yearly
fires on functional
platform
tests
11.5 FMEA | 271
11.6 Dow index methods
The Dow Fire and Explosion Index (Dow F&E) [5] and Chemical Exposure Index (Dow
CEI) [6] are internationally used ranking systems to evaluate the hazard category of a
process plant and expected losses due to potential fires and explosions or chemical
exposure. The Dow F&E uses the NFPA material factor which is a measure of the haz-
ards of the substance which falls between 0 and 40 in a matrix of instability/reactivity
and flammability [16]. The material factor is then modified by penalties and credit fac-
tors which depend on the operating conditions. If the results of the assessment are an
index value higher than 100, the risk is deemed to be unacceptable and risk reducing
measures need to be applied. Other outcomes of the assessment are an estimate of the
potential area of damage, value of the equipment in the area and maximum proba-
ble property damage and business interruption. The Dow F&I is a useful tool during
early engineering stages since it helps to assess multiple process options and identi-
fies which process unit is the most hazardous. Modification or additional protection of
mitigating measures can then be considered. The Dow CEI assesses the relative acute
toxicity risks due to liquid spill, evaporation and dispersion of a vapor cloud.
11.7 Bowtie assessments
The Bowtie method is a risk assessment method that can be used to analyze and com-
municate how high risk scenarios develop. The essence of the Bowtie method consists
of plausible risk scenarios around a certain hazard, and ways in which the organiza-
tion stops those scenarios from happening. The method takes its name from the shape
of the diagram that you create, which looks like a man’s bowtie [17]. The methodology
in this paragraph is based, for the most part, on the work described in the CGE Bowtie
methodology manual [17], in which more detail is provided on creating Bowties and
how to facilitate Bowtie sessions.
The Bowtie method has several goals:
– Provide a structure to systematically analyze a hazard.
– Help decide whether the current level of control is sufficient (or, for those who are
familiar with the concept, whether risks are as low as reasonably practicable, or
ALARP).
– Help identify where and how investing resources would have the greatest impact.
– Increase risk communication and awareness.
– Provide a framework to plot/collate safety related information from audits, inci-
dents, maintenance backlogs, etc.
The advantage of using a Bowtie is that it creates an easy picture to understand and
communicate on multiple levels of the organization. A complete Bowtie diagram,
linked to the management system, is like a graphical table of contents, like a map,
11.7 Bowtie assessments | 273
showing everything an organization does to control its major risks. During engineer-
ing projects, it can be used in various stages of design and is useful in helping to
identify safety critical systems. It contains operational hardware barriers, behavioral
barriers and organizational management systems, which makes it a useful tool to
holistically look at where investing resources would have the greatest impact.
Bowties are best created during a team brainstorm session using input from var-
ious engineers and operations personnel. Constructing a Bowtie can be done by ap-
plying the following steps:
(1) Specify the hazard which will be assessed (anything that has the potential to
cause harm e.g., stored flammable hydrocarbons, toxic compounds and object at
height).
(2) Specify the top event, which is the event that occurs at the point in time when
control over the hazard is lost (e.g., loss of containment or object falling).
(3) Identify threats, such as a possible cause that can release the hazard by producing
the top event. (e.g., corrosion, overflow or breaking rope).
(4) Assess consequences after the top event has occurred, which are unwanted events
caused by the release of the hazard (e.g., fire, explosion, toxic exposure to person-
nel or fatality from dropped object falling on personnel).
(5) Identify control barriers against the threats which can prevent or reduce the like-
lihood of the top event occurring (e.g., proper material selection and proper rope
certification).
(6) Identify mitigation barriers against the consequences (e.g., fire walls, evacuation
procedures and dropped object protection frame).
(7) Identify escalating factors, which are specific conditions or circumstances which
could cause a barrier to fail or be less effective.
(8) Identify escalating factor barriers, which are barriers that can help to prevent or
manage escalating factors.
A generic overview of an example Bowtie is presented in Figure 11.2, created using the
CGE software Bowtie XP.
Unfortunately, barriers are seldom 100% effective. Escalation factors take this into
account by stating a condition that reduces the effectiveness of a barrier. The following
escalation factor categories are typically considered:
– Human factors: anything a person does to make a barrier less effective.
– Abnormal conditions: anything in the environment that causes a barrier to be put
under strain.
– Loss of critical services: if a barrier relies on an outside service, losing that service
might cause it to lose effectiveness.
When identifying escalation factors, the guideline is to keep in mind that they should
not cause a top event or consequence, need to be credible, that lessons learned from
other incidents are used and to focus on the critical barriers. Questions that a team
Hazard
Consequence
Threat
Barrier Barrier Barrier Barrier
Escalation Top event

Consequence
factor
Barrier Barrier Barrier
Escalation
Threat
factor
Barrier Barrier Barrier
Fig. 11.2: Generic Bowtie.
should ask themselves are: if there are any circumstances under which the barrier
will not work, how the barrier can be ‘destroyed’ and if the barrier has failed in the
past.
Once the escalation factors have been established, the barriers for the escalation
factors can be identified. Escalation factor barriers can be divided in the categories
of human factors (e.g., training, supervision, etc.), abnormal conditions (e.g., main-
tenance, spares, etc.) and loss of critical services (backup systems, shutdown, etc.).
Bowties can be created for process related hazards but also for product related
hazards. The challenge is to identify the proper hazard and top event. Some products
are not hazardous by themselves and one should look more towards the intended use
of the product to identify consequences.
11.8 Consequences assessment
Having identified the hazards, the next step in the risk management process is to
assess the consequences. From a process or product safety perspective, the conse-
quences of toxic exposure, fire and explosions will have to be reviewed. Depending
on the engineering stage this can be a high level assessment for a risk screening pur-
pose or a detailed one as input to a full QRA. The models discussed in this chapter
provide the reader with relatively simple means to quantify consequences which can
be used as input for high level assessments. However, modeling performed as input
for quantitative risk assessments or as input to engineering decisions such as safety
distances or specifications should consult sources where consequence assessment is
discussed in more detail such as Lees’ Loss Prevention [7], Cameron and Raman [18],
the Yellow Book [19] and Casal [20]. When performing consequence modeling, it is
very important to know what the limitations and underlying assumptions in the mod-
els are. This is to prevent using models that are not suitable for the case assessed,
leading to an underestimation or overestimation of the consequences and therefore
the risk.
Fire models
Fire consequences assessments allow the calculation of heat radiation transfer to a
target with heat radiation expressed in kW/m2 . Different models are used depending
on the type of fire. The following industrial fires are commonly considered during pro-
cess risk assessments:
– pool fires
– jet fires
– fireballs resulting from boiling expanding liquid vapor explosions (BLEVE)
– flash fires
Simplified models are presented in the following section for pool fires, jet fires and
fireballs. Flash fires are discussed only briefly. CFD models such as Fire Dynamic Sim-
ulator and KFX are used for more advanced assessments but are not discussed in this
chapter.
To provide some feeling of what the effect of different levels of exposures are, some
common tolerability levels used in the process industry are provided in Table 11.6.
Tab. 11.6: Heat radiation exposure effects [21].
Heat radiation Effects

2
35 kW/m Significant chance of fatality for people exposed instantaneously (can assume
100% lethality).
10 kW/m2 1% lethality, structural failure of equipment can start to occur at these radiation
levels and above.
6 to 9 kW/m2 Pain within approximately 10 seconds, rapid escape only is possible.
3 to 5 kW/m2 Will cause pain in 15–20 seconds and injury after 30 seconds exposure, special
protected clothing of first responders provide just enough protection at 3 kW/m2 .
1 kW/m2 Heat felt comparable to exposure to the sun during a hot summer day.
Pool fire
Pool fires result from the leakage and subsequent ignition of liquid substances. In an
industrial facility they might be caused by mechanical impact, over pressurization,
leaking flanges or valves, small bore fittings, etc. Fire models can consider the flame as
a point source, multiple point sources and solid flame or modeled dynamically using
computational fluid dynamics. The point source model is discussed in this section.
The point source radiation model models a flame as a single point radiator. The
heat flux Q at a distance x from the flame center to the observer can be calculated
by [18]:
M ⋅ ∆Hc ⋅ f ⋅ τ
Q=
4 ⋅ π ⋅ x2
Where:
Q = heat flux (kW/m2 )
M = rate of combustion (kg/s)
∆Hc = heat of combustion (kJ/kg)
f = fraction of thermal energy radiated (–)
x = distance from flame center to observer (m)
τ = atmospheric transmissivity
The above expression assumes that the position of the observer towards the flame is
such that the observer is at maximum exposure to the flame radiation.
The radiation fraction will be dependent on the type of fuel, pool diameter, flame
temperature, etc. Larger pool diameters will lead to more soot formation in some type
of fuels and thereby reduce the fraction of thermal energy radiated. The fraction is a
crucial factor and, in practice, difficult to estimate. The following expression has been
proposed to calculate the fraction of thermal energy radiated for hydrocarbon pools
with a pool diameter Dp (m) [20]:
f = 0.35 ⋅ e−0.05Dp
In practice, the value of f ranges between 0.15–0.35 for various hydrocarbon pool
sizes [18].
The atmospheric transmissivity τ accounts for the absorption of thermal radiation
by the atmosphere. An approximate relation is [18]:
τ = log 10[14.1 (RH)−0.108 ⋅ r−0.13 ]
Where:
RH = relative humidity (%)
r = distance (m)
The maximum rate of combustion M for a liquid pool fire on land can be estimated by
the burning rate mb and the pool surface area A (m2 ) [22]:
M = mb ⋅ A
Where:
M = maximum rate of combustion (kg/s)
mb = burning rate (kg/(s m2 ))
A = pool surface area (m2 )
The burning rate mb can be calculated by:

∆Hc
mb = 1 ⋅ 10−3
∆H ∗
Where:
∆H ∗ = modified heat of vaporization at the boiling point of the liquid (kJ/kg)
The modified heat of vaporization ∆H ∗ can be calculated by:

Tb
∗
∆H = ∆Hv + ∫ Cp dT
Ta
Where:
∆Hv = heat of vaporization of the liquid at the ambient temperature (kJ/kg)
Cp = heat capacity of the liquid (kJ/kg K)
Ta = ambient temperature (K)
Tb = boiling point temperature of the liquid (K)
A limitation of the point source model is that it overestimates the intensity of the ther-
mal radiation near the fire, since it does not consider the flame geometry [20]. It is
recommended that the model is only used for far field prediction at distances greater
than 5 pool diameters from the center of the fire, thus using it to perform conservative
calculations of danger to personnel as opposed to establish distances between adja-
cent equipment [20].
Jet fire
Jet fires are high momentum ignited releases. Their size and flame intensity are a func-
tion of the release rate and fuel properties. Jet fires can occur in facilities where pres-
surized flammable gasses and/or liquids are stored. Unlike pool fires which can, in
principle, be extinguished using deluge and foam systems, jet fires are difficult, and
in most cases impossible, to extinguish by active fire protection systems.
The best validated jet fire models are the Chamberlain model for vertical and an-
gled jet fires [23] and the Johnson model for horizontal jet fires [24]. Calculation of
thermal radiation effects on a target is an involved process and the reader should again
refer to sources such as Lees’ Loss prevention [7] and the Yellow Book [19] for details.
The models treated here can best be used when an approximation of the flame dimen-
sions is required or as input for a flame radiation model [25].
A simple correlation for the length L of a jet fire is [25]:
L = 18.5 ⋅ m0.41
r
Where:
L = flame length (m)
mr = mass release rate (kg/s)
A generalized formula for different fuel types is [25]:
L = 0.00326 ⋅ (mr ⋅ ∆Hc )0.478
Where:
∆Hc = heat of combustion (J/kg)
Fire balls/BLEVE’s
A boiling liquid expanding vapor explosion (BLEVE) occurs when a container holding
a superheated liquid or liquefied gas experiences a sudden loss of containment. If the
medium is flammable, a fireball can occur. The most well known disaster involving
BLEVE’s occurred in San Juanico, Mexico on 19 November 1984, in which the town
of San Juan Ixhuatepec housing 40,000 people located adjacent to a facility owned
by the PEMEX State Oil Company was almost destroyed due to a series of BLEVE’s.
Approximately 500–600 people died and between 5,000 and 7,000 suffered injuries.
The models presented here assume a BLEVE of a flammable material. The fire-
ball diameter, duration and height can be calculated using the following correlations
based on the TNO Yellow Book model [19]:
Dfb = 6.48 (⋅mf )0.325 t = 0.852 ⋅ (⋅mf )0.260 H = 0.75 ⋅ Dfb
Where:
Dfb = fireball diameter (m)
mf = mass of fuel (kg)
t = fireball duration (s)
H = height of fireball center from ground (m)
The radiation received by the individual, assuming that the fireball is in full sight of
the receiver, can be calculated using the emitted flux E of the fireball and the view
factor F, as illustrated in Figure 11.3 [18]:
I =τ⋅E⋅F
Where:
I = received flux (kW/m2 )
F = view factor = (Dfb )2 cos θ/4x2 for θ ≤ 90∘ − ϕ
E = emitted flux ≅ ε ⋅ σ ⋅ Ts4
Where:
ε =emissivity (∼ 1.0)
σ = stefan-boltzman constant (5.67 ⋅ 10−11 kW/(m2 K4 ))
Ts = flame surface temperature (K)
τ = atmospheric transmissivity (–)
Dfb
x
H φ
θ Fig. 11.3: Simplified model of fireball.
Flash fires
When a cloud of flammable gas forms over an area and finds an ignition source, the
mass between the flammability limits will burn rapidly. People in the gas cloud will
be exposed to very high heat fluxes for a short duration, but this is enough to assume
that anybody inside will suffer fatal injuries. Under some conditions, the flash fire can
transform into an explosion. This is treated in the following section.
Explosion models
The difference between a fire and an explosion is that the latter occurs in a premixed
combination of the fuel and oxidant. The power of the explosion is dependent on the
proximity of the fuel and the oxidant. Explosions common to the chemical industry
can be divided into two types: deflagrations and detonations. Deflagrations are explo-
sions in which the reaction front travels through the unreacted fuel in subsonic speeds
and the propagation method is by heat transfer [26]. A detonation is an explosion with
an extremely high, supersonic reaction speed. The propagation method is not heat
transfer but an extremely rapid and sharp compression occurring in a shock wave.
Under certain conditions, a deflagration could transition to a detonation. The defla-
gration type occurs most commonly in vapor cloud explosions and the detonation in
dense phase explosions in cases where TNT or ammonium nitrates are involved. Thus
various models have been developed over the years to model the diverse types of ex-
plosion behaviors and the most recent ones can be divided into two classes: blast curve
models and computational fluid dynamic (CFD) models. The most common models are
provided below:
– blast curve models
– TNT equivalency
– multi energy (ME)
– Baker–Strehlow–Tang (BST)
– computation fluid dynamics models
– AutoReaGas (not licensed anymore)
– FLACS
This section will discuss briefly the TNT model which, in specific cases, is still used
in the process industry and ME model which is used for modeling vapor cloud explo-
sions. The reader should refer to other sources such as Lees [7] for more information
on the other types of explosion models. Dust explosions are another important type
of explosion which are not treated in this section. The reader should again refer to
sources such as Lees [7] for further information. Damage levels from explosion effects
are also discussed by Lees.
TNT equivalency
The TNT equivalent method is based on the relation between the flammable material
and TNT factored by an explosion efficiency term. Since a lot of information is avail-
able with respect to TNT explosion characteristics and damage potential, this model
is still widely used but the user should keep in mind that it is a poor model for vapor
cloud explosions. TNT explosions are dense phase explosions producing a shockwave
from the initial blast of the explosion. Vapor cloud explosions behave differently. The
initial flame front is laminar and its speed is slow but, as it propagates into congested
areas, the flame front will become more turbulent and the peak pressure will rise. Thus
the explosion overpressure in a vapor cloud explosion is more dependent on the plant
geometry. The blast shockwave has a different pressure time profile than that of a TNT
explosion. In the near field it is higher and, in the far field, lower compared to a vapor
cloud explosion. Because of this, the TNT model underestimates overpressures in the
far field of a vapor cloud explosion. When the results of this consequence assessment
are used for building design against vapor cloud explosions, the TNT method is not a
suitable model and the ME method and Baker–Strelow–Tang or CFD models are better
suited for that purpose. The TNT method is better suited for condensed phase explo-
sions such as with ammonium nitrate. Even then, it is important to know that there is
at least an order of magnitude margin of error in the TNT model due to large scatter in
the results of the underlying experiments [7].
There is experimental evidence that if a spherical explosive charge of diameter Dc
produces a peak overpressure ∆P at a distance R from its center, as well as a positive-
phase duration t∗ and an impulse i (the integral of the pressure time history), then
a charge of diameter kDc of a similar explosive in the same atmosphere will produce
an overpressure wave with a similar form and the same peak overpressure p0 (as well
as a positive phase duration kt∗ and an impulse ki at a distance kR from the charge
center. This is called the Hopkinson or ‘cube root’ scaling law and illustrated in Fig-
ure 11.4 [20].
Typical blast curves for TNT have been obtained by military field experiments and
these curves apply the cube root scaling law to provide blast characteristics versus
scaled parameters. The first step is to calculate the equivalent TNT mass based on the
heat of combustion, the explosion efficiency and the mass of the fuel [22]:
η ⋅ mfg ⋅ ∆Hc
mTNT =
∆HTNT
p0
P (bar)
ip
t*
Fig. 11.4: Blast curve characteristics of a shock wave showing the
side on peak overpressure p 0 , the positive phase duration t ∗ and
t (s) positive impulse ip .
Where:
mTNT = equivalent mass of TNT (kg)
η = empirical explosion efficiency (–)
mfg = mass of hydrocarbon gas kg)
∆HTNT= heat of combustion TNT (4437–4765 kJ/kg)
The explosion efficiency depends on the substance. Cameron and Raman [18] propose
the following classes based on various sources of information:
– class I: η = 0.05 (propane, butane, flammable liquids)
– class II: η = 0.10 (ethylene, ethers)
– class III: η = 0.15 (acetylene)
With the TNT equivalent mass calculated, the next step is to calculate the scaled over-
pressure properties [7]:
R
z= 1
(mTNT ) 3
Where:
R = real distance from the center of the explosion origin (m)
z = scaled distance from the explosion origin (m/kg(1/3) )
p0
ps =
Pa
Where:
p0 = real peak overpressure (Pa)
ps = scaled peak overpressure (–)
Pa = atmospheric pressure (Pa)
t∗
ts = 1
(mTNT ) 3
Where:
t∗ = positive phase duration (s)
ts = scaled positive phase duration (s/kg(1/3) )
ip
is = 1
(mTNT ) 3
Where:
ip = real impulse (s)
is = scaled impulse (Pa s/kg(1/3) )
ta
τs = 1
(mTNT ) 3
Where:
ta = real arrival time (s)
τs = scaled arrival time (s/kg(1/3) )
Kingery and Bulmash [27] created blast parameter diagrams where, using the scaled
distance z, the scaled blast parameters can be read from the diagram and the real
explosion parameters calculated from these using the above equations. The diagram
for a surface burst (hemispherical symmetry) is provided in Figure 11.5, which is the
most representative type of geometry in case an explosion occurs at a process facility.
Vapor cloud explosions: ME method

The most common type of explosion in the process industry is the vapor cloud explo-
sion. A vapor cloud explosion occurs when a flammable gas, between the concentra-
tion of the lower explosion limit (LEL) and upper explosion limit (UEL), ignites. The
LEL is the minimum concentration of a flammable gas in the air that is capable of
igniting and causing an explosion. Below LEL the gas concentration is too lean for ig-
nition. The UEL is the maximum gas concentration in the air at which the gas can be
ignited. Above the UEL the gas is too rich to ignite. The flammable mass for a vapor
cloud explosion is the gas that is within the LEL and UEL concentration range. The ex-
plosion overpressure from a vapor cloud explosion will be more severe the closer the
mixture is to the stoichiometric combustion concentration in air. But analysis from va-
por cloud explosion incidents have shown that the fuel mass of the vapor cloud alone
is not the only parameter that defines the effect of the explosion. Depending on the
confinement and congestion in different areas where the gas is located, the explosion
characteristics can differ widely. Figure 11.6, shows its principle.
A vapor cloud explosion in an area of no congestion and confinement will have a
low laminar burning speed, which is too low to cause any significant blast overpres-
Scaled arrival time, тa (s/kg⅓); scaled duration time тd (s/kg⅓); scaled impulse is (Pa s/kg⅓)
103 тa 10–1
ps
is
102 10–2
тd
Scaled peak overpressure, ps
101 10–3
1 10–4
10–1 10–5
10–2 10–6
0.01 0.1 1 10 102
Scaled distance, z (m/kg⅓)
Fig. 11.5: Some side on blast parameters for a TNT explosion for a surface burst, hemispherical sym-
metry explosion (reused with permission from Elsevier) [28].
Fig. 11.6: Principle of the ME method showing that the blast characteristics of a vapor cloud explo-
sion is assessed individually for each region on the facility. Adapted from [19].
Dimensionless maximum static overpressure (ΔPs)

10
10 p
5 9
2 8
1 7
time
0.5 6 t1
0.2 5
0.1 4
0.05 3
0.02 2
0.01 1
Fig. 11.7: Blast chart of ME method showing the
0.005
scaled peak side on overpressure and combus-
0.002 tion energy-scaled distance for ten different
0.001 blast source strengths. Similar charts are avail-
0.1 0.2 0.5 1 2 5 10 20 50 100 able for blast characteristics such as impulse
Rn and dynamic overpressure. (Reused with per-
Combustion energy-scaled distance (R) mission from Elsevier) [28].
sures. The vapor cloud explosion will behave like a large flash fire in this case. On
the other hand, a vapor cloud explosion in a heavily confined and/or congested envi-
ronment can create high blast overpressures approaching the characteristics of dense
phase explosions. To consider this difference between dense phase explosion behav-
iors and vapor cloud explosions, specific models have been developed for the mod-
eling of vapor cloud explosions. One of the most accepted models is the ME method
developed by TNO [30]. The method is applicable to unconfined vapor cloud explo-
sions and not to vented vapor cloud explosions or internal explosions in piping or
vessel [30].
The ME model is based on numerical simulations of blasts which produced scaled
blast parameter characteristics versus a combustion energy scaled distance (Fig-
ure 11.7). The ME method is applied by identifying each blast source, characterizing
the congestions and confinement in each source and finally determining their contri-
bution to blast generation.
The following steps are followed to determine the explosion characteristics of a
vapor cloud explosion using the ME method [19]:
(1) determine cloud size
(2) identify the congested areas
(3) determine the free volume of each congested area (unobstructed areas)
(4) estimate the explosion energy E of each area: volume of cloud times the explosion
energy of the fuel (3.5 MJ/m3 combustion energy of most stoichiometric hydrocar-
bon mixtures in air assumed in the ME model)
(5) determine the free volume of each congested area
(6) determine the source strength of each area (1 to 10)

(7) determine the location of the center of each area
(8) calculate the combustion energy scaled distance from the center of each area
(9) calculate the blast characteristics at the real distance by using the ME blast charts
The combustion energy scaled distance r󸀠 is calculated as follows:

r
r󸀠 = 1
( EPexa ) 3
With:
Eex = Vcloud ⋅ Ev
Where:
r = real distance from center of area (m)
r󸀠 = scaled distance (m(2/3) )
Eex = explosion energy (J)
Vcloud = cloud volume in area (m3 )
Ev = combustion energy of hydrocarbon mixture = 3.5 ⋅ 106 J/m3
Pa = atmospheric pressure (Pa)
The scaled side on overpressure can then be calculated as follows:

∆PS
P󸀠S =
Pa
Where:
∆PS = real peak side on overpressure (Pa)
P󸀠S = scaled peak side on overpressure (–)
Critical things to consider during the use of the ME model are the validity limits of
the model, potential blast interaction between blast sources and the selection of blast
source strength.
The ME is best suited for vapor cloud explosions where overpressure effects need
to be considered at relatively large distances from the blast source. The ME model is
not suitable for assessing overpressure effects close to the explosion source and in
complex geometries such as offshore production facilities, drilling rigs and floating
production storage and offloading vessels (FPSO’s). CFD models need to be used in
such cases.
A point of consideration is up to what distance individual blast sources can still
be independent. If two blast source areas have sufficient separation distance, the ex-
plosion will start to build up pressure in the area where ignition occurred, then reach
the free field between the sources and lose flame speed and overpressure before reach-
ing the second area. If the two sources are close to each other, the blast wave will lose
some flame speed but then continue in the second area. In this case, the blast sources
cannot be considered independent and behave as one large explosion source. Thus,
the potential for interaction between blast sources also needs to be considered. The
Yellow Book [19] defines the critical separation distance as the parameter to help make
this distinction. The project ‘RIGOS’ [31] concluded that the guidance provided in the
Yellow Book [19], with respect to critical separation distance, is not conservative and
could lead to underestimating the explosion overpressures. The research showed that
critical separation distance is dependent on the overpressure of the donor explosion.
For further guidance, the reader should refer to the research report [31].
11.9 Toxic exposure
Another hazard that is important to assess is toxic substances. Toxic substances can
enter a biological organism by various routes [26]:
– inhalation – uptake via the lungs
– dermal contact – penetration through the skin
– ingestion (oral) – uptake via gastrointestinal tract
– in utero – uptake by the fetus via the placenta
In process risk assessments, the inhalation and dermal contact routes are the most
common. In product engineering, other routes can also become relevant. An example
is in the assessment of pharmaceutical products. In pharmaceutical practice, addi-
tional routes of exposure are also possible such as intravenous, intraperitoneal, sub-
cutaneous and intramuscular [26].
From a consequence assessment perspective, the exposure time and concentra-
tion of the substance are the parameters to consider. Threshold values have been de-
veloped for different circumstances such as working environment concentration limits
and emergency concentration limits. Some examples are:
– TLV-TWA – time weighed average exposure over 8 hours or 40 hours
– TLV-STEL – short term exposure limit, not longer than 15 minutes
– TLV-C – threshold limit value, ceiling limit which may not be exceeded at any time
– IDHL – immediate danger to life and health, exposure up to 30 minutes
– ERPG – emergency response planning guidelines, exposure up to 1 hour assumed
A threshold limit value is selected based on the goal of the assessment and, in some
cases, regulatory requirements. Calculations then need to be performed to assess the
concentration and exposure time and these are compared to the threshold limit value
to estimate the effects.
To assess the consequences of toxic exposure to people, dose response relation-
ships are sometimes used. A ‘dose’ is the amount of substance that is taken up by an
organism. The dose is expressed as a concentration c, raised to a chemical specific
parameter n, multiplied by the exposure time t. Since the amount taken up is time
11.9 Toxic exposure | 287
dependent [20]:
Dtoxic = c n ⋅ t
Where:
Ttoxic = the received dose
c = concentration of toxic substance exposure
t = exposure time
It should be noted that the toxic concentration in limit values or dispersion models
are sometimes expressed in mg/m3 or in ppmv. To transform the units from ppmv to
mg/m3 and vice versa, one can use the following equation assuming ideal gas behav-
ior [20]:
22.4 T 1.013
cppmv = ⋅ ⋅ ⋅c 3
M 273 Pa_bar mg/m
Where:
M = molecular weight (kg/kmol)
T = ambient temperature (K)
Pa_bar = atmospheric pressure (bar)
Since the absorbed dose is time dependent due to a varying concentration exposure
during an emergency, the expression becomes:
t
Dtoxic = ∫ [c(t)] n dt
0
The dose is used in so called ‘probity relations’, which will be explained briefly in
this section. Probit stands for ‘probability unit’ and probit functions are empirically
derived mathematical expressions linking risk (effect and probability) to exposure
(intensity, time, and duration) [26]. It is assumed in the probit function that the log-
exposure-risk relation follows a Gaussian normal distribution. By using the probit re-
lation, one can calculate the probability expressed as a percentage of an exposed pop-
ulation which will exhibit a given adverse effect as related to a given exposure. Probit
functions are expressed in the form:
Pr = a + b ⋅ ln (Dtoxic )
Where Pr is the probit value, a and b are regression constants and Dtoxic is the toxic
dose. The regression constants depend on the toxic effect that is assessed and are
available for a limited number of substances. Transformation tables for probits to re-
sponse fractions (%) have been developed by Finney [32] and adopted in the TNO
Green Book [33]. Analytical expressions are also available for calculations, which are
useful in cases when computer models are used for the calculations [34].
The reader should refer to the Green Book [33] in which the subject is treated in
more detail. Note that toxic exposure relations provided in 1992 might have changed
so be advised to mainly use the source to get familiar with the methodology and check
more recent literature sources before using the dose response relations mention in the
Green Book.
Note that the application of probit relations is not limited to toxic effects. There are
probit relations available in literature that can be used to calculate the consequence
for fire and explosions, such as those provided in the Green Book [33] and Lees [7].
11.10 Codes, standards and designing ALARP
Having identified the hazards and assessed the consequences, how does one choose
the relevant engineering design cases? What is, and when is it ‘safe enough’. First,
there are several information sources, approaches and methods that can assist engi-
neers making these choices, such as:
– regulations, codes and standards
– company specific design engineering standards
– ALARP and risk analysis
These will be elaborated on in the following sections.
Regulations, codes and engineer standards

Engineering regulations, codes and standards all serve to ensure that products and
processes are safe and of superior quality. Each has their own characteristics and there
are differences between them, as follows:
– Technical regulations: mandatory requirements set by a regulatory authority,
defining the characteristics and/or performance requirements of a process, prod-
uct or service.
– Engineering codes: laws and regulations specifying minimum requirements to
protect safety, health and the environment. Codes will typically refer to standards
for specific details on requirements that have not been specified in the code itself.
Codes are enforced by one or more governmental entities and are important for
establishing best engineering practices.
– Engineering standards: technical documents defining technical methods or char-
acteristics of a product, process or service. They are not laws and are thus not
enforced by regulatory authorities but are considered as engineering practice.
In other words: codes specify what needs to be done and standards detail how to ac-
complish this.
Regulations, codes and standards specify in some cases what the design cases
should be. But in many instances, this is not the case. And even when they do, there
are cases during engineering projects where there are conflicting design requirements
11.10 Codes, standards and designing ALARP | 289
between standards or they provide too little information on how to deal with a specific
design challenge. In such cases, further analysis or reviews are required to make the
proper design choices. This is where the ALARP principle and risk analysis comes into
the picture when reviewing design aspects that could have an impact on safety, health
or the environment.
ALARP principle
ALARP has its origin in the UK Health and Safety at Work, etc. Act from 1974 [35].
The principle’s goal is to avoid making sacrifices grossly disproportionate to the ben-
efits gained. An extreme example might be to spend $1m to prevent five staff suffering
bruised knees, which is obviously grossly disproportionate. But to spend $1m to pre-
vent a major explosion capable of killing 150 people is obviously proportionate [36].
The principle is outlined in Figure 11.8:
Unacceptable
region
Increasing risk
Tolerable
region
Broadly acceptable
region
Fig. 11.8: ALARP triangle.
The risk is divided into three regions, shown in Figure 11.8 [37]:
– Unacceptable region: risks are too high and therefore risk reduction measures are
required, unless there are exceptional reasons for the practice or activity to be
retained.
– Tolerable region: risk reduction should be applied if the sacrifices are not dispro-
portionate to the benefits gained.
– Broadly acceptable region: risks falling into this region are generally regarded as
insignificant and adequately controlled. They are typical of the risk from activities
that are inherently not very hazardous or from hazardous activities that can be,
and are, readily controlled to produce very low risks.
The challenge of the ALARP principle lies in the practical application of the principle.
What is considered ‘reasonably practical’ can be highly subjective. The UK’s Health
and Safety Executive website provides some guidance to inspectors to assess if duty
holders have reduced risks to as low as reasonably practical. Camaron and Raman [18]
state factors that come into consideration when assessing practicability under the
ALARP principle:
(1) Can the consequences be eliminated?
(2) If elimination is not possible, can they be mitigated?
(3) What are the possible mitigation measures?
(4) Are the mitigation measures within engineering practicability?
(5) If it is not practicable to design the mitigation measure due to engineering con-
straints, then can the likelihood of an occurrence resulting in the consequence be
reduced?
(6) What are the possible likelihood reduction measures?
(7) Are the likelihood reduction measures within engineering practicability?
(8) What are the benefits of the risk reduction measures? This may not be a reduction
in public risk alone, but should also address risk to employees, benefits gained in
minimizing interruptions to operations, etc.
After having established what is deemed ‘practical’ it is required to assess if these

measures are ‘reasonable’. Cost benefit analysis and risk assessment are some of the
important tools to assess this. The UK Health and Safety Executive provides guidance
on how to perform the cost benefit analysis [36] but warns that it should not be the
sole criteria to determine if the ALARP level has been reached.
11.11 Risk analysis
Risk analysis is a tool that can be used as part of an ALARP assessment and help in
making design decisions. There are two types of assessments:
– Qualitative, where risk is expressed in a nonquantified way (e.g., low, medium
and high).
– Quantitative, where the result of the assessment is a number and compared to a
risk acceptance criterion.
The following sections provide a brief overview of the methods and their advantages
and disadvantages.
Qualitative assessments
A commonly used qualitative tool in the chemical and process industry is the risk ma-
trix. The risk matrix is a graphical representation of the risk as a function of the con-
sequences and frequency or probability of occurrence. Consequence categories can
be defined for people, environment, assets and reputation. An example is provided
in Figure 11.9. It has three risk categories (low, medium and high) but more could be
Consequences Frequency of occurrence

Severity level 1 2 3 4 5
People
Very unlikely Unlikely Possible Likely Very likely
-4 -4 -3 -3 -2 -2 -1 -1
< 10 /yr 10 /yr -10 /yr 10 - 10 /yr 10 - 10 /yr >10 /yr
More than
5
5 fatalies
Between 1
4 and 5 High
fatalies
Major
3 Medium
injury
Minor
2 Low
injury
Slight
1
injury
0 No injury
Fig. 11.9: Example risk acceptance matrix defining consequences for people.
allocated. The categorized regions on the matrix correspond with the regions of the
ALARP triangle (Figure 11.8). When the outcome of the risk assessment is in the ‘high’
category, the risk is unacceptable and risk reducing measures are required. If a risk is
in the ‘low’ category, the risk must be managed for continual improvement. If a risk is
‘medium’ category, risk reduction measures must be incorporated until ALARP.
The matrix in our example has a 5 × 5 dimension but other dimensions such as
3×3, 4×4, 6×6, 4×6, etc., are also possible with more or less allocated severity and fre-
quency categories. Other examples are available in industry guidelines such as from
the International Organization of Oil and Gas Producers [38] and ISO standards [39].
The same principles apply to all the matrixes, irrespective of their dimensions. Major
chemical and oil and gas producing companies have a corporate defined risk matrix
which is used as a basis for all the process safety related risk assessment studies. The
advantage of having a corporate defined risk matrix is that it helps to achieve an in-
crease in consistency between risk assessments in a company.
Qualitative risk assessments, such as the risk matrix, have the advantage that they
are relatively easy and quick to use and can help to prioritize actions. Despite its ap-
parent ease of use, there are in fact several areas where the risk could be incorrectly
addressed leading to a wrong prioritization of actions and incorrect screening of high
risks. A critical point to remember when using the risk matrix is its limited suitability
for assessing complex and high risks. In those cases, decisions should be based on
more quantified risk assessment methods to decide if the risks have been reduced to
ALARP.
Quantitative assessments
In case of complex systems with the potential for major hazards, quantified risk as-
sessment can be employed to assist in making engineering decisions. This section will
briefly discuss some common techniques and their advantages and disadvantages.
Quantitative risk analysis (QRA) requires failure frequency of equipment and an
assessment of the consequences as input for the assessment. Consequence assess-
ment models have been discussed previously in this chapter. Frequency data can be
obtained from online and offline information sources. Several databases containing
failure frequencies are available free of charge from sources such as from the interna-
tional association of oil and gas producers [40] and the UK Health and Safety Execu-
tive [41]. The quantitative assessments that will be discussed in the following sections
are the fault tree analysis (FTA), event tree analysis (ETA) and the QRA.
Fault tree analysis

A FTA investigates how a top event (fault), such as an exothermic runaway, loss of con-
tainment, etc., originates from individual causes. It is a deductive technique meaning
that one starts at the outcome. The goal of the FTA is to calculate the probability or
frequency of the top event. It also provides insight into how a top event scenario de-
velops from its individual causes. An example is provided in Figure 11.10. The name is
derived from the fact that the top event is the apex that branches to individual causes,
thus resembling a tree. A fault tree considers that the system components are in one
of two states: a correct state or a fault state. Note that there is a difference between
a failure and a fault in a component. A fault is an incorrect state which may be due
to a failure of that component or may be induced by some outside influence. Thus,
fault is a wider concept than failure. All failures are faults, but not all faults are fail-
ures [7]. A fault tree is constructed by using ‘gates’. Gates are points in the diagram
where branches meet. Two common gates are the ‘AND’ and ‘OR’ gates. At the ‘AND’
gate, all the components meeting at this gate must fail to progress further towards the
top event. At the ‘OR’ gate, one of the components meeting at this point must fail to
further progress towards the top event.
To effectively use the fault tree the following needs to be prepared and decided
before the analysis:
– The goal of the analysis such as screening of scenarios, verification of codes, com-
paring design alternatives, etc.
– Scope of the system to be investigated.
– Frequency and probability data.
A fault tree diagram can then be constructed through the following steps:
(1) Define the top event.
(2) Decide where the analysis will start on the system diagrams such as Piping & In-
strumentation Diagrams (P&ID) and Process Flow Diagrams (PFDs). For safety re-
lated assessments, this will typically be the point where the safeguarding system
starts to act.
Top event frequency

Overpressurization = F(x)*P(x)
reactor
AND gate
F(x)=A+B+C
P(x)
Closed Fail
valve pressure
A,B or C relief
OR gate
Control Control Closed

failure failure valve C = C1+C2
A B C
A B
Control Mechanical
failure failure
C C
C1 C2 Fig. 11.10: Example fault tree.
(3) Move through the system against the flow of liquid, gas, energy, current, etc., and
consider the faults of all relevant components.
(4) Continue modeling up to the point where the goal of the analysis has been met.
The FTA and has the advantage that risk contribution to the top event from individual
components can be quantified. The limitation of a FTA is that only Boolean states are
assessed (fail/correct) and thus no account is given to partial functioning, temporal
effects and spare parts [16]. There are various references for the construction of fault
tree diagrams such as Lees’ Loss Prevention in the chemical industries [7].
The FTA also forms the basis of the layers of protection analysis (LOPA). The LOPA
is becoming more common in the process industry as a tool to determine how reliable
independent barriers will have to be to prevent the realization of a single process haz-
ard. The steps of a LOPA consist of:
– calculate initial event frequency without taking barriers into consideration
– compare result to a quantitative criterion and determine the ‘gap’
– calculate if current barriers reduce the risk sufficiently
– if not, redesign, increase reliability of barriers and/or introduce new independent
barriers to ensure criteria are met
The technique is sometimes used during safety integrity level (SIL) assessment. Dur-
ing a SIL assessment, the level of required protection of an instrumented system is
based on the IEC61508 and IEC61511 standards. For more information on the LOPA
methodology, the reader should refer to the book ‘Layer of Protection Analysis: Sim-
plified Process Risk Assessment’ [42].
ETA
The ETA is a complementary method to the FTA, employing an inductive method to
assess branching consequences from an initial cause. Probabilities or frequencies can
be assigned to each of the branches. An example for a hydrocarbon gas scenario is
provided in Figure 11.11.
Construction of an event tree consists of the following steps:
(1) Define the initiating event.
(2) List the sequence of protections of the system.
(3) Start from the first layer and branch for a yes/no for the success or failure of the
layer.
(4) Continue branching for each layer until the sequence has been completed.
(5) Describe the consequences at each end of the event tree. Note that it is possible
that multiple branches have the same consequences, meaning that multiple path-
ways to these consequences are possible.
Fail process Fail Ignition Immediate Consequence

control ESD ignition P =
0,80
Jet fire
F1 =2,4 ⋅10-3/yr
P=
0,30
P= P=
0,10 0,20 Vapor cloud explosion
F2 =6,0 ⋅10-4/yr
P= P=
yes 0,10 0,70 Hydrocarbon gas release
F3 =7,0 ⋅10-3/yr
P=
0,90
Shutdown production
F4 =9,0 ⋅10-2/yr
P=
0,90 Continue
no
normal production
Initiating cause: F5 =9,0 ⋅10-1/yr
high pressure
F0 = 1,0 /yr
Fig. 11.11: Example event tree for a pressure peak in a gas production facility. Frequencies and proba-
bilities are only for illustrative purposes.
(6) Assign numeric values to the branches and calculate the frequency of the con-
sequences. Calculation of the consequence frequencies is performed by multipli-
cation along the branches. One should check that each probability node on the
event tree sums up to unity and that the consequence frequencies sum up to the
initial event frequency.
The advantage of the event tree is that it is a good method to show many possible
consequences from a single scenario. The event tree is also used as a precursor for
QRAs.
Individual and group risk

The tools mentioned in this chapter can be used as input to a full QRA of a system.
A QRA will assess the frequencies and consequences of a variety of scenarios and is
therefore a more involved study compared to other assessments. The goal of the QRA
can vary from land planning use to assessing if the risk to operators on the indus-
try site is acceptable. In the process industry, it is common to investigate the risk of
a fatality, since this is a severe consequence that is more easily definable compared
to the risk of injury. General assumptions in QRA assessments are that contribution
of all incidences on the facility are additive and that the risk receiver is exposed to
the presence of the risk source over a certain period (usually one year). There are two
categories of risks that can be investigated:
– Individual risk (IR): the frequency at which an individual may be expected to sus-
tain a given level of harm from the realization of specified hazards [43].
– Group risk (GR): expresses the relationship between the frequency and the num-
ber of people suffering from a specified level of harm [18].
If consequence severity values are calculated for a meshed area around a risk source,
points of equal risk can be connected to produce individual risk contours at various
levels, for example, 10−4 , 10−5 , 10−6 or 10−7 [16]. If directional effects are absent, the
contours are circular, and if a dominating wind direction exists, or directional blast or
fragment projection occurs, contours become ellipsoidal or even irregular [16].
Group risk (also called societal risk) differs from IR in the fact that it expresses
the relationship between the frequency and the number of people suffering from a
specified level of harm as opposed to the individual approach [16].
[1] Pat-El IE, et.al, Major Accidents and Their Impact – The EU Directive for Offshore Safety, Society
of Petroleum Engineers conference paper 172094, 2014.
[2] CCPS. Guidelines for Chemical Transport Analysis, American Institute of Chemical Engineers,
New York, 1995.
[3] CCPS. Chemical Process Quantitative Risk Analysis, 1999

[4] Klotz TA. Inherently Safer Design – The Growth of an Idea, Process Safety Progress, 15, 1996,
5–8.
[5] AIChE. Dow’s Fire & Explosion Index Hazard Classification Guide, 7th edn, June 1994.
[6] AIChE. Dow’s Chemical Exposure Index Guide, 1st edn, 1994.
[7] Lees FP. Lees’ Loss Prevention in the Process Industries, edited by Sam Mannan, Burlington,
Butterworth-Heinemann, 3rd edn, 2005.
[8] Urben PG (ed). Bretherick’s Handbook of Reactive Chemical Hazards, Oxford, Academic Press,
7th edn, 2007.
[9] CCPS. Essential Practices for Managing Chemical Reactivity Hazards, 1st edn, 2003.
[10] CCPS. Chemical Reactivity Evaluation and Applications to Process Design, 1995.
[11] CCPS. Guidelines for Safe Storage and Handling of Reactive Materials, 1995.
[12] Leggett, DJ. Chemical Reaction Hazard Identification and Evaluation: Taking the First Steps.
AIChE Loss Prevention Symposium, New Orleans. March. 2002.
[13] IEC 61882, Hazard and Operability Studies – Application Guide, 2001.
[14] ISO 14224, Petroleum, Petrochemical and Natural Gas Industries – Collection and Exchange of
Reliability and Maintenance Data for Equipment, 2016.
[15] CCPS. Guidelines for Hazard Evaluation Procedures, 2nd edn, 1992.
[16] Pasman HJ. Risk Analysis and Control for Industrial Processes – Gas, Oil and Chemicals, 1st
edn, 2015.
[17] CGE. Bowtie XP Methodology Manual, Rev.15, March 2015.
[18] Camaron IT and Raman, Process System Risk Management, Volume 6, 1st edn, Elsevier, 2005.
[19] VROM. Methods for the Calculation of Physical Effects – ‘Yellow Book’ Publication Series on
Dangerous Substances PGS 2, 3rd edn, 2005.
[20] Casal J. Evaluation of the Effects and Consequences of Major Accidents, Industrial Safety Se-
ries, Vol 8, Amsterdam, Elsevier, 2007.
[21] HSE. Methods of Approximation and Determination of Human Vulnerability for Offshore Ma-
jor Accident Hazard Assessment, 2010. Sourced 26 January 2018 from: http://www.hse.gov.
uk/foi/internalops/hid_circs{/\penalty\exhyphenpenalty}technical_osd/spc_tech_osd_30/
spctecosd30.pdf.
[22] CCPS. Guidelines for Consequence Analysis of Chemical Releases, 2nd edn, Hoboken, Wiley,
1999.
[23] Chamberlain, GA. Developments in Design Methods for Predicting Thermal Radiation from
Flares, Chem. Eng.Res.Des., 65, 1987.
[24] Johnson, AD, et.al. A Model for Predicting the Thermal Radiation Hazards from Large Scale
Horizontally Released Natural Gas Jet Fires, Trans.IChemE, 72, Part B, 1994.
[25] OGP. Consequence Modeling, Report No. 434-7, March 2010. Sourced 26 January 2018 from:
https://docslide.com.br/documents/ogp-434-7-consequence-modelling.html.
[26] Lemkowitz SM, and Pasman HJ. Chemical Risk Management, Faculty of Applied Sciences,
Chemical Engineering, Delft University of Technology, 2002.
[27] Kingery CN, and Bulmash G. Air Blast Parameters from TNT Spherical Air Burst and Hemispher-
ical Surface Burst, Tech. Rep. ARBRL-TR 02555, US Army, Ballistic Res. Lab., Aberdeen Proving
Ground, MD, 1985.
[28] Lees FP, Lees’ Loss Prevention in the Process Industries, edited by Sam Mannan, Burlington,
Butterworth-Heinemann, 4th edn, 2012.
[29] Van den Berg, AC. The Multi-Energy method – A Framework for Vapor Cloud Explosion Blast
Prediction, Journal of Hazardous Materials, 12, 1985.
[30] TNO. Application of Correlations to Quantify the Source Strength of vapor Cloud Explosions in
Realistic Situations, Final Report for the Project ‘Games’, Report PML 1998-C53, October 1998.
[31] HSE. Research to Improve Guidance on Separation Distance for the Multi-Energy Method (RI-
GOS), Research Report 369, 2005.
[32] Finney DJ. Probit Analysis, 3rd edn, Cambridge, Cambridge Press 1971.
[33] Green Book, Methods for Determining Possible Damage, 1st edn, 1992, Sourced 26 January
2018 from: https://www.scribd.com/doc/61170131/Green-Book-Methods-for-the-
Determination-of-Possible-Damage-CPR-16E.
[34] Vilchez JA, et.al. J. of Loss Prevent. Proc., 14, 2001.
[35] HSE. Guidance, Sourced from: http://www.hse.gov.uk/risk/theory/alarpglance.htm, accessed
on 18-12-2016.
[36] HSE. Cost Benefit, Sourced from: http://www.hse.gov.uk/risk/theory/alarpcheck.htm, ac-
cessed on 18-12-2016.
[37] HSE. Reducing Risks, Protecting People – HSE’s Decision Making Process, London, HMSO,
2001.
[38] OGP. HSE Management – Guidelines for Working in a Contract Environment, report 423, June
2010.
[39] ISO 31000, Risk Management, 2009.
[40] OGP. Risk Assessment Data Directory, Report 434, March 2010.
[41] HSE. Failure Rate and Event Data for use within Risk Assessments, Sourced from:
http://www.hse.gov.uk/landuseplanning/failure-rates.pdf, accessed January 2017.
[42] CCPS. Layer of Protection Analysis: Simplified Process Risk Assessment. 1st edn, Hoboken,
Wiley, 2001.
[43] Jones DA. Nomenclature for Hazard and Risk Assessment in the Process Industries, 2nd edn,
Rugby, England, Institution of Chemical Engineers, 1992.
The following references are recommended for further reading: [9, 10, 15, 16, 18–20,
22, 28, 33, 42]
12 Evaluating for environmental, social and
sustainable development aspects
12.1 Environmental evaluations
12.1.1 Introduction to environmental evaluations
The scope of environmental impact assessment has increased enormously in the last
60 years. In the 1960s, legislation was set by individual countries and was very limited.
In the Netherland, for instance, laws were implemented to reduce low level smog, by
limiting low level acid gas emissions. Very tall factory chimneys, of 200 meters and
higher, were then constructed to make sure that acid flue gases no longer caused local
low layer smog. In the seventies and eighties, these tall chimney emissions caused
acid rain and thereby caused forests to die in other countries; notably in Sweden. This
in turn caused stringent flue gas emission legislation in Western European countries,
which were later adopted by the European Union. Other industrialized regions, such
as North America, East Asia and South America followed suit.
For product design, such as for refrigerants, long range environmental effects,
such as atmospheric ozone layer depletion and river water quality deterioration due
to laundry washing products, were considered. In the nineties and beyond, enhanced
global warming by burning fossil fuels caused worldwide concerns, as expressed by
the United Nations and global companies.
Presently, the scope of environmental impact assessment is on a global scale
and concerns environmental effects on the Earth’s atmosphere, on water, soil, and
space [1]. Environmental evaluation of designs is now an essential element all over
the world. Companies not only comply to environmental laws, but go beyond laws as
they anticipate that in future these laws will be more stringent. As such, it is more
economical to make designs more robust to meet the requirements of these future leg-
islations, rather than having to invest later in end of pipe solutions. Many companies
have now have their own environmental strategy, policy and regulations regarding
environmental impacts of novel product and process operations. Some do it to keep
their license to operate, some to anticipate new, more stringent environmental laws
and some as part of their contribution to sustainable development. The next sections
provide methods and guidelines for environmental evaluation for each innovation
stage gate.
12.1.2 Environmental evaluation: discovery stage
In the discover stage, very little environmental information is available or can be gen-
erated on the novel product and/or process idea. On the other hand, only little infor-
https://doi.org/10.1515/9783110467741-012
mation is needed as the only question to be answered is whether the idea is better than
the commercial product and the process with which the idea competes.
Here are two simple metrics to evaluate the novel idea on environmental impact. It
is based on an estimate of the maximum amount of product (hence, minimum amount
of waste) that will be produced by the novel idea. For a product, this can be directly
estimated from output/input mass ratio. For a simple chemical product, it can be de-
rived from the stoichiometric reaction equation. The metric is the mass efficiency of
the stoichiometric reaction (MEST) determined from the equation [2]:
MEST = product mass/(total mass of all molecules produced)
The MEST value should be calculated for the novel product or process and be com-
pared with the MEST value of the reference case.
An alternative metric often used is atom efficiency [3]. This also is defined as ratio
of product mass and total mass of molecules generated by the reaction. If the stoichio-
metric equation is used to determine the atom efficiency, then the atom efficiency will
have the same value as the MEST value.
It is also useful to calculate the actual mass efficiency (ME) of the reference case.
The mass efficiency is the inverse of the E factor of [3]. The ME metric has the advantage
that it can be directly compared to MEST values.
ME = actual product mass/(total actual mass waste and product)
If the MEST value of the novel idea is lower than the MEST value of the reference case,
than the novel idea scores poorly on the environmental criterion, even if the MEST
value of the novel idea is higher than the ME value of the reference case.
The reasoning for this latter statement is that the ME value of the reference case
will increase in time through improvements to the reference case. This may well hap-
pen within the time needed for the idea to become implemented.
In the discover stage the environmental impact of the novel product and process
should be enormously lower than reference case, if the reference case has a large envi-
ronmental impact. In the present production of fine chemicals and pharmaceuticals,
the amount of waste produced far exceeds the amount of product due to stoichiomet-
ric consumption of additional chemicals [2, 3]. By applying catalysis this waste can be
reduced enormously.
12.1.3 Environmental evaluation: concept stage
In the concept stage, the novel product and process design should be evaluated on
its environmental impact in comparison with the reference case on all aspects. How-
ever, limited resources are available so the evaluation should still be simple. The best
method available is the conceptual life cycle assessment method (LCA), which is pro-
posed in the next section.
300 | 12 Evaluating for environmental, social and sustainable development aspects
Environment Inputs
Outputs
Extraction
Atmosphere
Outputs
Water Processing
Outputs
Land Use
Outputs
Discarding
Outputs
Fig. 12.1: LCA system elements.
LCA for concept stage

There are three levels of LCA: conceptual, simplified and detailed. For the concept
stage, the conceptual LCA is most suitable. In my courses on sustainable design for
products, processes and systems, I worked out a simple stepwise approach to con-
ceptual LCA, called rapid LCA [4]. First a simplified systems description for LCA is
provided in Figure 12.1.
System elements of LCA are the environment (with subsystems, atmosphere, wa-
ter and land) and the steps (extraction, processing, use and discarding). The concept
LCA considers all mass inputs from the environment and all mass outputs to the envi-
ronment for these life cycle steps. In a complete LCA, far more steps will be involved,
such as transport between the major steps and more process steps. Other feeds will
also be involved, but these are neglected in this limited LCA. In most cases, the con-
clusion that the concept is better than the reference case will appear to be correct,
when in latter stages a complete LCA is executed.
The inputs and outputs are at first instance defined as mass flows with their com-
position the same as streams in process design. Then inputs and outputs are further
defined on their environmental impacts, such as eco toxicity.
The rapid LCA involves:
(1) defining the functional unit
(2) defining all life cycle steps from primary production to end of life and preferable
recycle steps
(3) selecting one or two key environmental impact types
(4) identifying all major emission streams for each life cycle step
(5) quantifying the environmental impact
(6) comparing the rapid LCA of the novel design with the reference case
(7) concluding whether the novel design has a lower environmental impact or not
Ad 1: Functional unit. For a comparison, it is very important to have the same func-
tional unit definition for the alternatives and the reference case. For a process design
the functional unit will be the product produced by the process. The functional unit
definition is then given by the amount of product to be produced and the product spec-
ifications. The definition of the amount of product produced can be simply design ca-
pacity expressed as the amount in tons produced per year.
A word of warning should be stated here; economist often express the amount
in absolute mass dimension (ton), while they mean the amount per year (ton/year).
Chemists also often express the amount in absolute mass (grams), but they really
mean absolute mass. For a comparison between different processes with the same
product, the functional unit definition should be the same. As all inputs and outputs
will be linearly related to the amount of product, the selected dimension and size of
the functional unit can be any if it is the same for all alternatives considered.
For novel product design, the definition of the functional unit is in general far
more complicated as often it is an action, such as a single hair wash, rather than an
absolute amount of a material. If the novel product design replaces several existing
products, then the functional unit should be defined in such a way that it is clear how
much of each product of the reference case makes up the functional unit.
Ad 2: All life cycle steps. For a novel product, most upstream life cycle steps can
be found by a literature search on LCA about the input components. The downstream
steps after product use may need some imagination. For a novel process for an existing
product, the LCA of the existing product can be used and adapted to the input and
output streams of the novel process compared to the reference process case.
Ad 3: Selection of the key environmental impact types will, in most cases, directly
follow from the major environmental impact types (see Table 12.1) and is obtained from
LCA data of the reference product(s).
Ad 4: Literature on feedstock stream manufacturing descriptions helps to find
emission streams data.
Ad 5: Crude estimates of each stream for the selected impact types will be made
and added up.
Ad 6: A comparison on environmental impacts of the novel design and the refer-
ence case will be made
Ad 7: From the comparison, a conclusion on the environmental impact of the novel
design compared to the reference case for the key impact type will be made. A quali-
tative statement about the assumptions for the other environmental impact types will
also be made.
A helpful method for a rapid evaluation in the concept stage is to fill in template
Table 12.2 for all environmental impacts for which data is available. For impacts with
no data, ‘no data yet’ is filled in. For some of those impacts it may be possible to reason
that the impact for the modern design will be lower than the reference case. Then
‘lower’ can be filled in for the 4th column.
Tab. 12.1: Environmental impact types for LCA [5].
Atmosphere Water Biotic Soil

Global warming Eco toxicity Eco toxicity Eco toxicity
Ozone layer depletion Nitrification Habitat reduction Physical degradation
Photochemical pollutants Heat Biotic resource reduction Dehydration
Acidification Physical ecosystem reduction Resource depletion
Radiation
Noise
Smell
Visual landscape
Tab. 12.2: Template environmental impact ratio innovative design/reference case.
Impact type Reference impact Innovative design Impact ratio

impact new/reference
The numbers of the last column are used for the evaluation. If all ratios are less than
one then the environmental evaluation has a positive outcome. If for some impacts the
ratio is less than one, but for at least one impact the ratio is higher, than the outcome
is negative. It has a resemblance to a Pareto improvement.
The advice for this choice of improvement is based on the experience that it is
nearly impossible to convince managers, stakeholders and/or customers that the
product is environmentally better if it has a lower environmental impact for some im-
pact types but has a higher impact for at least one impact type. It also avoids lengthy
discussions on how to use weight factors for the individual impact types to get to
single environmental impact figures.
Process output evaluation using atom balances

The process concept design can be quickly evaluated on the completeness of all out-
put streams by a simple atom balance over the process input and output streams. Any
atom that enters the process will leave the process. Often when a crude input such as
biomass is used as input to the process, the concept designer neglects or forgets that
the biomass contains atoms, such as potassium and sulfur, which will leave the pro-
cess in one of the outlet streams. A simple atom balance analysis then reveals that the
output streams are not sufficiently defined and that the envisaged output destination
of some output streams are infeasible.
Fugitive emissions evaluation process concept design

For the process concept itself, fugitive emissions via valves, etc., can also be estimated
from simple process flow sheets, where major unit operations or process equipment is
defined. Environment protecting agency (EPA) provides for most unit operations and
process equipment directly the amount of fugitive emissions in kg/h [6].
12.1.4 Environmental evaluation: feasibility and development stages
A simplified LCA method can be chosen to assess the environmental impact of the
novel design compared to the reference case.
The simplified method consists of:
– Screening for important inputs and outputs of the life cycle.
– Simplifying by selecting a few most important impact categories for the inputs
and outputs.
– Assessing reliability of simplified LCA for comparison innovative design with ref-
erence case.
The simplified method still looks like the concept method. It takes more impact cate-
gories in consideration and uses more quantified literature data. It is, however, prone
to uncertainties whether the novel design is better than the reference case. Because
the purpose of the feasibility stage is to take a decision of a major investment in prod-
uct prototype testing and/or process pilot plant purchase, these uncertainties are in
general not acceptable. Therefore, a full detailed LCA is recommended.
In the follow up stages, a detailed LCA is the best method to fully assess environ-
mental impacts associated with all the steps of a product’s life from cradle to grave,
i.e., from raw material extraction through materials processing, manufacture, distri-
bution, use, repair and maintenance and disposal or recycling. The procedures of LCA
are part of the ISO 14000 environmental management standards [7]. It is an elaborate
method and only provides an accurate result if the ISO 14040 method is applied [8].
A complete LCA of the novel product should then be determined for the feasibil-
ity stage, using the certified well defined LCA method ISO 14040 and 14044 [9]. ISO
14040 and ISO 14044 are the leading standards for LCA. In addition to applying the
LCA method properly, reliable LCA databases are needed in addition to get reliable
LCA of the novel product and the reference case. These databases may be purchased
from institutes specialized in LCA data gathering and processing.
The quality of the LCA result not only depends on the quality of the databases
used but also on the choice of database type, and the origin and future developments.
The database can be based on measured data for the case at hand (called foreground
data) or based on averaged data, holding data for a certain country or region. As such,
choosing the relevant country is very important. For breakthrough innovations, which
will take several years to be implemented and then will be applied for a long time, the
LCA database will not be very relevant for the long term future. This is due to emis-
sions and intakes from other steps in the LCA changing due to innovations. Given this
uncertainty, the best approach is to make sure that the innovation scores are far better
on environmental impacts.
The LCA results will reveal impacts for each impact type. There are ways to nor-
malize these different outcomes to a single impact figure, but this is not recommended.
A novel product that is far better on one or more impact types, but worse on one
or more others, compared to the reference case, will lead to endless discussions in-
side and outside the company (NGO’s). It is far better to ensure that the novel product
and/or process is not worse on any of the impact types and considerably better on the
major impact types. The product and process will be easily communicated as environ-
mentally improved.
For a detailed description and application, the reader is referred to the LCA hand-
book [5]. For a nice overview of all LCA methods and ranking, the reader is referred to
a booklet generated by the European Commission [10].
12.1.5 Environmental evaluation: detailed design stage
In addition to the full LCA, environmental laws and regulations of the countries where
the product will be marketed, where the process will be located and the countries
where the feedstock will be obtained will also be considered. Often a public environ-
mental impact report must be made to facilitate local civilian’s objections to certain
emissions or risks of emissions.
In the detailed design stage, fugitive emissions for the process should also be de-
termined. These are emissions from flanges and seals. The EPA method for this is easy
to use [11].
12.2 Evaluating for social acceptance
12.2.1 Importance of social acceptance
Social acceptance of novel products and processes is enormously important. It can be

the show stopper. Think, for instance, about nuclear power plants in Germany, which
are no longer accepted by the government and will be phased out. Also consider the
Brent Spar case of Shell. Because German petrol buyers did not accept the Shell stance,
dumping the platform as waste in the ocean seemed the best solution to Shell. Petrol
buyers bypassed Shell petrol pumps and bought petrol and diesel elsewhere. Within
six weeks, Shell reviewed its position and invited parties to come up with better solu-
12.2 Evaluating for social acceptance | 305
tions for the platform which led to many solutions being presented. The best selected
solution was moving the platform to a Fjord in Norway, dismantle it there and reuse
all steel and concrete parts in other applications.
For new processes, a license to operate for decades is therefore essential. That
license is not only governed by law but also by social acceptance. For consumer prod-
ucts in a free market with competition, social acceptance is the first criterion for suc-
cess or failure. It cannot be separated from market acceptance. Social acceptance is
related to the beliefs and cultures of societies in which the product and process will
be located. When, for instance, using animal material for products while depending
on the cultures in which the products will be marketed, religious criteria should be
considered.
The next section shows a comprehensive list of items relevant for social accep-
tance. This will be of use in the concept stage as a checklist for the novel design. For
the feasibility and development stage, social and market acceptance for novel prod-
uct is of major importance. For these stages, market surveys and inquiries are the key
methods for determining acceptance. In other sections of this book those surveys are
described.
12.2.2 Social acceptance aspects
Aspects relevant for social and market acceptance are shown in Table 12.3 and briefly
discussed thereafter. The word innovation used here stands for the novel product, pro-
cess, value chain and life cycle.
Tab. 12.3: Checklist for social and market acceptance of novel product and process.
Derived from [Verkerk [12], Potter [13] and de Soto [14].
Modal aspect Essence of social and market acceptance

Beliefs The innovation is acceptable to believers
Ethical The innovation agrees with moral values
Juridical The innovation agrees with laws
Aesthetic The innovation is appealing
Economic The innovation price fits its functions (value for money)
Social The innovation fits in the social context
Linguistic Symbols and language provide the correct perception of the innovation
Historic The history of the innovation and company is attractive to customers
Logical The logic of the innovation is clear
Sensitive Senses (smell, taste, hearing, sight, feeling), and emotions associated to the
product and process are appealing
The upper part of the modal aspects of artifacts of Verkerk [12] has been used for the
first column of the checklist (Chapter 4, Table 4.8 contains the full list of modal as-
pects). The second column has been filled in using information from Potter [13] and
de Soto [14] and my own observations. A brief description for each modal aspect re-
lated to novel product acceptance follows hereafter.
Beliefs of a society can have a strong influence on social acceptance of the novel
product and/or process. This holds for societies with a named governing belief, such
as Hinduism (no animal derived feeds), Islam (no ethanol) and Humanism (no human
rights violations).
Ethical and moral aspects of a product over its value chain are often very impor-
tant for social acceptance. Think, for instance, of child labor associated with produc-
tion of an intermediate in the value chain.
Juridical aspects such as laws and regulations can make the product unaccept-
able, because the product (or its manufacturing process) violates laws related to
safety, for instance. Also, the laws on property and land ownership regulations are
very important [14].
Aesthetic aspects such as the shape of the product (packing) can make or break a
consumer product.
Economics of a novel product are often very important for acceptance. The prod-
uct price should fit the product function. It should also be at least like the reference
product for the same function.
Societal aspects of the novel product should be such that it fits with the local so-
ciety and preferably enhances its welfare. Contributing to socially sustainable devel-
opment goals will help in this respect. Also, the market type (centrally governed, free
market or regulated market, informal market and worker union presence) is relevant
for introducing a novel product or process. Furthermore, the government structure,
i.e., democracy/dictatorship, is of importance for introducing novel products and pro-
cesses.
Linguistic aspects are particularly important for new consumer products. The
product is often not seen as an artifact fulfilling a function, but as providing meaning
to the person and its social relations. For example, a novel nail polish may send a
message to the person who sees it on the nails of the lady.
Historic aspects can be very important. When the new product is intended to re-
place an existing product, while the history related to the reference product has such
strong ties in society that the new product is not accepted.
Logical aspects of the new product and process should be applied. If the novelty
makes no sense, the new idea will not survive.
Sensitivity is an enormously important aspect for consumer products. The impres-
sion it has on the human senses will make the product attractive or unacceptable.
12.3 Sustainable development assessment | 307
12.3 Sustainable development assessment
12.3.1 Sustainable development history and consensus
Sustainable development was, for the first time, defined by Brundtland for the United
Nations in 1987 [15]:
“Sustainable development is not a fixed state of harmony, but a process of change
in which the exploitation of resources, the direction of investments, the orientation
of technological development and institutional change are made consistent with fu-
ture as well as present needs. Sustainable development is development that meets
the needs of the present generation without compromising the ability of future gener-
ations to meet their own needs.”
From then onwards in subsequent UN world conferences the definition has been
further refined.
Since then, this definition has been applied and adapted several times in UN
general assemblies. The UN general assembly of September 25th 2015, set up the Sus-
tainable Development Agenda for 2030 with the aim to end poverty, protect the planet,
and ensure prosperity for all as part of a new sustainable development agenda. This
time the World Business Counsel for Sustainable Development (WBCSD) was also
involved in setting this agenda and in defining the Sustainable Development Goals
(SDGs) for 2030. These SDGs are summarized in the Table 12.4 [16].
For a design evaluation, these 17 goals can be used. The question is then does the
design, when implemented, contribute to these SDGs. To make this practical for rapid
evaluations, the following questions can be used.
Evaluating a novel product, process or service on sustainable development crite-
ria means positively answering the questions:
– Does the novel product, process or service meet a need or requirement?
– Does it not harm the environment over the life cycle?
– Does it create prosperity over the value chain?
– Does it benefit societies over the value chain?
If the answer to any of these questions is negative, then that technology does not con-
tribute to sustainable development. If the answers are all positive, a comparison with
the commercial reference case should be made.
SDG number nine focusses on innovation. So academic research, contract re-
search, technology innovators, startups and technology centers of companies all can
place this SDG on their strategy.
Apart from this consensus on the definition of sustainable development there are
numerous other descriptions of sustainability, often only addressing the environmen-
tal aspect. These descriptions are of little use for companies and for education as they
lack global consensus and therefore have no authority.
Tab. 12.4: SDGs for 2030 [16].
Goal 1. End poverty in all its forms everywhere.

Goal 2. End hunger, achieve food security and improved nutrition and promote sustainable
agriculture.
Goal 3. Ensure healthy lives and promote wellbeing for all at all ages.
Goal 4. Ensure inclusive and equitable quality education and promote lifelong learning
opportunities for all.
Goal 5. Achieve gender equality and empower all women and girls.
Goal 6. Ensure availability and sustainable management of water and sanitation for all.
Goal 7. Ensure access to affordable, reliable, sustainable and modern energy for all.
Goal 8. Promote sustained, inclusive and sustainable economic growth, full and productive
employment and decent work for all.
Goal 9. Build resilient infrastructure, promote inclusive and sustainable industrialization
and foster innovation.
Goal 10. Reduce inequality within and among countries.
Goal 11. Make cities and human settlements inclusive, safe, resilient and sustainable.
Goal 12. Ensure sustainable consumption and production patterns.
Goal 13. Take urgent action to combat climate change and its impacts.*
Goal 14. Conserve and sustainably use the oceans, seas and marine resources for
sustainable development.
Goal 15. Protect, restore and promote sustainable use of terrestrial ecosystems, sustainably
manage forests, combat desertification, and halt and reverse land degradation and
halt biodiversity loss.
Goal 16. Promote peaceful and inclusive societies for sustainable development, provide
access to justice for all and build effective, accountable and inclusive institutions at
all levels.
Goal 17. Strengthen the means of implementation and revitalize the Global Partnership for
Sustainable Development.
A word of warning regarding the loose use of the word sustainability. In industrial-
ized countries, the meaning of the word is often reduced to living in harmony with
the environment. What I also have noticed is that due to the many other definitions of
sustainability, some people think that the word is meaningless or that they can make
their own definition and then work with that definition. It will be clear that this ap-
proach will bear little fruit as that definition has no common ground and has not been
reached by global consensus.
The only powerful term and definition of sustainable development, reached by
global consensus, is the one reached within the United Nations and which has been
accepted by the World Business Counsel of Sustainable Development (WBCSD).
Sustainable development hot spot scan

The sustainability hot spot scan is a comprehensive tool that provides insights in the
most important potential ‘hot spots’, i.e., risks and opportunities in sustainability in
a company’s existing and future worldwide value chains.
It is based on many databases from all over the world on safety, health and social
impacts and LCAs. The databases are from the UN, WBCSD, BASF social impact tool
SEEBALANCE, and many others. The scan provides information about which of the 17
SDGs are addressed by the company and where are the major areas of concern (hot
spots) are. It is an excel spreadsheet and can be downloaded for free from the MVO
Netherlands website [17].
The tool can be used in many ways. For example, it can be used to do a sensitiv-
ity analysis of design parameters on the hot spots. By simply changing the country of
origin of a feedstock, for instance, or by changing the choice of a solvent, sensitivity
analyses can be carried out and directions for improvement can be identified. It can
be used by the designer on its own, but also with others from marketing and legal de-
partments. The choice of feedstock and the related choice of countries can be quickly
evaluated. For example, it can give direct insight into the risk of feedstock produced
by child labor in a certain country. By changing the country and or feedstock type,
informed decisions regarding sustainable development can be made.
[1] Vogler J. The global commons, Chichester: J. Wiley & Sons, 2000.
[2] Sheldon RA. Atom efficiency and catalysis in organic synthesis, Pure Appl. Chem., 72, 2000,
1233–1246.
[3] Sheldon RA. The E factor: fifteen years on, Green Chem., 9, 2007, 1273–1283.
[4] Jonker G, Harmsen J. Engineering for sustainability: A practical guide for sustainable design,
Amsterdam: Elsevier, 2012.
[5] Guinee JB, et.al. Handbook on life cycle assessment. Operational guide to the ISO standards.
Dordrecht: Kluwer Academic Publishers, 2002.
[6] EPA. Protocol for Equipment Leak. Emission Estimates. Emission Standards, EPA-453/R-95-017.
1995. Sourced from: https://www3.epa.gov/ttnchie1/efdocs/equiplks.pdf, accessed on 25
January 2017.
[7] Madu C. Handbook of Environmentally Conscious Manufacturing, Berlin: Springer Science &
Business Media, 2012.
[8] Niederl-Schmidinger A, Narodoslawsky M. Life Cycle Assessment as an engineer’s tool? Journal
of Cleaner Production 16(2), 2008, 245–252.
[9] ISO. Environmental management – Life cycle assessment – Principles and framework, sources
30-1, 2017, sourced 26 January 2018 from: http://www.pqm-online.com/assets/files/lib/std/
iso_14040-2006.pdf.
[10] ILCD. General Guide to LCA, 2010. Sources 26 January 2018 from: http://publications.jrc.ec.
europa.eu/repository/bitstream/JRC48157/ilcd_handbook-general_guide_for_lca-detailed_
guidance_12march2010_isbn_fin.pdf.
[11] EPA. Methods for Estimating Air Emissions from chemical manufacturing plants, EIIP Vol.
II: Chapter 16 2007. Sourced 26 January 2018 from: https://search.epa.gov/epasearch/
epasearch?querytext=Methods+for+Estimating+Air+Emissions+from+chemical+
manufacturing+plants&areaname=&areacontacts=&areasearchurl=&typeofsearch=epa&
result_template=2col.ftl.
[12] Verkerk MJ, et.al. Philosophy of Technology, New York: Routledge, 2017.
[13] Potter D. Dimensions of society, Sevenoaks: Hodder and Stoughton, 2nd edn, 1978.
[14] De Soto H. The Mystery of Capital, London: Bantam Press, 2000.
[15] Brundtland GH. Our common future, report of the World Commission on Environment and De-
velopment (WCED), Oxford: Oxford University Press, 1987.
[16] UN. Transforming our world: the 2030 Agenda for Sustainable Development, Resolution
adopted by the General Assembly on 25 September 2015. Sourced 26 January 2018 from:
http://www.un.org/sustainabledevelopment/sustainable-development-goals/.
[17] MVO. Sustainability Hotspot Scan, 2017. Sourced 26 January 2018 from: https://mvonederland.
nl/publicatie/sustainability-hotspot-scan.
13 Communicating
13.1 Communicating: project team and stakeholders
A design project comprises many activities carried out by the team members and in in-
teraction with stakeholders. This was made clear in Chapter 4 and also becomes clear
from all other chapters. In this Chapter, we want to focus not only on the reporting of
the (final) design results, but also emphasize that ‘communicating’ is far more than
completing the final project documentation.
In order to support the communication during the project with project team mem-
bers and with the project’s stakeholders, a number of communication tools and tips
will be discussed. From experience, it is clear that in order to arrive at the best design
solution within resources (time, tools and designers) made available, communicating
about the design goal, approach, intermediate results, evaluation, decision making
and argumentation in the team and with stakeholders are crucial activities. Only by
sharing this information in an effective and efficient way, reflection and improvements
by team members’ and stakeholders’ feedback makes improving the design possible.
Based on our experience, project teams may spend up to 40% in communication ac-
tivities.
In this Chapter, we will link a number of key communication tools to the Delft
Design Map (DDM) (see Section 4.3).
13.2 Communicating using the Delft Design Map (DDM)
In the DDM (see Section 4.3) a ‘reporting’ step is included in all design levels (see
Table 13.1). This is done for a special purpose, and links to this Chapter’s introduction.
By introducing the ‘reporting’ step, we want to emphasize the need for capturing
the goal, approach, results, decisions and next steps in a series of design level reports.
It invites the project team to collect and report their findings at least at each design
level.
These reports also form the basis for, e.g., Stage-GateTM project review meetings
and decisions about the project’s next steps.
As explained in Section 8.1, ‘designing’ is ‘taking decisions’ about many different
aspects, and at many different levels to arrive at the optimal design. Reporting these
design results, design models (schematic, mathematical, verbal and sometimes even
physical/iconic models), the design decisions that were made, and the argumentation
why these were made is crucial.
Many design projects will not make it to full implementation, and will be stopped
in the Stage-GateTM process because of various business, technical and environmental
reasons. However, the knowledge gained from these projects is extremely valuable for
https://doi.org/10.1515/9783110467741-013
312 | 13 Communicating
Tab. 13.1: Delft Design Map (DDM) design levels, design steps, including the ‘reporting’ step (AR:
Activity Report(s), DLR: Design Level Report).
Framing
AR: Activity report(s) Supply chain Process Process Final
DLR: Design Level imbedding technology engineering
Report
Design levels ⇒ (PF) (CW- (PQ- (I-O) (SP) (TN) (UN) (PI) (ED) (OI) (FO) (Final)
PQ) PF)
⇓ Design steps/
activities
1 Scope AR AR AR AR AR AR AR AR AR AR AR AR
2 Knowledge AR AR AR AR AR AR AR AR AR AR AR AR
3 Synthesize AR AR AR AR AR AR AR AR AR AR AR AR
4 Analyze AR AR AR AR AR AR AR AR AR AR AR AR
5 Evaluate AR AR AR AR AR AR AR AR AR AR AR AR
6 Select AR AR AR AR AR AR AR AR AR AR AR AR
7 Report DLR DLR DLR DLR DLR DLR DLR DLR DLR DLR DLR DLR
the company’s decisions at that time and in the future. A well-documented design
project that includes what and for what reasons the design decisions were made, and
the learnings and future outlooks, is highly valuable. Business conditions can change
so that these design results and teachings can be brought to life again. With many team
members having moved on to other parts of the company (or to other companies),
it is only the documentation that remains. With this documentation, a new design
team can have a head start in tackling this project or, in absence of this historical
documentation, can be outperformed by competitors.
It should be stressed that only reporting design activities at the end of each de-
sign level is far from sufficient in the communication between the team members, and
with the stakeholders. These reports are a collection of information and decisions from
many more activities. In fact, in each of the design steps design activities take place
at each level that should be communicated.
A good working habit for a design team is to use a mix of reporting and communi-
cating tools. These can be used by each team member and by the project team in order
to support the communication.
In the next paragraphs the communication tools that are in use by MSc students
and PDEng trainees in the TU Delft design courses will be discussed:
– activity reports
– meetings: agenda and minutes of meeting (MOM)
– models (including brain writing and drawings)
– presentations
– reports
13.3 Activity reports | 313
13.3 Activity reports
Well organized and written project documentation is not only valuable at the end of
each design level and at completion of the entire project. During the project, reports
and other documentation are needed even more.
After project planning meetings (see Section 13.4), team members will start exe-
cuting their (individual or team) design activities. Reporting design activities in ‘ac-
tivity reports’ assists the designer(s) to first formulate the objective and deliverables
of the/their activity. Then he/she/they can choose and report the approach and re-
sources needed. Finally, results, discussions and teachings from the activity can be
reported. Typically, activities that take a couple of hours to a day will be concluded
with a concise activity report. A format for such a design activity report is shown in
Section Appendix A13.1).
The ‘activity reports’ greatly assist the designer to focus on activities’ targets in
relation to the overall design. Reporting the results forces the designer to think over
these, and gives time for alternative options or argumentation to surface. It also avoids
executing too many and possibly unnecessary activities, and thereby losing valuable
time. A proven guideline for designers to consider is: ‘if a design activity is not worth
reporting, it is not worth executing’. The creation and use of ‘schematic models’ like
figures, drawings and tables is highly recommended. The proper use of these models
cannot be stressed enough. Schematics and drawings take time to prepare, but are
far more effective in communicating design results and teachings than text. It is very
tempting to create schematic models that are very close/similar to information found
in other sources. It is much more productive to construct these schematic models to
what is required for your design, always focusing on your target(s). Some examples
can be mentioned here:
(1) House of Quality (HoQ) diagrams (see Section 4.3.3 and Section 8.3).
(2) Stream specifications table (streams crossing battery limit, see Table A13.1).
(3) Concepts/criteria decision matrix or Pugh matrix [1] (see Table A13.2):
(a) Do list your requirements first, and indicate per design option or technology
how each option fits in meeting these requirements.
(b) Do not make a long list of available technologies, or of design options and
compare their strengths/weaknesses (in general). Link it directly to your
project’s needs.
(4) Pure component properties table (see Table A13.3):
(a) Do make a list of components properties that are essential/useful for your de-
sign and fill out what is easily retrievable and what is not (or not yet retriev-
able).
(b) Do not make a list of pure components with only properties that are easily
retrievable.
The activity reports are not only valuable for the team member executing the activity.
They are equally valuable for the other team members, as they can keep abreast with
other activities than their own and they can do this at their own convenience. Read-
ing is much faster than listening to a team members’ presentations in team meetings.
Team members can come well prepared to team meetings, and these meetings can be
used for taking decisions on the design direction and future planning. This greatly
improves the information exchange and knowledge sharing.
The storage of the activity reports can be very well organized using the DDM’s de-
sign level/design step coding: reports can be coded from PQ–PF/Scope, all the way to
final report. At each design level the ‘report’ design step can be used to consolidate the
various activity reports in e.g., the design level report: ‘PQ–PF/report’ (see Table 4.3).
Reporting/documentation when done with the ‘target setting and knowledge de-
livering approach’, as explained above, communicates a clear result oriented mindset
in the project. In the DDM the target setting approach becomes already very apparent
in the first design level, ‘project framing’. Before starting the actual design work the
overall project goals, objectives, deliverables, etc., are determined and ‘framed’.
Reporting and documenting the results of this design level gives an ‘intermediate
result’ that can always be consulted by the team members. From here on, all team
members will adopt the approach of reporting while designing so that design activities
results become available to the team in a continuous workflow.
Last but not least, it is very good practice to decide very early on in the project
on the reporting tools/software to use. This ensures that all stakeholders (also the
ones outside the organization) are able to open the files, read the report(s) and provide
(digital) feedback. Of course, this is equally important for mathematical modeling and
simulation software, spreadsheets and presentation software.
13.4 Meetings: agenda and minutes of meeting (MOM)

Standard agenda, MOM structure (internal and external)
During a design project a multitude of meetings are organized. The design team will
schedule (internal) progress meetings for the team and (external) meetings with stake-
holders. Reports of these meetings (MOM) are special versions of activity reports. In
order to set the objective(s) of the meeting time planning is essential, as is the prepa-
ration of an agenda. Also, meeting rooms and other facilities (audio/video equipment
for presentations and access for remote audiences) need to be arranged. The devil is
in the detail as each hiccup in the availability or proper functioning of these items
will interfere with the meeting’s effectiveness. Meticulous care should be put in these
preparations, especially for the stakeholder meetings, and nothing should be taken
for granted.
The preparation and recording can be enhanced by using standard templates for
agenda and MOM (see Section A13.5).
13.6 Presentations | 315
13.5 Models
Models play a very important role in design as explained in Section 8.1. The verbal,
schematic, mathematical and physical models form the backbone of the design re-
sults and decisions. Preparation of the schematic, mathematical and physical models
and communication thereof are important tasks. These models should be prepared
and explained in such a way that they can be understood by stakeholders without
any additional extensive text. The designers need to give this sufficient thought and
attention, as misunderstanding about insufficient clear models will lead to confusion,
mistakes, delays and frustration.
Many models have been discussed in previous chapters. Some further examples
are:
– mindmap
– brain writing or brain drawing maps
– fishbone or Ishikawa diagrams
– process block scheme (PBS) (Figure 8.1)
– process flow sheet (PFS) (Figure 8.2)
– process stream summary (PSS) [2]
13.6 Presentations
During the course of the design project, the team not only prepares and delivers re-
ports, but also presentations. These presentations will be held at different phases
of the project, and for various audiences/stakeholders. The design team should be
aware that their presentations should always be prepared with their audiences in
mind. Also the limitations of presentations and their effectiveness in conveying in-
formation, knowledge and messages to the audience should be known to the design
team members.
The goal of the presentation to the audience should be clear. During design
projects the presentations at the key project milestone meetings are to inform the
audience about project progress, design decisions made, proposals about how to
continue the project and to seek stakeholders’ approval and feedback.
The main ‘storyline’ can be supported (but not overshadowed) by supportive in-
formation. The relationship between design results, decisions and the overall project
context and objectives should always stand out. Supportive information, mostly on
design options analysis and evaluation, comes second: it is needed but not at the ex-
pense of omitting the main storyline.
The discussions and feedback of the stakeholders should be promoted, as this in-
formation is crucial for the next phases. This implies that the information presented
and discussed should be easily understood by the audience, so the exchange of ideas
is realized. Recording the discussions and feedback in MOM (Section 13.5) is the main
deliverable of most presentations/meetings. The next section provides further guide-

lines for preparing and delivering presentations.
13.6.1 Quality checks: FOOFI list for presentations
At regular intervals, formal presentations are given to the design project’s stakehold-
ers. The preparation and delivery of these presentation should not be underestimated.
The structure, content and layout of the slides need particular attention. Audiences
have a limited concentration span. The presentation’s content should be very clear,
and the audience should be focused on the main ‘storyline’ and message. The stake-
holders need to be taken from a helicopter view of the project’s objective to the re-
sults, conclusions and recommendations. The detailed information is only supportive
of this.
Over the years, a quality checklist for design project presentations has been com-
posed by P. Swinkels while supervising and coaching MSc and PDEng trainee design
project teams. This list has been named FOOFI, which translates to ‘frequently occur-
ring opportunities for improvements’. FOOFI is the phonetic equivalent of (the Ams-
terdam) slang word for the Dutch ‘foefje’, meaning ‘knack or agility’.
This FOOFI list for presentations is presented in Section Appendix A13.6. A sim-
ilar list is also available for reporting quality checks (see Section 13.7.3 and Section
Appendix A13.8).
13.7 Reporting in stage gate reviews
In the following sections the contents of the consolidated reports for the ‘concept and
feasibility stages’ are presented.
13.7.1 Concept stage
For the ‘concept’ stage gate, a consolidated report is prepared for the concept ‘stage
gate’ review. This consolidated report will be composed of selected contents from the
consolidated reports at each of the design levels:
– framing
– product concept/consumer wants
– product concept/product function
– input/output, and sub processes
Often, this document is also called ‘concept stage report’, or ‘basis of design’ report. It
is important that, at this stage, all key data for the design, as agreed by the design team
and client/stakeholders, are tabulated and provided with background information.
As can been seen from the previous chapters, the first design levels concern the
consumer needs, requirements and specifications, possible product compositions and
structures and possible processing technologies using various feed streams, etc.
Not only the specifications of the product, but also of (saleable) byproducts, com-
mercially available feed streams and utilities, and last but not least any environmental
limits on gaseous, liquid and solid waste streams should be clearly specified (for an
example, see Table A13.1). At TU Delft MSc and PDEng level design projects a report
structure for this report is suggested, and standardized tables and formats are pro-
posed (see Table 13.2 for product focused design and Table 13.3 for a process focused
design). Further information on this topic can also be found in [3].
Tab. 13.2: Suggested concept stage report structure and contents: product design.
Section Contents Topics

(DDM levels:
Table 4.3)
1 Description of the Project brief/charter: background; framing (BOSCARD –

PF design; background, objectives, scope, constraints, assumptions,
project incen- risks, and deliverables); stakeholder needs; design type and
tive (opportunity driver (market pull, technology push, innovation map).
assessment) Market assessment (customers, markets, competitors,
‘SHEETS’ (safety, health, environment, economic, technically
feasible, and social criteria) based opportunity definition);
market segmentation; value chain analysis.
2 Design approach Design challenges, design methodology, creativity methods,
PF group methods and tools (SWOT – strength, weakness,
opportunities, and threats); project and time planning.
3 Consumer needs Stakeholder/consumer needs (voice of customer); product
CW–PQ and requirements; quality requirements in relation to product application and
project opportu- entire life cycle.
nity assessment
4 Product require- House of quality (HoQ) (composed by quality function
PQ–PF ments deployment (QFD), relating consumer needs, product quality
performance, product property function and manufacturing
requirements.
5 Database creation Literature and patent search; product application process
PF; CW–PQ; conditions and knowledge; chemical components; materials;
PQ–PF pure component properties (physical, chemical, SHE (safety,
health, and environmental criteria), price); formulation and
physical/chemical models and data; thermodynamic models
and data; stoichiometry.

(DDM levels:
Table 4.3)
6 Preliminary product Based on the product application process’

CW–PQ; concepts synthesis understanding and functional requirements, generate
PQ–PF product concepts – both top down idea generation, and
bottom up, based on molecular identity, product (nano)
and microstructures, combined into a
product/application process combination.
7 Product concepts and Product concept/criteria evaluation/selection; chosen
CW–PQ; preliminary process superior product concept (components, structure,
PQ–PF; synthesis physical/chemical product properties, in use
I-O; SP; TN; performance).
UN Based on product structuring/manufacturing technology
choices, stoichiometry, yields and split factors,
determine: plant capacity and location; battery limit;
available utilities; other design determining
assumptions; quantitative Input/output diagrams;
(sub)process block diagrams; stream summaries.
8 Preliminary business Product: performance versus requirements in use;
I-O; SP case (product and pro- product and process: performance verses ‘SHEETS’:
cess design) ‘SHEETS’ safety (‘Bowtie’), health, environment, economy (gross
analysis margin), technology, social analysis using concept stage
knowledge.
9 Planning review and Review project and time planning, and use lessons
PF update learned to improve future project and time plan; work
allocation between team members.
10 Creativity and group Review the team’s creativity and group performance.
PF process tools Plan any improvements for the remainder of the project.
11 Conclusions and rec- At completion of the concept phase and feed forward to
PF to UN ommendations the subsequent phases.
List of symbols
Literature
Appendices
Tab. 13.3: Suggested concept stage report structure and contents: process design.

(DDM levels:
Table 4.3)
1 Description of the Project brief/charter: project background; framing

PF design project in- (BOSCARD); stakeholder needs; customer requirements;
centive (opportu- technical requirements.
nity assessment) Market assessment (customers, markets, competitors;
‘SHEETS’ based opportunity definition.
PF group methods and tools (SWOT); project and time planning.
3 Database creation Literature and patent search; chemical components;
PF; CW–PQ; materials; pure component properties (physical, chemical,
PQ–PF; I-O; SHE, price) thermodynamics; stoichiometry.
SP
4 Preliminary pro- Based on technology choice, stoichiometry, reaction yields
I-O; SP; TN; cess concepts and separation split factors, determine:
UN synthesis Plant capacity and location; battery limit; available utilities;
other design determining assumptions; quantitative
input/output diagrams; (sub)process block diagrams; stream
summaries.
5 Preliminary Business case analysis based on ‘SHEETS’: safety (‘Bowtie’),
I-O; SP; TN; business case health, environment, economy (gross margin), technology,
UN ‘SHEETS’ analysis social analysis using concept stage knowledge.
6 Planning review Review project and time planning, and use lessons learned to
PF and update improve future project and time plan; work allocation
between team members.
7 Creativity and Review the team’s creativity and group performance. Plan any
PF group process improvements for the remainder of the project.
tools
8 Conclusions and At completion of the concept phase and feed forward to the
PF to UN recommendations subsequent phases.
List of symbols
Literature
Appendices
13.7.2 Feasibility and development stages
A suggested report structure and contents for the feasibility and development stage
reports can be found in Tables 13.4 (product focus) and 13.5 (process focus). The con-
cept stage report is further extended. Parts of the concept stage report becomes part
of the feasibility and development stage report.
Tab. 13.4: Suggested feasibility and development stages report structure and contents: product
design focus.

(DDM levels:
Table 4.3)
Summary In not more than one page the highlights of the project
(objective, approach, results, including quantification of the
key design results and performance criteria), conclusions
and recommendations should be mentioned.
Table of contents Use descriptive chapter and section titles.
1 Description of the Project brief/charter: background; framing (BOSCARD);
PF design; stakeholder needs; design type and driver (market pull,
project incen- technology push, innovation map).
tive (opportunity Market assessment (customers, markets, competitors,
assessment). ‘SHEETS’ based opportunity definition); market
segmentation; value chain analysis.
3 Consumer needs Stakeholder/consumer needs (voice of customer); product
CW–PQ and requirements; quality requirements in relation to product application and
project opportu- entire life cycle.
nity assessment
4 Product require- HoQ (composed by QFD), relating consumer needs, product
PQ–PF ments quality performance, product property function and
manufacturing requirements.
5 Database creation Literature and patent search; product application process
PF; CW–PQ; conditions and knowledge; chemical components; materials;
PQ–PF pure component properties (physical, chemical, SHE, price);
formulation and physical/chemical models and data;
thermodynamic models and data; stoichiometry.
6 Preliminary prod- Based on product application requirements and
CW–PQ; uct concepts syn- understanding of application process, generate product
PQ–PF thesis concepts – both top down idea generation, and bottom up,
based on molecular identity, product (nano) and
microstructures, combined into a product/application
process combination.

(DDM levels:
Table 4.3)
7 Product concepts Product concept/criteria evaluation/selection; chosen

CW–PQ; and preliminary superior product concept (components, structure,
PQ–PF; process synthesis physical/chemical product properties, in use performance).
I-O; SP; TN; Based on product structuring/manufacturing technology
UN choices, stoichiometry, yields and split factors, determine:
plant capacity and location; battery limit; available utilities;
summaries.
8 Preliminary busi- Product: performance versus requirements in use;
I-O; SP ness case (prod- product and process: performance versus. ‘SHEETS’: safety
uct and process (‘Bowtie’), health, environment, economy (gross margin),
design) ‘SHEETS’ technology, social analysis using concept stage knowledge
analysis
PF improve future project and time plan; specify work allocation
tools
11 Detailed product Superior product concepts (concept stage) are further
CW–PQ; design and vali- designed in detail. Product performance validation (by
PQ–PF; dation; product/ product modeling, product testing and product prototyping)
I-O; SP; TN; process interac- is carried out.
UN tion The design of the manufacturing process, the processing
equipment and their interactions with the product physical
properties is further modeled, tested and validated.
12 Detailed business Product: performance verses requirements in use;
case evaluation Product and process: performance verses ‘SHEETS’: safety
(‘Bowtie’), health, environment, economy (gross margin),
technology, social analysis using concept stage knowledge
13 Conclusions and Based on the design knowledge gathered at this phase,
PF to Final recommendations present the strengths and weaknesses of the
product/process design (in relation to the needs and
requirements). Depict the major risks (technical, financial,
etc.) and their consequences in failing to achieve the project
objectives. Suggestions are given how to mitigate/eliminate
these risks, and what should be the next steps.
List of symbols
Literature
Appendices
Tab. 13.5: Suggested feasibility and development stage report structure and contents: process de-
sign focus.

(DDM levels:
Table 4.3)
Summary In not more than one page the highlights of the project
(objective, approach, results (including quantification of the
key design results and performance criteria), conclusions
and recommendations should be mentioned.
Table of contents Use descriptive chapter and section titles.
1 Description of the Project brief/charter: project background; framing
PF design project in- (BOSCARD); stakeholder needs; customer requirements;
centive (opportu- technical requirements.
nity assessment) Market assessment (customers, markets, competitors;
‘SHEETS’ based opportunity definition.
3 Database creation Literature and patent search; chemical components;
PF; CW–PQ; materials; pure component properties (physical, chemical,
PQ–PF; I-O; SHE, price) thermodynamics; stoichiometry.
SP
4 Preliminary pro- Based on technology choice, stoichiometry, reaction yields
I-O; SP; TN; cess concepts and separation split factors, determine:
UN synthesis plant capacity and location; battery limit; available utilities;
summaries.
5 Preliminary Business case analysis based on ‘SHEETS’: safety (‘Bowtie’),
I-O; SP; TN; business case health, environment, economy (gross margin), technology,
UN ‘SHEETS’ analysis social analysis using concept stage knowledge.
PF improve future project and time plan; specify work allocation
tools
8 Detailed and in- Superior process concepts from the preliminary process
I-O; SP; TN; tegrated process concepts evaluation are designed, developed in further
UN; PI synthesis detail, and analyzed and evaluated (SHEETS) in the task
network and unit network levels, leading to (a) superior
process flow scheme(s) (PFS), process stream summary
(PSS), and mass and energy balances. Mass and heat
integration are applied in the process integration level.
Utility requirements and product yields are discussed.
SHEETS performances are reassessed

(DDM levels:
Table 4.3)
based on the more detailed analysis. SHEETS performances

are reported in subsequent sections.
The superior PFS is described in detail using the PFS and
PSS. Design choices and possible alternatives are described
in relation to the SHEETS performance.
9 Equipment sum- Based on the functional specification of the equipment in the
ED mary lists and PFS, the equipment design variables are determined and
equipment unit sized to fulfill these specifications. It is especially important
design specifica- to identify the key equipment design variables (sizes,
tion sheets internals, material of construction, etc.), and how values are
assigned to these variables for the best performance.
See Section A13.7 for equipment summary lists and
equipment specification sheets.
10 SHE and S Safety, health, environment and sustainable development
I-O; SP; TN; detailed evaluation (HAZOP, DOW’s Fire and Explosion Index
UN; PI; (F&EI).
11 Economic evalua- Detailed economic analysis: in addition to gross margin at
tion this stage, also equipment costs, total capital investment
and operating costs can be calculated and economic
performance criteria – rate of return (ROR), payout time
(POT), discounted cash flow rate of return (DCFROR), net
present value (NPV) – can be determined (see Chapter 10).
12 Conclusions and Based on the design knowledge gathered at this phase,
PF to Final recommendations present the strengths and weaknesses of the design (in
relation to the needs and requirements). Depict the major
risks (technical, financial, etc.) and their consequences in
failing to achieve the project objectives. Suggestions on how
to mitigate/eliminate these risks, and what should be the
next steps are given.
List of symbols
Literature
Appendices
13.7.3 Quality checks: FOOFI list for reports
At major design project Stage GateTM milestones, design reports are prepared for the
stakeholders by the design team. In the previous section the contents for the concept,
feasibility and development milestones were presented. In this section guidelines are
provided on ways to improve the quality of how the information is presented, taking
into account the different roles of the readers of the report. What many (junior) de-
signers do not appreciate is that very few readers will read the complete report (except
university lecturers in design courses). Different stakeholders are interested in differ-
ent parts of the report, and need to be able to navigate efficiently through the report
and also to absorb and understand the information they are after with minimal effort.
This means that, in order of priority, the following parts of the report need to be self-
explanatory and understandable as a standalone text:
– summary
– table of contents
– should contain descriptive text about content
– project objectives
– conclusions and recommendations
– figures and tables, especially of the design requirements, and design results:
– stakeholder needs, project and product and process (performances) require-
ments
– stream tables
– process block scheme and PFS
– etc.
Attention to figures and tables is compulsory, as (scanning) readers tend to focus on

these report components first. Figure and table titles together with the content should
be sufficient for the reader to understand without reading the main text. The text
surrounding figures and tables also receives attention by the scanning reader, so it
makes sense to put important information in these areas. The readers should be able
to quickly retrieve the main storyline first, and then focus on the domain knowledge
that is important for their role(s), e.g., related to the SHEETS performance criteria.
These detailed chapters should be clear and make sense when read in isolation. One
should include key information from other chapters that is crucial to understand the
individual chapters.
As already mentioned in section 13.6.1, a quality checklist for design project re-
ports has been composed (see Section Appendix A13.8).
Abbreviations
BOSCARD Background, objectives, scope, constraints, assumptions, reporting,
deliverables
CW–PQ Consumer wants – product quality (performance function) (DDM)
DCFROR Discounted cash flow rate of return

DOW’s F&EI DOW’s Fire and Explosion Index
ED Equipment design (DDM)
FO Flow sheet sensitivity and optimization (DDM)
FOOFI Frequently occurring opportunities for improvement
HAZOP Hazard and operability study
HoQ House of quality
I-O Input-output structure (DDM)
MOM Minutes of meeting
NPV Net present value
OI Operability integration (OI)
PBS Process block scheme
PF Project framing (DDM)
PFS Process flow scheme
PI Process integration (DDM)
P&ID Piping and instrumentation diagram
POT Payout time
PQ–PF Product quality (performance function) – (product) property
function (DDM)
QFD Quality function deployment
ROR Rate of return
SHE Safety health environment
SHEETS Safety health environment economy technology social
S Sustainable development
SP Sub processes (DDM)
SWOT Strengths weaknesses opportunities threats
TN Task network (DDM)
TOC Table of contents
UN Unit network (DDM)
[1] Pugh S. Creating innovative products using total design, New York, NY, USA, Addison-Wesley-
Longman, 1996.
[2] De Haan AB, Swinkels PLJ, de Koning PJ. Instruction manual conceptual design project CH3843 –
from idea to design. Department of Chemical Engineering, Delft University of Technology, Delft,
Netherlands, 2016.
[3] Seider WD, Seader JD, Lewin DR, Widagdo S. Process and Process Design Principles: Synthesis,
Analysis and Evaluation, 3rd edn Hoboken, NJ, USA, John Wiley & Sons Inc., 2010, 681–692.
|
Part D: Education
14 Education
In this Chapter, the authors give an overview of the BSc, and truly international MSc,
PhD and PDEng programs at TU Delft in the field of (bio)chemical engineering, and
with a focus on the interaction between education, research and design. The coopera-
tion with industry and public/private partnerships and the valorization of knowledge
in support of innovation is presented from a ‘systems point of view’. We hope this
forms an inspiration for other universities and countries to adopt a similar approach.
14.1 (Bio)chemical design education: a long history
The authors of this book have built their diverse product and process design and inno-
vation skills through our education at the Universities of Technologies in Delft, Eind-
hoven and Twente, through our industrial careers at various companies (Corbion, Co-
sun, DSM, ICI, Shell, Unilever). We have nearly two decades of experience in teaching
product and process design at Delft University of Technology (TU Delft), and also at
the Technical University Eindhoven (TU/e), University of Twente (UT) and Rijksuniver-
siteit Groningen (RUG). Inspired by our predecessors, and also having worked closely
with many of them in the last decades, we also would like to mention them in this
chapter on education. At TU Delft, the process design courses, including a large open
ended capstone design project, combined with the setup of a computer simulation
course and facilities, have been in existence since the mid 1980s. As a special recogni-
tion for their drive to set up and build a world class design course including projects,
we would like to mention, the predecessors and current colecturers in curricula of
(bio)chemical product and process design at BSc, MSc, and Professional Doctorate in
Engineering (PDEng) (see Table 14.1).
In this Chapter, the developments in product and process design education in the
(bio)chemical engineering discipline, and the fit in the TU Delft BSc, MSc and PDEng
programs are described.
14.2 Education programs
Since the year 2000, higher education at the Technical Universities in The Netherlands
has quickly converted to the Bologna 1st/2nd and 3rd cycles of education (Figure 14.1):
– 1st cycle: BSc: three years (180 European Credit Transfer System (ECTS) credits
– 2nd cycle: MSc: two years (120 ECTS credits)
– 3rd cycle: PhD: four years, research orientation
– 3rd cycle: PDEng: two years (120 ECTS credits), design orientation
https://doi.org/10.1515/9783110467741-014
330 | 14 Education
Tab. 14.1: TU Delft lecturers in (bio)chemical product and process design since the mid 1980s – pre-
decessors and current lecturers.
Predecessors in lecturing (bio)chemical product Current (co)lecturers in (bio)chemical product

and process design at TU Delft: and process design:
Em.Prof.ir. A.G. Montfoort Ir.drs. G. Bierman PDEng

Dr. F.A. Meijer Dr.ir. G.M.H. Meesters
Em.Prof.dr.ir. J. de Graauw Dr.ir. H.W. Nugteren
Em.Prof.ir. J. Grievink Dr. P.J. Hamersma
Dr.ir. P.J. Verheijen Dr. A.J. Houtepen
Prof.ir. K.Ch.A.M. Luyben Dr.ir. J.L.B. van Reisen PDEng
Dr.ir. C. van Leeuwen Dr.ir. P.J. Daudey
Dr.ir. J.C. Goebel Dr. A.J. Groenendijk
Ing. M. Krijnen Prof.dr.ir. L.A.M. van der Wielen
Ing. M. Valentijn Dr.ir. A.J.J. Straathof
Em.Prof.dr. G. Frens Dr.ir. M. Ottens PDEng
Em.Prof.dr.ir. P.W. Appel Prof.dr.ir. G.J. Witkamp
Prof.dr.-ing. J. Gross Dr.ir. M.C. Cuellar Soares PDEng
Prof.dr.ir. P.J. Jansens Prof.dr.ir. H.J. Noorman
Ir. C.P. Luteijn Prof.dr.ir. M.C.M. van Loosdrecht
Ir. J. Dijk A.M. Echavarria Alvarez PDEng
Em.Prof.dr.ir. K. van ’t Riet Dr.ir. L.D.M. van den Brekel
Dr.ir. J.F. Jacobs Prof.ir. G.J. Harmsen
Prof.dr.ir. J.J. Heijnen Prof.dr.ir. A.B. de Haan
Ir. M.W. Lambrichts Ir. P.L.J. Swinkels
Dutch university education:

3 Bologna cycles
4 years
PhD
2 years
PDEng
4 years 2 years
2 years
Master of Science
2 years
3 years
Bachelor of
Science
3 years
university industry
Fig. 14.1: Three cycles of higher education at the Dutch universities of technology (1st (BSc), 2nd
(MSc) and 3rd (PhD or PDEng: Professional Doctorate in Engineering) cycle, based on the Bologna
declaration.
The three cycles, based on the Bologna declaration, provide the opportunity to stu-
dents to complete different parts of their studies at different universities within differ-
ent countries. In addition to the European Erasmus program (European Region Action
Scheme for Mobility of University Students), this is stimulating the international mo-
bility of students between European countries. It has been named the largest success
of the European Union.
The connectivity and curricula of the BSc, MSc and the third cycle programs (PhD
or PDEng) will be discussed with a special focus on the design content.
14.2.1 BSc programs: TU Delft
The TU Delft BSc programs ‘Molecular Science & Technology (MST)’ and ‘Life Science
& Technology (LST)’ are offered in cooperation with Leiden University. Students attend
courses and take exams at both universities and, at graduation, earn a joint degree
from TU Delft and Leiden University. This cooperation was started over ten years ago
to attract more students to these education fields. It allowed high school graduates
to delay their choice to the 2nd year for the more science related chemistry or life sci-
ence direction in Leiden, or the more technology oriented direction at TU Delft. This
has been a very large success. Student numbers rose quickly to larger values than the
separate bachelor programs had before. TU Delft saw even a substantial increase after
this cooperation. Being exposed to technology oriented lectures and research group
research activities made more students choose to continue their BSc education at TU
Delft. The same holds for the biotechnology focused BSc program, LST.
The MST and LST BSc curricula have a study load of 180 ECTS credits, are full
time and take nominally three years to complete. The program is held in Dutch, but
textbooks are mostly in English. The MST program combines chemistry and (chemi-
cal) technology. The LST program combines life sciences (cell biology and chemical
processes) with biotechnology. Students are introduced to the specializations within
chemistry and chemical technology (MST) and life sciences and biotechnology (LST)
listed below. Students also gain ample laboratory experience in the very well equipped
laboratories in Delft and Leiden. The exposure to research work and to the staff in the
research sections starts early in the first year. The experience shows that this boosts
the students’ understanding of, and interest in, the field. The application of knowl-
edge and research questions inspires the students. It greatly helps their decision mak-
ing process about whether to continue in the chemistry (MST) and life sciences (LST)
field at Leiden University or in the technology directions at TU Delft.
The BSc-MST chemistry track focuses in more depth at disciplines like inorganic,
theoretical physical, organic and molecular chemistry. The BSc-MST technology track
covers a wider range of mathematics, numerical techniques, thermodynamics, physi-
cal transport phenomena, chemical reaction engineering, separation technology and
a (chemical) product design project [1]. As mentioned before, building competencies
332 | 14 Education
in experimentation skills and participating in scientific research projects are impor-

tant in both tracks. In the 3rd BSc program year a more in depth or a broad minor can
be selected and the BSc final research project completes the curriculum.
The BSc-LST curriculum [2] is also run at Leiden and Delft, but all students follow
the same program in the first two years. The more fundamental topics and laboratory
projects (cell biology, biochemistry, molecular cell biology and immunology, micro-
bial physiology and molecular genetics) are covered in Leiden, and the more biotech-
nology topics (mathematics, statistics, thermodynamics, transport phenomena and
biotechnology process design project) are taught in Delft. An introductory course on
the design of sustainable biotechnology processes is a key design project in year two.
The choice of the third year direction (minor and BSc final research project) depends
on the MSc choice. MSc-LST in Delft requires the third year BSc to be completed in
more technology related topics at TU Delft. The MSc-LST in Leiden requires the third
year BSc to be completed at Leiden University.
14.2.2 MSc programs: TU Delft
The MSc programs, Chemical Engineering (CE) and Life Science & Technology (LST),
are truly international programs and fully taught in English. The fraction of inter-
national students joining this program is currently about 30%. This number is ris-
ing, as is the total number of MSc students. The very high rankings of the TU Delft
(bio)chemical engineering field in various indices like QS World University Rankings,
and Shanghai Ranking, supported by the excellent research and teaching facilities,
have a large attraction on international talent. Both MSc programs cover 120 ECTS
credits and take two years to complete.
The MSc Chemical Engineering has three tracks: chemical product engineering,
process engineering, and nuclear science and engineering. The process engineering
track is traditionally the largest, but the other two tracks are steadily growing. Focus of
the product engineering track is on the design and synthesis of products, ranging from
new construction materials, materials for energy conversion & storage to pharmaceu-
ticals. The nuclear science and engineering track focuses on nuclear engineering and
its applications making use of radioisotopes. The TU Delft nuclear reactor is a key fa-
cility for these radioisotopes for, e.g., medical applications, and the field is developing
rapidly into other health and industrial applications.
The MSc LST has three tracks: biocatalysis, biochemical engineering, and cell fac-
tory. These two MSc programs comprise of deepening domain knowledge, executing
a large design project, an internship placement and an individual research project.
Top talent students in these MSc programs can apply for the Bologna third cycle
of education: PhD (Section 14.2.3) or PDEng positions (Section 14.2.4). MSc students
with exceptional academic performance are invited after the first two quarters of the
MSc program to participate in the ‘honors’ program. During this program the students
select an (additional) 20 ECTS credits. This program should be completed within the
two year MSc program. ‘Honors’ specializations can have a ‘academic research focus’
(preparing for a PhD position), a ‘design focus’ (preparing for a PDEng position) or an
‘industry focus’ with a research project with an industrial partner.
14.2.3 PhD programs: TU Delft
Research in the area of (bio)chemical engineering at TU Delft is clustered in the ‘Delft

Process Technology Institute’ (DPTI) [3]. DPTI is a partnership and encompasses 19 re-
search groups from the departments Biotechnology, Chemical Engineering, Process
and Energy and Energy Technology (Figure 14.2). DPTI’s research, education and de-
sign activities are focused on three main areas:
– biochemical process engineering
– process intensification and information intensification
– technologies for advanced materials
Crystallization
Energy storage
Energy technology
Engineering
thermodynamics
TU Delft Process Technology Institute (DPTI) Intensified reaction &
Multidisciplinary collaboration in 3 areas separation systems
Solids processing
Separation technologies
Process & information intensification
Bioprocess engineering
Bioseparation technology
Environmental biotechnology
Biochemical process engineering Equipment design
Process systems
modeling & design
Water technology
Technologies for advanced materials
Advanced materials
Catalysis & reaction
engineering
Fluid mechanics
Materials for energy
conversion & storage
Product, process design
& engineering
Transport phenomena
Fig. 14.2: Overview of the three key research areas of the TU Delft Process Technology Institute
(DPTI), based on [3].
334 | 14 Education
DPTI aims for aligning the activities in the research groups, and intensifying collabo-
ration with other (inter)national universities and industries. DPTI offers industry and
knowledge institutions a wide variety of partnership opportunities. It ranges from BSc
or MSc internships to participation in large international consortia for PhD and PDEng
projects.
DPTI also has a strong impact on the visibility and impact of its research and ed-
ucation. Over the last few years, the TU Delft in the Chemical Engineering subject al-
ways reached top fifteen ranks in the QS World University Ranking [4].
In general, the PhD research tracks are four year projects, funded by national, Eu-
ropean and worldwide initiatives. PhD tracks are mostly part of larger projects with
many partners in public/private partnerships (PPP). Top performing MSc graduates
are hired as PhD students after a careful selection process. The applicants’ educational
performance, research, and personal competences and motivation are tested in inter-
views. Once hired, the PhD student will soon setup a personal development plan in
addition to their research project plan. PhD students select several personal develop-
ment courses organized by the TU Delft’s Graduate School. After one year, a ‘go/no go’
presentation and assessment is held. After the ‘go ahead’ decision, PhD students are
expected to complete their research project with minimum delay. The PhD students’
progress in executing their personal development plan, research project and realizing
publications is monitored by their scientific supervisors. The Graduate School organi-
zation aids in recording said plans, execution progress, and in advising PhD students
and supervisors during this process.
14.2.4 PDEng programs: TU Delft
During more than thirty years, PDEng programs are a very successful cooperation be-
tween the Dutch technical universities and industry. Because of its unique setup, and
the fact that these programs hardly exist in other countries, a rather detailed overview
is presented here.
The text below is largely cited from Swinkels [5], who has given a concise descrip-
tion on the history, nature and current situation of the over thirty year old post Master
PDEng traineeships at the Dutch technical universities. In some places, the text has
been adapted to the current (2018) situation. At the end of this section, the contents of
the TU Delft PDEng traineeship programs in the (bio)chemical product/process design
fields are presented.
“The three universities in The Netherlands – Delft University of Technology, Eind-
hoven University of Technology and University of Twente – offer two year post MSc
technological design programs (traineeships) to selected top level MSc graduates re-
cruited from all over the globe. The Dutch Government and the Dutch industries jointly
initiated these traineeships in 1986, with the aim to train top talent in design compe-
tencies, and in business and personal skills, to prepare them as key ‘knowledge work-
ers’ who strongly contribute to the design of new products and processes, and can
boost innovation. The PDEng traineeships are designed to train and educate top MSc
graduates to become qualified excellent designers capable of designing ‘fit for pur-
pose’ and ‘first of its kind’ products, processes and systems. The PDEng trainees are
encouraged to actively look beyond the perimeters of their own disciplines and rec-
ognize the challenges and restrictions imposed by product chain management, time
and money.”
“Over a period of thirty years many of these programs are still going strong – sup-
ported by long lasting and multinational industrial partners who actively and contin-
uously pursuing (open) innovation and cooperation with the technical universities.
Other programs were terminated, and yet new ones matching with emerging innova-
tion areas were started. This lifelike dynamic very closely follows industrial demand
and opportunities for innovation, the demand for technical innovation talent and the
universities’ research, design and education capabilities and strengths. Contribution
from, and participation by these stakeholders in innovation, is a prerequisite for every
successful PDEng program. Regular contacts and active cooperation projects between
the industrial partners, universities and PDEng trainees prove to be essential in this
process.”
“In the late 1990s the Dutch Universities started to internationalize their MSc cur-
ricula, and the PDEng (and PhD) recruitment followed suit. Currently over 90% of the
PDEng trainees originate from outside the Netherlands, and both genders are equally
represented. The PDEng traineeships have turned into truly global and gender bal-
anced doctoral training centers. This has increased the attractiveness of the trainee-
ships for the industry, university and last but not least for the PDEng trainees them-
selves.”
“The position of the PDEng programs in the Dutch technical university education
system is shown in Figure 14.3. It builds on the first (BSc) and second (MSc) cycles of
the Bologna agreement, and forms together with the research oriented PhD track, the
third tier of academic education. The design oriented PDEng education reaches out
into the industry domain, as after one year of competence building and knowledge
acquisition, the full second year is spent on an individual design project initiated by
and executed (mostly) at the industrial partner’s site.”
“On successfully completing the PDEng program, the graduates receive a certified
diploma and are entitled to use the academic degree ‘Professional Doctorate in Engi-
neering (PDEng)’. All diplomas are registered in a central register kept by the Royal
Dutch Society of Engineers (KIVI: ‘Koninlijk Instituut Van Ingenieurs’).”
“Since 2006 the PDEng designer programs have been coordinated under the flag
of the 3TU.School for Technological Design – Stan Ackermans Institute (3TU.SAI), a
cooperation between the three technical universities in the Netherlands (Delft, Eind-
hoven, Twente). Currently also Wageningen University & Research (WUR) has joined
the now called 4TU.SAI cooperation.”
336 | 14 Education
Professional Doctorate in Engineering

Post-MSc designer programmes
1st year:
• Design methods & tools
• Deepening &
broadening
• Socioeconomic aspects
PhD • Professional skills
4 years
• Group design project

4 years 2nd year:
PDEng • Industrial individual
2 years design project
Highlights:
2 years
Master of Science • Top 25% international

2 years talent
• Female/male ratio: 1/1
Bachelor of • >95% graduate without
3 years
Science delay
3 years • Academic & industrial
network
• Innovation savvy
university industry
Fig. 14.3: Position of PDEng traineeships at TU Delft (bio)chemical engineering fields and specifics
regarding graduation efficiency and gender balance in the Dutch technical universities’ education
system.
“Nowadays between 200–250 trainees join – after a stringent recruitment and selec-
tion process – one of the 20 PDEng programs in the various disciplines every year.
They work as a salaried employee (PDEng trainee) for one of the three technical uni-
versities, on strengthening their skills through workshops and design cases, and on
solving a real design case for (and at) an industrial partner during the full second year
of their employment.”
“On October 1st 2014 TU Eindhoven Rector Magnificus Prof.dr.ir. C.J. van Duijn
awarded the 3500th PDEng (Professional Doctorate in Engineering).” On October 6th
2017, the 4000th PDEng graduate was celebrated.
“It should be noted that over 80% of the 3500 PDEng graduates originate from
countries outside the Netherlands, and that the female/male ratio has steadily in-
creased to over 50%. More than 75% (and rising) of the PDEng graduates find tech-
nological challenging and permanent jobs in the Netherlands. From a recent alumni
overview from the (then) 23 year old program, Process & Equipment Design (TU Delft),
it was also evident that the large majority of these alumni progress their industrial ca-
reers in the technology or technology management career track. This is a very positive
observation, as traditionally many (MSc) graduates from Dutch universities move into
operation management, financial or general management roles rather soon in their

careers, and stop contributing directly to technological innovations. As their business
and financial counterparts don’t make the reverse career switch, this loss of technical
talent could easily become a threat. PDEng graduates seem to very naturally and ef-
fectively occupy these positions, and keep pursuing a technical career track for a very
significant part of their careers. Their technological design talent and ambitions leads
and keeps them in this high in demand career track.”
“Training PDEng designers parallel to educating PhD researchers is an approach
that has catalyzed and delivered many innovation benefits to the industries based in
the Netherlands. With vast worldwide challenges ahead of us in the areas of energy,
food, water, scarce resources and health, it is an example that could be adopted world-
wide by more countries. The startup and implementation of such PDEng programs is
not an easy task. To initiate these programs it requires joint initiative and conversion
on program content, but also financial and intellectual property aspects between uni-
versities, companies and government. To sustain these programs a continuous dia-
logue between these stakeholders is absolutely necessary. These are conditions that
are not quickly and easily met. Dutch industry took the initiative in 1986 and universi-
ties and government converged quickly. Over twenty-eight years these partners man-
aged to keep aligned in the interest of the Dutch innovation agenda and of the many
(inter)national technological design talents that found their career for life (and are
living it).”
“The PDEng programs thrive on the industry’s interest in cooperating, and sup-
porting (both in supervising the PDEng trainee, and in financial contribution) the
one year individual design assignments as topic for the PDEng design thesis. Many
of DPTI’s multinational industrial partners in large multinational consortia, also reg-
ularly provide design problems for the PDEng traineeships. Also, small and medium
enterprises and startup companies are able to find their way via the TU Delft Valoriza-
tion Center to this cooperation with the PDEng traineeships. All industrial partners
(small or big) not only benefit from a novel design solution for their design assign-
ment (by the inputs of fresh minds, a sound design methodology, project management
approach, and the latest academic research results), but also build and maintain a
strong and lasting network with the university and with top level designer trainees
who mostly start their industrial career in the Netherlands.”
“In order to prepare the PDEng trainee for the challenging one year individual
design project (IDP) for an industrial partner, over the years a technical training pro-
gram has been developed to strengthen selected knowledge and competence areas.
The structure of the programs as described by Swinkels [6] hasn’t undergone major
changes over the last decade. Strengthening the design methodology tools, and fine
tuning and extending the advanced domain knowledge into the quickly developing
research and education strengths are the major developments. The specific example
338 | 14 Education
listed below is taken from the PDEng traineeships in the (bio)chemical product and
process engineering at TU Delft) [7]:
– Personal skills (personal effectiveness, technical writing and editing, presenta-
tion skills).
– Business skills (project management, technoeconomic evaluation).
– Advanced methods for product and process design (design methodology for sus-
tainable product and process design and computer tools).
– Advanced domain subjects (process intensification, catalysis and reactor engi-
neering, nano engineered materials and products) linked to the research focal
areas of the TU DPTI and completed with a design assignment.
– A real design case from industry, executed as a group design project (GDP) for
three full months, and intensively coached by industrial principals and TU Delft
scientists and experienced design coaches.”
“The PDEng trainee’s knowledge and competencies continue to be trained and

coached during the full year Individual Design Project executed at the industrial
partner.”
The contents of the key PDEng design courses are presented in paragraphs 14.3.3
and 14.3.4.
14.3 Design oriented courses at TU Delft’s BSc, MSc, PhD and

PDEng level
14.3.1 BSc molecular science and technology (MST) and BSc life science and
technology (LST)
14.3.1.1 Chemical product design (4052TLEOY, 6 ECTS)

The course provides insight into the aspects that play a role in developing chemical
products. Teams of five to six students need to effectively use their creativity and ac-
quired chemical technology knowledge to develop an innovative, but realistic, new
product. Course topics comprise: performing as a team, product design and develop-
ment, relation between product properties and physical chemical composition, devel-
oping physical chemical models to describe product properties, identifying customer
needs and idea generation and selection. The course setting mimics the company sit-
uation by assigning a product development task to a development team of engineers
(i.e., students) by their manager (i.e., a staff member of the Department of Chemical
Engineering). The teams are graded on their report (75%) and presentations (25%). All
students will fill in a peer review form on their fellow teammates. This peer review is
used to adjust the average group grade and to determine the grade of the individual
students.
14.3 Design oriented courses at TU Delft’s BSc, MSc, PhD and PDEng level | 339
14.3.2 MSc chemical engineering
14.3.2.1 Product and process design (CH3804, 5 ECTS)

This course is a preparatory course for the Conceptual Design Project (CH3843). Topics
covered include:
– product and process design methodology
– material supply chains
– opportunity identification
– relation product performance and composition
– process integration
– process flow sheet modeling (Aspen Plus® )
– energy and mass integration
– process and product evaluation and optimization
– health, safety, environmental and sustainability aspects of design
– economic performance estimation and evaluation
– creativity methods and team performance
Lectures are combined with multiple graded product and process design assignments
in which teams of students directly apply the obtained knowledge. The students are
graded through a combination of individual tests (50%, product, and flow sheet mod-
eling) and team assignments (50%). An individual mark for the team assignments is
established by conducting a peer review in which all students evaluate the contribu-
tion of their fellow teammates.
14.3.2.2 Conceptual design project (CH3843, 12 ECTS)

The design project involves a conceptual design generation for a chemical product,
device or process. To offer students a realistic experience, all assignments are of high
practical relevance and mostly delivered and cosupervised by experts in the industry
acting as ‘clients’. The projects are carried out in teams (between five and six students
from mixed educational backgrounds and nationalities) with a stage gate based in-
termediate reporting and presentations to train teamwork and communication skills.
Each project is supervised by a university team coach. Furthermore, the team is sup-
ported by multiple university scientific staff members who are selected based on the
project content and act as technical advisors. Creativity is stimulated through a creativ-
ity workshop and active application of the various creativity tools during the project
execution. Final grading is based on:
– Design skills, theoretical knowledge and understanding (context, problem anal-
ysis, alternatives generation and design quality)
– Design project organization, planning and execution (realization, motivation, cre-
ativity and open mindedness)
– Communication and social skills (reporting, presentations and team perfor-
mance)
340 | 14 Education
With input collected from all supervisors and advisors. Individual marks for the team
members are established by conducting a peer review in which all evaluate the con-
tributions of their fellow teammates.
14.3.3 Process and product design PDEng courses
14.3.3.1 Advanced principles of product and process design (ST6064, 6 ECTS)

The ‘Advanced Principles in Product and Process Design’ (ST6064) course is divided
in three parts:
– Part 1: online individual product design assignment
– Part 2: the core of the course on product design and process design methodologies
– Part 3: comentor and critically review and evaluate a MSc level conceptual product
or process design (draft final report)
The first part of the course provides the PDEng trainees with the basics of design work
and is executed as an individual self study course with assignments. This first part of
the ST6064 course starts immediately after starting the PDEng traineeship.
The core of this course consists of four parts. Course participants:
– Read the course materials on product and process design principles and method-
ologies discussed in this book and based on the Delft Design Map (DDM) for prod-
uct and process design and other referenced text books. They engage in online
discussions on the material. This preparation is followed up in six half day work-
shops. During these workshops the material is summarized by the lecturer, exam-
ples are provided and a question and answer session is held.
– Apply the design methodology, project planning, team working skills, and com-
munication tools in an open ended team project. Teams consist of three to four
PDEng trainees who are coached by the lecturer(s). The team members each spend
80 hours on this project. In three workshop meetings, the teams present and dis-
cuss progress and receive feedback on the design approach, results and team co-
operation. The final results are reported, and formally presented and discussed
with a PDEng trainee peer group.
– Attend and contribute to a one day national conference or workshop organized
by, for example, Netherlands Network Process Systems Engineering – NL (PSE-
NL) [8], Process Intensification Network (PIN-NL) [9] and/or others.
14.3.3.2 Process simulation laboratory (ASPEN Plus® ) (ST6063A, 2 ECTS)

This course’s objective is to train students to build a flow sheet in Aspen Plus® using
a few example cases, to understand an existing flow sheet simulation, to simulate the
process with Aspen Plus® , and to solve problems related to the modeling and conver-
sion of these simulations. Also, practical skills in modeling and analysis of complex
systems is acquired, and insight into the simulation of several processes is gained.
The course covers mass and energy balances, sequential modular flow sheeting,
application of various thermodynamic models, use of simulation convergence meth-
ods, and finally a two week process design and simulation project on the simulation
of a complete (bio)chemical process.
The course is completed with presenting and discussing the project results at the
Fluor Haarlem office for a professional audience and with the assessment of the project
report.
14.3.3.3 Advanced process energy analysis and optimization (ST6101, 2 ECTS)

Industrial processes often have large utility requirements for heating and cooling pro-
cess streams to be able to convert, separate and transport raw materials to products.
Efficient use of utilities reduces operating costs and CO2 emission per unit product.
Since the 1970s, analysis and optimization techniques have been developed to assess
heat requirements and integration opportunities within process systems and indus-
trial sites. The basic principles are generally available in chemical engineering text-
books. This course gives an advanced guide to applying basic and advanced energy
analysis techniques to common industrial cases. It uses these techniques to optimize
industrial heat exchanger networks, both in grassroots and retrofit designs. Addition-
ally, attention is given to the design of common utility systems for process heating
and cooling. Cases are elaborated on using SuperTargetTM Software. The course cov-
ers both individual and team assignments. The team assignments are presented at the
CB&I The Hague office to an audience of experienced designers.
14.3.3.4 Technology management, economical evaluation in the process industry

(ST6612, 6 ECTS)
This course develops an understanding of a broad range of economic, financial, strate-
gic and marketing issues. It also examines the characteristics and working of the pro-
cess industry business in general and of the development, design and construction
funnel in particular. The course covers best practice in capital and operating cost es-
timating, and assessing project profitability in lifecycle terms. Furthermore, it offers
an introduction to financial methods for nonaccountants, and discusses a company’s
financial performance in the light of its financial statements.
The course has been developed specifically for postgraduate engineers and re-
searchers wishing to pursue a career in process development and design in a manu-
facturing or engineering/construction environment. At the same time, it is also geared
towards the needs of those already working in the industry for some years.
Being able to determine and evaluate investment costs is of vital importance for
the process industry. Participants in the course will learn to think in nontechnical
terms as costs and market share. The course consists of three parts. In the first part
342 | 14 Education
of the course, the structure and characteristics of the process industry are analyzed.
Next, attention is paid to organization and layout of projects and to the role of eco-
nomics. The second part aims at methods for determining investment costs, annual
costs and profitability. The third part deals with evaluating the financial performance
of a company as a whole. Also, attention is paid to strategic aspects of investment de-
cisions. In concordance with the course participants, other subjects can be dealt with
(for example, the ‘Porter industry analysis’ and ‘mergers and acquisitions’).
The course starts with interactive lectures combined with workshops. The work-
shops comprise of exercises and case studies. The course is concluded with an assign-
ment in which the economic feasibility of an actual project is assessed.
Course outline
– Part 1
– introduction industry/economy general
– process (chemical) industry business
– case study
– project organization
– economic evaluation/investment appraisal
– Part 2
– capital cost estimating
– course project
– case studies
– manufacturing cost estimating
– case studies
– evaluation profitability/discounted cash flow (DCF) analysis
– Part 3
– cost of capital, risk and uncertainties
– finance and accounting fundamentals
– case study
– finance and accounting fundamentals (continued)
– management accounting
– case study/project
14.3.3.5 (Personal and) project management (ST6111, 2 ECTS)

The (Personal and) Project Management training is set up to learn how to combine per-
sonal, project and process management theory with daily (project) work. In addition,
it demonstrates how to apply these methods into practice during the various projects
in the PDEng traineeship. ‘Hard’ project management skills like project structuring,
project planning and scheduling, are covered. However, an analysis of personal com-
petencies and preparing and working on a personal improvement plan are also key
parts of this course. The course builds on multicultural differences, and peer shar-
ing and learning. Individual coaching sessions are held with all participants in the
first year and the personal improvement plan is further monitored during the design
projects.
14.3.3.6 Sustainable design of processes, products and systems PDEng course

(ST6792, 4 ECTS)
Educating design in general and educating sustainable design in particular to stu-
dents is paved with difficulties. On the one hand the students should be free to use
their creativity to come up with breakthrough novel designs that meet a need and are
a factor four to ten better in environmental, economic and social impacts than con-
ventional solutions. On the other hand, because they are still inexperienced in de-
sign, they can become stifled due to this difficult task and their little experience in
designing in general. Moreover, students learn designing in various ways. Dorst [10]
distinguishes six different forms of intelligence relevant for design: linguistic, logical,
spatial, musical, bodily and personal. He also observed different characters such as
pragmatic problem solvers, autonomous designers and visionary artists. It is therefore
important to have the teaching and coaching adapted to these differences in personal
make up. To teach the students, a method has been developed that facilitates differ-
ent learning methods and that provides an atmosphere of trust between students and
teachers. The design course is set up in six steps, each taking half a day, with two to
three weeks in between for assignments. Each step consists of:
– A plenary teaching lecture of a course element group design work starting in the
classroom with the teacher being available for coaching on request
– Assignment task work: a course book by Jonker and Harmsen [11] is used which
allows students to look up information
– Presentation of intermediate results
– Feedback by the teacher on these results
In this way, the design effort is cut up in smaller steps and allows for a large amount
of coaching specifically geared towards each group.
Course program schedule:
– Day 1: Theory of triple P sustainable development and relation to product design
– Team task 1: Define triple P sustainable design scope, goal and constraints for
given case
– Day 2: Presentation group results team task 1
– Theory of industrial symbiosis
– Team task 2: Redefine design scope, with all inputs and outputs and destina-
tions using industrial symbiosis theory
– Day 3: Presentation of results team task 2
– Theory precaution principle. Theory lifecycle assessment (LCA) and interac-
tive workshop. Theory closing material cycles and atom balances
– Team task 3: Define LCA functional unit and all LCA steps for a case and make
an LCA block flow diagram with all input and output streams. After day three:
PDEng trainees elaborate at home the problem definition and LCA for the ref-
erence case
344 | 14 Education
– Day 4: Presentations of results team task 3

– Theory on people dimension: social fit with culture and acceptance. Theory
scenario set building for future robustness test
– Team task 4: Describe the social and economic context relevant to the design
and redesign the solution to fit with the social context.
– Team task 5: Make a set of two scenarios, then analyze the design solution for
robustness using scenario set.
– Day 5: Presentation of results team tasks 4 and 5
– Question and answer session
– Team task 6: Finalize design case. After day five, the PDEng trainees will fi-
nalize their design and design report/presentation in teams. The final design
report is a substantial amount of work to earn the 4 ECTS credits (which totals
112 hours). Six half days of coursework (24 hours) leaves 88 hours per student
for the home problem and final project assignment work. The end deliverable
is a written report to be sent to the course lecturer four days before the final
oral presentation.
– Day 6: Presentation of the triple P design result to the PDEng peer groups and
lecturer
14.3.3.7 Group design project (GDP) (ST6802, 21 ECTS, ST6814, 17 ECTS, or ST6815,
17 ECTS)
The objective of the group design project (GDP) for all PDEng programs (Process &
Equipment Design – ST6802, Designer in Bioprocess Engineering – ST6814 and Chemi-
cal Product Design – ST6815) is to enable the participants to develop their design skills
in a real life design case in teams consisting of three to five members. The group as a
whole is responsible for the final outcome of the project. In turn, each member of the
group will be assigned the group leadership. The participants are forced to use their
deepened skills of product and process fundamentals and acquired design skills in
an integrated way. They are made aware of other engineering disciplines in the areas
between conceptual product and process design and an operational plant. The theory
and learnings from the prerequisite courses Advanced principles of product and pro-
cess design (ST6064) and (personal and) project management (ST6111) are applied as
follows:
– Each participant spends 600 hours on the assignment.
– The total amount of (elapsed) time for the assignment is limited to half a year,
so each participant spends at least 24 hours a week (60%) on the assignment on
average.
– The hours not spent on the project (400 hours) are used for other PDEng program
subjects.
– The implementation of the assignment is the responsibility of the team as a whole.
– Generally, three reports are written during the project and two intermediate and
one final presentation(s) is held at the following milestones meetings:
– Scope and basis of design

– Results of conceptual design (including product composition and structure,
process flow diagrams and balances)
– Final review meeting: evaluation of the design, final conclusions and recom-
mendations
– These meetings take place at TU Delft or at the location of the (industrial) principal
and for a wider audience of the principal’s employees.
– The team members manage their own planning and project execution. During the
project the supervisors act solely as consultants. After project completion, the su-
pervisors review and assess the results and the team performance.
Progress and consultation meetings are scheduled on a regular basis and are attended
by the participants and supervisors (at least every three weeks although often more
frequently). The project is carried out under supervision of TU Delft staff members,
who are in an advisory role of an experienced process engineer/designer from the
industry. The industrial supervisors join in the progress meetings (at TU Delft or at
their location). In between meetings, communication takes place by telephone, on-
line video communication or email. The supervision by (an) experienced designer(s)
from industry is a directive of the Netherlands Committee for Certification of (post
MSc) Technological Designer Programs (CCTO).
14.3.3.8 Loss prevention in process design (ST6042, 5 ECTS)

The main issue of the course Loss Prevention in Process Design, is the design of safe
chemical processes. Important subjects are: inherent safe process design, equipment
design, selection of material, piping systems, heat transfer systems, thermal isolation,
process monitoring and process control, High pressure systems, electrical systems,
control of dangerous emissions (for instances flares) and documentation are also in-
cluded.
The goal of the course is to familiarize PDEng trainees with advanced knowledge
of risks of fire, explosion and toxicity of chemical products and processes. Students
should be able to design a safe and environmentally friendly chemical product and
plant using modern loss prevention design methods upon completion of the course.
The course consists of the following parts:
– lectures/exercises on chemical risks (fire, explosion, toxicity)
– student presentation on chosen safety topic
– preparation of written conceptual design for hazardous chemical process, with
detailed analysis of and solution to safety, health and environmental risks, mak-
ing use of various approaches: layer of protection analysis (LOPA), inherently
safer design (ISD), green chemistry and green chemical engineering
– presentation of project assignment results. Students will work in groups of two to
four.
346 | 14 Education
The objective of the assignment is to create a conceptual design of a chemical process

for safe and environmentally friendly production of a hazardous substance (where
‘hazardous’ refers to toxic, flammable and explosive substances), using a modern de-
sign philosophy. The scope of the assignment comprises of: synthesis route and reac-
tion stoichiometry, fire/explosion/toxicity data on all substances involved (reactants,
products and byproducts), basic thermodynamics, basic kinetics, process flow sheets
including most the important unit operations, description of process including pro-
cess conditions, with special emphasis on the design of the most hazardous part of
process.
14.3.3.9 Individual design project (IDP) (ST6902, or ST6903, or ST6904, 60 ECTS)

All PDEng trainees are carrying out an individual design project (IDP) during the sec-
ond year of their PDEng traineeship. The necessary skills needed for this project are
provided during their prior studies and during the first year of the traineeship.
The IDP aims for the (re)design of a (bio)chemical product (material, formula-
tion and configured product device) and process or method, including attention for
the imbedding in the supply chain.In each case, it should comprise a design problem
that requires a novel design for a client, rather than applying already existing designs.
Each individual design assignment generally contains design as well as research and
development aspects. In all cases, the research and development must be beneficial to
the design: e.g., acquiring required knowledge or experimental data. The design part
should amount to at least 50%, leaving at most 50% for the research and development
part.
The design task can be part of a larger project: e.g., a product and process develop-
ment/design project. The designer will always interact with other project participants.
14.3.3.9.1 Confidentiality and working location

At the request of the principal, all knowledge obtained during the project will be
treated as confidential and will not be disclosed to third parties without prior consent
of the cooperating partner (principal). To this end, a cooperation agreement is signed
by both the TU Delft and cooperating partner. This agreement also contains a nondis-
closure agreement for the designer and TU Delft members of the supervising steering
team. Detailed guidelines on how to act are given by the program management team.
For the IDP, the PDEng trainee will be mostly seconded to industry or to a research
section at TU Delft. When working with the industrial partner, relocation may need to
be considered.
14.3.3.9.2 Composition of steering group

The steering group of an IDP consists of a mentor from the principal engineer from the
industrial party, a professor/mentor from university, a consultant with expert knowl-
edge on the subject and an experienced design mentor from the PDEng program (the
consultant and professor may be the same person). The PDEng program management
will compose the steering team in close cooperation with the principal.
14.3.3.9.3 Milestones and implementation

During the IDP, at least four milestones can be identified: The kick off, basis of design
(BOD), the Intermediate and the Final reviews (BOD, Intermediate and Final Reports
are to be submitted). They consist of meetings with both the industry and university
supervisors, and include a presentation by the PDEng trainee. The Kick-off meeting
is scheduled within the first two to four weeks after the start of the project. The BOD
review is scheduled two to three months after the start of the project, the Intermediate
review is after 8 months, and the Final review is at the end of the project. All members
of the steering group attend these meetings.
Generally, the Principal of the project will provide a project description, terms of
reference, or proposal at the very beginning of the project. Often the description is
relatively brief (concise) and not clear on all details. One of the first activities of the
designer will be to lay down the detailed project description in close cooperation with
the principal. The project description addresses the objective of the project, deliver-
ables and a (time) plan how to obtain the deliverables (project approach). The project
content will be formally discussed during the Kick-off meeting, which is to be orga-
nized preferably during week two and four of the project.
The main purpose of the Kick-off meeting is to make clear the objectives, and to
agree upon them. The definition stage of a project leaves much room for misunder-
standings between parties. Therefore, the description must be as explicit as possible.
In order to avoid any misunderstandings, it is helpful to provide a description of the
different project activities and the duration thereof. In this way, the parties involved
will get a clear idea about on what subject and how time is spent, and whether parties’
expectations are satisfied. Focus points during the Kick-off Meeting are:
– Project objective
– Scope of the project, delineation
– Approach
– Project planning, time table
– First results
Several other meetings with the supervisors and/or experienced design engineers are
organized throughout the year the design project takes. The meeting frequency de-
pends on the need and the opportunity. It is also the PDEng-trainee’s responsibility to
organize these meetings. Minimum will be a monthly meeting with the project man-
ager/design mentor from the PDEng program (bi-weekly contact is advised).
Nontechnical skills needed for the execution of the project are to a large extent ob-
tained during the first year, with dedicated courses on the social-economical aspects,
and personal and professional skills as part of the core program. The designer is re-
sponsible to manage the project and to communicate adequately with all stakeholders
348 | 14 Education
involved. The designer has to organize and plan progress meetings. Moreover, the de-
signer will also chair the meetings, and ensure to distribute agenda and minutes of
meeting.
14.3.3.9.4 Reporting and oral presentations

The purpose of reporting is to document and transfer obtained knowledge from the
designer to the project principal. To give evidence of skills, the designer is requested
to prepare a BoD report and an intermediate report. In this way, the reporting effort is
started at an early stage and is adequately spread over the project span.
Alongside the written report, the designer also will report orally during the steer-
ing group review meetings. At the end of the project two, final presentations are re-
quired. One presentation is to be held at the principal and one at the TU Delft. During
the presentation at the principal, generally, representatives from both management
and coworkers from the principal will be attending.
At the university, a colloquium must be given. As a rule, the colloquium is open
to the public. Care must be taken to have the contents cleared by the principal before-
hand, so as not to disclose any confidential information.
14.3.3.9.5 Assessment
After the completion of the IDP, the work will be reviewed with respect to theory, im-
plementation and communication in accordance with the TU Delft PDEng assessment
criteria. For all three aspects a mark will be given. The appraisal of the work will be
carried out by the steering committee and under the responsibility of the professor/
mentor.
14.3.4 OSPT courses (PhD, PDEng and participants from industry)
The OSPT (which is the Dutch abbreviation for the Research School Process Technol-
ogy) organizes advanced two to five days long courses and workshops for PhD and
PDEng students and for participants from industry.
The courses take place at TU Delft, TU Eindhoven or University Twente. The Chem-
ical Product Centric Sustainable Process Design course (by R. Gani) held at TU Delft
is described below in detail. The other courses are listed below with a reference to the
OSPT course plan website [12].
14.3.4.1 Chemical product centric sustainable process design (PhD/PDEng course)

The objective of this PhD/PDEng course is to give the participants a view of chemical
product design and the important process design issues related to their development
(product centric process design). The course will highlight how to define the needs
of a chemical product, how to identify the candidate chemicals and/or mixtures of
chemicals and how to quickly evaluate the important process design issues so that
decisions related to product development can be made in the early stages of product
development. The objective is also to highlight the currently available methods and
tools that can be applied to solve various types of problems associated with product
process design in a systematic and integrated manner. Different case studies will be
used as application examples. The ICAS software developed by the Computer Aided
Process Engineering Center (CAPEC) at the Technical University of Denmark (DTU) will
be used. Also, two new software tools, ProCAFD (a tool for synthesis-design of chemi-
cal processes) and ProCAPD (a tool for synthesis-design of chemical products), made
available through the PSE for SPEED Company Ltd [13], will be used in the course.
In chemical product design and development, one first tries to find a product that
exhibits certain desirable or targeted behavior and then tries to find a process that can
manufacture it with the specified qualities. The product may be a single chemical, a
mixture, or a formulation. For the later product type, additives are usually added to
an identified active ingredient (molecule or mixture) to significantly enhance its de-
sirable (target) properties. Examples of chemical products, such as functional chemi-
cals (solvents, refrigerants, lubricants, etc.), agrochemicals (pesticides, insecticides,
etc.), pharmaceuticals and drugs, cosmetics and personal care products, home and
office products, etc., can be found everywhere. The workshop will only cover chemi-
cal based products and issues related to their design, development (including process
design also) and analysis (product performance). Even though it is possible to iden-
tify many chemicals and/or their formulations as potential chemical products, only a
small percentage actually becomes commercial products. Finding a suitable process
that can sustainably (that is, reliably, efficiently and economically) manufacture the
identified chemical with the desired product qualities, as well as evaluating product
performance during application and analyzing market trends, play important roles
in product design and development. From a process point of view there are products
where the reliability of the quality of the manufactured chemical may be the deciding
factor (for example, drugs and agrochemicals), while there are others where the cost
of manufacturing the product is at least as important as the reliability of the prod-
uct quality (such as solvents, refrigerants, lubricants). This means that product cen-
tered sustainable process design is important because identifying a feasible chemical
product is not enough. To make it sustainable, the process needs to be efficient, reli-
able, economically feasible and environmentally acceptable. Also, while in the case
of functional chemicals the identified molecule or mixture is the final product, in the
case of chemicals based consumer products (drugs, cosmetics and personal care prod-
ucts, etc.), they are intermediate products from which the final products are obtained
through additional processing. Therefore, the performance of the manufactured prod-
uct, when applied, needs to be tested and validated. For some functional chemical
products (such as solvents and refrigerants) this may be straight forward, but for some
consumer products (such as drugs and food products), it may not be so straight for-
ward.
350 | 14 Education
14.3.4.2 Other OSPT courses (see [12])

– Computational Fluid Dynamics for Chemical Engineers
– Advanced Thermodynamics
– Fundamentals and Practice of Process Intensification
– Molecular Affinity Separation
– Numerical Methods for Chemical Engineers
Abbreviations
3TU.SAI Federation of three Dutch technical Universities – Stan Ackermans
Institute
4TU.SAI Federation of four Dutch technical Universities – Stan Ackermans
Institute
BSc Bachelor of Science
BOD Basis of design
CAPEC Computer Aided Process Engineering Center
CB&I Chicago Bridge & Iron Company
CCTO Certification Committee for (post MSc) Technological Designer Programs
CO2 Carbon dioxide
DCF Discounted cash flow
DPTI Delft Process Technology Institute
DTU Denmark’s Technical University
ECTS European Credit Transfer System
GDP Group design project
ICAS Software tool developed at CAPEC/DTU
IDP Individual design project
ISD Inherent safer design
KIVI Koninlijk Instituut Van Ingenieurs (Royal Dutch Society of Engineers)
LCA Lifecycle assessment
LOPA Layer of protection analysis
LST Life Science & Technology (BSc and MSc programs)
MSc Master of Science
MST Molecular Science & Technology (BSc program)
OSPT Onderzoeksschool Procestechnologie (Dutch for: Research School
Process Technology)
PDEng Professional Doctorate in Engineering
PhD Doctor of Philosophy
PIN-NL Process Intensification Network – Netherlands
PPP Public/private partnerships
ProCAFD Software tool for computer aided process synthesis design

ProCAPD Software tool for computer aided product synthesis design
PSE-NL Process Systems Engineering – Netherlands (network)
RUG Rijksuniversiteit Groningen
TU Technical university
TU/e Technical University Eindhoven
UT University Twente
[1] Molecular Science and Technology BSc curriculum, TU Delft, 2017. Accessed October 8, 2017,
at https://www.tudelft.nl/en/education/programmes/bachelors/mst/bsc-molecular-science-
technology/curriculum/.
[2] Life Science and Technology BSc curriculum, TU Delft, 2017. Accessed October 8, 2017,
at: https://www.tudelft.nl/en/education/programmes/bachelors/lst/bsc-life-science-
technology/curriculum/.
[3] TU Delft Process Technology Institute (DPTI), TU Delft, 2017. Accessed October 8, 2017, at http:
//dpti.tudelft.nl.
[4] QS world universities ranking by subject – chemical engineering. QS Top Universities, 2014,
2015, 2016). Accessed October 10, 2017, at http://www.topuniversities.com/university-
rankings/university-subject-rankings/2014/engineering-chemical,
https://www.topuniversities.com/university-rankings/university-subject-rankings/2015/
engineering-chemical, https://www.topuniversities.com/university-rankings/university-
subject-rankings/2016/engineering-chemical.
[5] Swinkels PLJ. Post MSc technological design (PDEng) traineeships by Dutch universities of
technology catalyze industrial innovation, QScience Proceedings (Engineering Leaders Con-
ference 2014) 2015: 21. Accessed October 8, 2017, at http://dx.doi.org/10.5339/qproc.2015.
elc2014.21.
[6] Swinkels P, Grievink J, Hamersma P. Technical and human factors in chemical process design
courses. In: Tu S-D, Wang Z-D (eds.), Total engineering education, Shanghai, East China Univer-
sity of Science and Technology Press, 2006, 59–65.
[7] Traineeships Engineering Design – PDEng programmes, TU Delft, 2017. Accessed October 8
2017, at http://www.pdeng.tudelft.nl.
[8] Netherlands Process Technology Networks – Process Systems Engineering, 2017. Accessed
October 8, 2017, at http://www.processinnovation.nl/pse.
[9] Netherlands Process Technology Networks – Process Intensification Network (PIN-NL), 2017.
Accessed October 8, 2017, at http://www.processinnovation.nl//homepinnl.
[10] Dorst CH. Design problems and design paradoxes. Design Issues, 22(3), 2006, 4–17.
[11] Jonker G and Harmsen J. Engineering for sustainability – a practical guide for sustainable de-
sign. Amsterdam, Netherlands, Elsevier, 2012.
[12] OSPT Courses, ISPT Innovation Academy. Accessed October 8, 2017, at http://www.ispt-
innovationacademy.eu/course-agenda.html.
[13] PSE for SPEED – Sustainable Product Process Engineering, Evaluation & Design. Accessed
October 13, 2017, at http://www.pseforspeed.com.
A3 Appendix to Chapter 3
Group design with Belbin team roles
Project groups are mostly constructed based on the required disciplines only. How-
ever, if groups have to cooperate closely for the desired results then they should get
along with each other. In addition to disciplinary design, Belbin team roles can be
used to populate the group or team [1].
Belbin role descriptions

– The Plant role is to be highly creative and good at solving problems in unconven-
tional ways. He may be the outsider planted in the team.
– The Monitor provides a logical eye to make impartial judgments about options.
– The Coordinator focuses on the team’s objectives and delegates work appropri-
ately.
– The Resources Investigator provides knowledge from outside the team.
– The Implementer plans a practical, workable strategy and carries it out as effi-
ciently as possible.
– The Finisher makes sure at the end of a task, to remove errors, subjecting it to
the highest standards of quality control.
– The Teamworker helps the team to gel. He may also fill in gaps left by others.
– The Shaper provides the necessary drive to keep the team moving to reach the
goal.
– The Specialist provides in-depth knowledge of a key area. His weakness is to
focus narrowly on the problem element and prioritize it over the team’s progress.
In the authors’ view it is in general better to consult the specialist rather than
making him a team member.
The preferred role of an individual can be determined by one’s self using the Belbin
self-perception method. This method can be found hereafter.
Arittzeta [2] provides a review of a large number of experimental studies on teams
designed with Belbin team roles. His conclusion is that Belbin’s team role theory has
been validated. Some roles, however, show intercorrelation. From the intercorrelation
information, the following major style groupings can be made:
Completer style: Completer-Finisher, Implementer, Monitor Evaluator
Bridge style: Teamworker, Monitor Evaluator, Coordinator
Innovative style: Resource Investigator, Plant, Shaper
Therefore, small teams should have one of each of these three major styles.
The authors have also had a powerful experience with groups designed using the
Belbin self-perception team roles. The most complementary team indeed produced
https://doi.org/10.1515/9783110467741-015
the best results, while the team with members of only one role (specialists) performed
worse by far.
Making inventory for Belbin team roles using self perception
Teams with complementary Belbin roles can be easily composed as follows. Each po-
tential team member fills in a questionnaire and concludes their preferred role(s).
Then a team or several teams can be composed with complementary roles. The ques-
tionnaire can be obtained from:
http://www.belbin.com/about{/\penalty\exhyphenpenalty}belbin-team-roles/
Creativity methods
The literature on creativity methods for innovation is vast. We limit ourselves here to
methods for which the authors have experience in industrial product-process prac-
tice. A good introduction to further information can be found in Nijstad [3] and Her-
mann [4].
Brainwriting 6-3-5
Creating ideas within the product-process industries is probably slightly different from
that within other industries in that process engineers as a majority are introverted peo-
ple, who feel uncomfortable in directly expressing ideas. Brainwriting is sometimes a
better method than brainstorming, in particular for introverts. On average brainwrit-
ing generates twice as many ideas as brainstorming [3]. For process engineers that
factor could be even higher.
The brainwriting 6-3-5 method is very simple to execute and needs no trained fa-
cilitator.
Step 1: Plenary, a clear problem to solve is defined and stated clearly to all partici-
pants.
Step 2: Groups of preferably 6 people are formed.
Each member gets a writing chart with at least 3 columns and a number of
rows at least equal to the number of members.
Step 3: In Round 1 each group member writes down silently 3 ideas in the first row on
his writing chart.
Step 4: After 5 minutes the writing charts are rotated to the neighboring members and
Round 2 is executed in which each member enriches the 3 ideas of the first row
by putting additional ideas in the second row
354 | A3 Appendix to Chapter 3
Step 5: This 5 minute cycle is repeated until the members are exhausted of ideas.
Of course additional clean sheets of writing charts can be provided to start
new sessions and rounds.
Step 6: All filled in sheets are gathered by the facilitator and recorded outside the
brainwriting session.
Potential problem analysis for start up
Potential problem analysis for the start-up phase is very useful for generating a robust
start-up plan. Here is a template that the author has prepared and used for making
robust start-up plans. It is derived from a problem analysis method [5] and turned into
a potential problem analysis method.
Potential problem analysis template:
Question Reference New design Difference in new Risk

product/process design
What
Where
When
Size
[1] Belbin, 2017. Sourced from: http://www.belbin.com/rte.asp.

[2] Aritzeta A, et.al. Belbin’s team role model: Development, validity and applications for team
building, Journal of Management Studies, 44(1), 2007, 96–118.
[3] Nijstad BA, Paulus PB. Group creativity: Innovation through collaboration. 2003, 326–339.
[4] Herrmann D, and Felfe J. Effects of leadership style, Creativity technique and personal initiative
on employee creativity, Brit J Manage, 25, 2014, 209–227.
[5] Kepner CH, Tregoe BE. The new rational manager, New Jersey, USA, Kepner-Tregoe Inc., 1981.
A4 Appendix to Chapter 4
History of (chemical) product and process design
leading to Delft Design Map
(Chemical) products and their manufacturing processes have been around for a long
time, long before formal product and process design methodologies were developed
and taught at universities of technologies since the end of the 1960s. Industry man-
aged to do this based on heuristic knowledge of and operational experience with these
products and processes. Experienced designers transferred their knowledge and expe-
rience through heuristic rules to novice designers. Design improvements were mostly
incremental.
Design courses at universities in the 1970s and 1980s focused heavily on the (very
powerful) process “unit operations” design, with attention toward shortcut methods
for process integration, and performance analysis mainly based on cost estimation
and techno-economic measures. The really good text books “Plant design and eco-
nomics for chemical engineers” [1] and “Strategy of process engineering” [2] were used
in many curricula in this period. The conceptual process design course taught at TU
Delft by Prof. A. Montfoort (TU Delft 1970–1990) was based on this approach.
Process design methodology development for education started to take off in
the 1980s. In this period, steady state process flowsheet simulatorsprocess flowsheet
simulator (like ASPEN™, ASPEN-PLUS™ (AspenTech) and PRO-II™ (Simulation Sci-
ences Inc.) emerged from research work at universities (like Massachusetts Institute
of Technology (MIT)), and were launched as commercial tools for analyzing mass and
energy balances of complete processes. Through these computational tools – includ-
ing key thermodynamic models – a ‘systems engineering approach’, built up from
unit operations and including process analysis and optimization became possible.
The development of the process integration methodologies (thermal pinch analysis)
at University of Manchester Institute of Technology (UMIST) by Linnhoff in 1968 [3],
with applications in Imperial Chemical Industries (ICI) in the 1980s also gave process
systems engineering thinking a clear boost. Here the systems engineering approach
was used to arrive at considerable energy conservation.
After this attention to process design methods took off (see Figure A4.1 below).
In the period 1985–1995, dynamic flowsheet simulators were also developed for
modeling process dynamics, startup, and control: SPEEDUP and gPROMS® were both
developed at Imperial College London (ICL), whereas Jacobian was developed at MIT.
The use of process flowsheet simulators allowed process systems engineering research
and education to focus again on the fundamental process-synthesis problem: what
unit operations and process structures were best suited for the required process task?
The widely used textbook “Conceptual Design of Chemical Processes” by Douglas,
issued in 1988 [4] successfully attempted to address these questions. In this book,
https://doi.org/10.1515/9783110467741-016
Task based, product performance driven product & process design
Heuristic,implicit design decisions, shallow decomposition, process design)

Chemical Product Design
DTU: Wesselingh et.al.,
Emphasising Chemical Product Design EFCE-PD&C: Bröckel et.al.)
ASPEN™, process (Cambridge: Delft
ASPEN-PLUS™ structures Cussler & Moggridge) Design
GAMS/DICOPT Map
PRO-II™ synthesis
SPEEDUP heuristics Product Driven Process Synthesis (PDPS) (TU Delft:
gPROMS®, (Unilever: van der Stappen, Bongers, Grievink,
Jacobina Almeida Rivera) Swinkels)
Driven by cost
and economic Pinch analysis
Delft Design ’Template’
performance Task-driven (TU Delft: Grievink, Swinkels)
analysis process design
Hierarchical (Siirola)
Decomposition
Method
Operational (Douglas)
experience
Optimisation
based
Process simulators For use in post-MSc PDEng Designer
method (MINLP)
Unit operations (steady state & Programmes (bio)chemical product &
(CMU)
process design dynamic) process design TU Delft, TU Eindhoven
(Rudd & Watson, In process design
Evolutionary Peters & (MIT, UMIST, ICL, ICI)
modification Timmerhaus)
method
<1970 1970 1980 1990 1995 2000 2005 2010 2015
Fig. A4.1: (Chemical) product and process design methods – a historical perspective.
he successfully translated his vast wealth of industrial design experience into a hi-
erarchical design decomposition scheme. Process structure options were generated/
synthesized using plausible heuristic rules in combination with the unit operations
concept. In 1999 Seider and Lewin published the first edition of their student textbook
on process design principles [5], taking a similar approach, but with more details on
process structure, equipment design and cost. In the new millennium, new editions
of this textbook have been upgraded over the years.
In the late eighties, computer-based optimization based methods were also de-
veloped to solve process design problems. Based on predefined superstructures en-
compassing all process flowsheet structure options, mixed integer nonlinear (MINLP)
programming techniques were developed and applied to find the optimal flowsheet
given the given optimization function. The GAMS/DICOPT MINLP algorithm was de-
veloped by Kocis and Grossman at Carnegie Mellon University [6].
In the mid-nineties, a novel task-driven process design approach was published
by Siirola [7, 8]. For the first time, unit operations were no longer the building blocks
for process design, but rather individual tasks or functions, such as adsorption, reac-
tion, heat and mass transfer, and diffusion. These tasks could be combined in single
columns generating novel design solutions, such as reactive distillation, in-situ crys-
tallizer fermentation, etc. Inside Eastman Chemical Company, this task driven design
approach had led to the design and implementation of large scale reactive distilla-
tion columns in which many tasks were combined, the well-known methyl acetate
production route [9]. This multifunctional design approach is now part of the process
intensification design tool box.
For product design, let alone concurrent product-process design, no generic de-
sign methods were yet developed and published in this period. For homogeneous
products, such as food products and chemical products, designs were carried out in-
side companies, based on experience and tacit knowledge. In the nineties, the product
design approach was at the phase where process design was in the late 1960s. Each in-
dustry has many heuristic rules, mostly limited to the existing product portfolio and
based on in-house experimental validation. Industrial researchers are reaching out
to academia where attempts were made to develop and understand application and
manufacturing of new molecules, structures, and materials.
With an increasing focus in the chemical industry and related academia on prod-
uct innovation, tools were becoming available and, once the desired product compo-
sition and structure is known, product driven and task based process design meth-
ods become available to tackle these design challenges. However, the product design
phase – how to arrive from a desired function/performance of the product to its com-
position and structure – is another matter. This is a field where chemical engineers
have historically not been educated and involved in, and is largely the domain of the
chemistry and material sciences scientists.
In the first years of the millennium, the first (chemical) product design courses
were initiated at several universities. Also, the first student textbooks were published
at Cambridge by Cussler & Moggridge [10] at RU Groningen and DTU Copenhagen by
Wesselingh et al. in 2007 [11]. In 2010 the third edition of the textbook by Seider et
al. [12] was for the first time focused on product design aspects.
The approach adopted by Cussler et al. and Wesselingh et al. starts with describ-
ing and explaining how chemical products work in performing their task, and how
changes can be made to improve them. Similar to the unit operation approach, this
product related knowledge was first presented and taught in a rather compartmen-
talized way, with little connection between different classes of products and applica-
tions. Wesselingh (with his transport phenomena background) tried to put more em-
phasis on the importance of an engineering (quantification) approach with transport
phenomena as a key part of product performance. He attempted to capture product
application properties in dimensionless numbers, and advocated the use of time con-
stant estimations of the ruling phenomena. This is a powerful tool process designer
use in the ‘scale-up by scale-down’ approach for processes, but is equally suited for
the characterization of product in-use processes.
Frens and Appel (both TU Delft) have also stressed in their product design courses,
the importance of physical chemistry and complex rheology in the formulation and
functioning of products in product design. Process systems engineering (PSE) scien-
tists/engineers with a focus on applying chemical thermodynamics in process synthe-
sis and engineering (Gani at Technical University of Denmark’s [DTU] Computer Aided
Process-Product Engineering Center (CAPEC) and K. Ng at Hong Kong University of
Science and Technology [HKUST]) started to extend their work to product design, de-
velopment, and simulation in the first decade of the new millennium.
In 2003 the European Federation of Chemical Engineering (EFCE) Section on Prod-
uct Design and Engineering (PD&E) was founded in Granada, Spain. This happened
at the first EFCE-sponsored PD&E Conference. In 2007 members of this section pub-
lished two volumes of a book on “Product Design and Engineering – Best Practices”
edited by Bröckel et al. [13, 14].
Starting in 2001 the cooperation between Grievink (ex-Shell) and Swinkels (ex-
Unilever) at TU Delft led to the combination of their complementary process and prod-
uct design knowledge and experience. The oil, gas, and chemicals company Shell has
a long history of designing and implementing large scale continuous processes for ho-
mogeneous products. Unilever, the fast-moving consumer goods company has a long
history of designing and developing novel structured and multiphase products for its
world markets. Grievink and Swinkels developed in this period at that time were called
hierarchically decomposed and task oriented “Delft Design Template for Conceptual
Process Design.” In 2014 a simplified version of the Delft Design Template for Concep-
tual Process Design was published in 2014 [15].
In cooperation with TU Delft, Unilever developed and tuned the Delft Design
Template approach to fast moving consumer products (mainly foods) and proposed
the “Product Driven Process Synthesis (PDPS)” methodology described by Meeuse in
2000 [16] and in 2017 by Almeida-Rivera [17].
Key developers van der Stappen, Bongers, and Almeida-Rivera at Unilever and
Boom et al. (Wageningen University & Research [WUR]) also adopted this approach in
their research into the design of new food products and their manufacturing.
A very similar approach in adopting a schematic conceptual model for chemical
product design has been taken and published by Bernardo at University of Coimbra
in 2017 [18], in which the same design levels/ functions are proposed (quality func-
tion, property function, and process function) as in the Delft method described in
Chapter 7. Bernardo also proposes a mathematical approach to solve the concurrent/
simultaneous product-process design optimization as described in Chapter 8 and 9.
The design methods; Delft Design Template (DDT) (now restructured and called
Delft Design Map, [DDM] and Unilever’s Product Driven Process Synthesis [PDPS])
have been taught for many years at the post-MSc PDEng programs at TU Delft (Pro-
cess & Equipment Design, BioProcess Engineering, Chemical Product Design) and TU
Eindhoven (Process and Product Design) and applied in the PDEng design projects
executed for and in industry. The advantage is that PDEng graduates learn a clear sys-
tematic approach to product oriented process design that is not dependent on prod-
uct and process circumstances and can be applied in the product-process industry
sectors, such as the chemical, biochemical, food, and others.
Delft Design Map for product and process design (DDM)
The Delft Design Map (DDM) is based on task-based design activity spaces, shaped in
matrix form. The horizontal axis corresponds with innovation stages and the vertical
axis corresponds with design steps, from problem definition to solution to reporting.
The empty blocks of the matrix are then filled during the design effort in a preplanned
order until the matrix is filled. The design problem at hand (new product, new process
and new supply chain, new reactor technology, heat integration) determines where
the emphasis of the design work should lay, but also emphasizes that designers should
be aware that implications of apparent small design changes (in reactor technology or
catalyst type, for instance) may have profound effects (positive, negative) on the en-
tire system. The design team can also use the DDM to map the route of the search
for suitable design solutions, encompassing: formulating the design problem, plan-
ning the design tasks, and following through on these activities including reporting
decisions and results. The DDM becomes then a planning, design management, and
communication tool. Therefore, the authors (after consulting Grievink) renamed the
‘Delft Template for Conceptual Process Design’ into the ‘Delft Design Map for Product
and Process Design.’
This map will be even more helpful for students and industry designers than its
predecessor. They get an overview of all design actions to be pursued to fill the matrix.
This gives them confidence that they can make a design because the whole problem
is cut up into small design action steps. The Delft Design Map is described in detail in
Chapter 4.
[1] Peters MS, Timmerhaus KD. Plant design and economics for chemical engineers, 2nd edn, New
York, NY, USA: McGraw-Hill, 1968.
[2] Rudd DF, Watson CC. Strategy of process engineering, New York, NY, USA: John Wiley & Sons
Inc., 1968.
[3] Linnhoff B, Flower FR. Synthesis of heat exchanger networks: I. Systematic generation of en-
ergy optimal networks, AIChE Journal, 24(4), 1978, 633–642.
[4] Douglas JM. Conceptual design of chemical processes, New York: McGraw-Hill, 1988.
[5] Seider WD, Lewin DR. Process design principles: Synthesis, analysis and evaluation, New York,
NY, USA, John Wiley & Sons Inc, 1999.
[6] Kocis GR, Grossman IE. Computational Experience with DICOPT: Solving MINLP problems in
process systems engineering, Computers and Chemical Engineering, 13(3), 1989, 307–315.
[7] Siirola JJ. An industrial perspective on process synthesis, 4th International Conference on
Foundations of Computer–Aided Process Design (FOCAPD), AIChE Symp Series USA 91(304),
1995, 222–233.
[8] Siirola JJ. Industrial applications of chemical process synthesis, In: Anderson JL, Bischoff KB
(eds.), Advances in Chemical Engineering Volume 23: Process Synthesis. Academic Press,
1996, 1–92.
[9] Agreda, VH, Partin LR, Heise WH. High-purity methyl acetate via reactive distillation, Chemical
Engineering Progress, 86(2), 1990, 40–46.
[10] EL, Moggridge GD. Chemical product design, Cambridge, UK: Cambridge University Press,
2001.
[11] Wesselingh JA, Kiil S, Vigild ME. Design & development of biological, chemical, food and phar-
maceutical products, Hoboken: John Wiley & Sons Ltd, 2007.
[12] Seider WD, Seader JD, Lewin DR, Widagdo S. Process and process design principles: Synthesis,
analysis and evaluation, 3rd edn, Hoboken, NJ, USA: John Wiley & Sons Inc, 2010.
[13] Bröckel U, Meier W, Wagner G. Product design and engineering – best practices, Volume 1:
Basics and technologies, Weinheim, Germany: WILEY-VCH Verlag GmbH & Co. KGaA, 2007.
[14] Bröckel U, Meier W, Wagner G. Product design and engineering – best practices, Volume 2:
Raw materials, additives and applications, Weinheim, Germany: WILEY-VCH Verlag GmbH & Co.
KGaA, 2007.
[15] Grievink J, Swinkels PLJ, Van Ommen JR. Basics of Process Design. In: De Jong W, Van Ommen
JR (eds.), Biomass as a sustainable energy source for the future: Fundamentals of conversion
processes, 1st edn, Hoboken NJ, USA: Wiley & Sons Inc., 2014, 184–229.
[16] Meeuse M, Grievink J, Verheijen PJT, Vander Stappen MLM. Conceptual design of process for
structured products, AIChE Symposium Series USA, 96, 2000, 324–328.
synthesis in foods manufacture. In: Martin M, Eden MR, Chemmangattuvalappil NG, eds. Com-
puter Aided Chemical Engineering Vol. 39 Tools for chemical product design – from consumer
products to biomedicine, Amsterdam, Netherlands: Elsevier BV, 2017, 417–441.
[18] Bernardo FP. Integrated process and product design optimization. In: Martin M, Eden MR,
Chemmangattuvalappil NG, eds. Computer Aided Chemical Engineering Vol.39 Tools for chem-
ical product design – from consumer products to biomedicine. Amsterdam, Netherlands: Else-
vier BV, 2017, 347–372.
A9 Appendix to Chapter 9: Cases of process model-
ing for simulation and optimization
Johan Grievink
This Appendix presents thirteen application cases. An upfront remark about the re-
strictive use of some terms is in order. Behavior of a process or product relates to
patterns of change in its chemical and physical quantities. The term performance is
intentionally reserved for factors that express some kind of value and appreciation
by society, including customers of a product and the operators and owners of a pro-
cess. Such factors are related to safety, health, economy and ecology. The literature
references appearing in this Appendix are listed at the end, independently of those in
Chapter 9. Commonly used mathematical symbols are also listed at the end. However,
several symbols are repeatedly in use in the various cases and sometimes have a dif-
ferent meaning. Therefore, to avoid ambiguity, care has been taken for each case to
have some additional description of its symbols as well.
A9.1 Linear model applications for concept stage
Before starting the application cases, it is expedient to summarize five different math-
ematical forms of linear models, all of which occur in process modeling practice. A
model is linear when each equation in the model is linear in all variables. Determinis-
tic variables and models will be covered only, excluding stochastic and logical models.
The five common linear deterministic model forms are:
(1) linear algebraic equations: Symbols

A⋅x =b A: m ⋅ n matrix;
x: n-vector; b: m-vector
(2) linear ordinary differential equations:
dx(t)/dt = A ⋅ x(t) + B ⋅ u(t) t: scalar coordinate
x: n-vector function;
u: m-vector function;
A: n ⋅ n matrix;
B: n ⋅ m matrix
(3) linear partial differential equations:
∂x(z, t)/∂t + v ⋅ ∂x(z, t)/∂z = A ⋅ x(z, t) + B ⋅ u(z, t) t, z: scalar coordinates;
v: scalar
u: m-vector function
A: n ⋅ n matrix;
B: n ⋅ m matrix
https://doi.org/10.1515/9783110467741-017
362 | A9 Appendix to Chapter 9: Cases of process modeling for simulation and optimization
(4) linear difference equations:

xk+1 = A ⋅ xk + B ⋅ uk k: integer index,
k ∈ {0, 1, 2, . . .}
u: m-vector function
A: n ⋅ n matrix; B: n ⋅ m matrix
(5) linear integer equations with binary variables:
Z⋅Y≤ ω Z: m ⋅ n matrix; Y: n-vector;
ω: m-vector
The matrices A and B, vectors x, and u, as well as the scalars t and z are all made up of
real numbers. Matrix Z and vector ω take integer values. The entries in vector Y take
binary values {0, 1}. Matrices A, B and Z are constant, as they have fixed entries.
The application cases with linear models in this Appendix deal with algebraic
equations and real variables. An exception to this is the modeling of the structure of
a process block diagram/flow sheet involving a linear model with binary variables.
Case A9.1: Exploring the stoichiometric space for an innovative chemical conversion
Reaction stoichiometry offers freedom in the design of a process concept. A simple

example will be given on a search over a stoichiometric space to find a suitable reac-
tion that serves best a particular conversion goal. In this case the goal is the chemical
conversion of carbon dioxide as a greenhouse gas into carbon monoxide as a chemi-
cal reactant in syngas. The reuse of carbon in chemical products helps in closing the
carbon cycle.
In analogy to steam reforming [CH4 +H2 O ⇒ CO+3H2 ], one can apply dry reform-
ing with CO2 as is done in the steel industry [CH4 + CO2 ⇒ 2CO + 2H2 ]. Unfortunately,
the conversion to CO by these reforming reactions is reduced by the water-gas-shift
reaction [CO + H2 O ⇒ CO2 + H2 ], shifting CO partially back to CO2 . All reactions
take place in the gas phase at high temperature. A disadvantage of the dry reform-
ing reaction is the low molar ratio of CO2 : CH4 = 1 : 1. It would be nice to achieve a
higher ratio of CO2 to methane as a way of recycling more CO2 molecules per methane
molecule. This issue can be explored by means of modeling the possible reaction stoi-
chiometry based on a given set of components. In this modeling exercise, the set {CH4 ,
CO2 , CO, H2 , H2 O} are the reacting species, comprising of {C, H, O} as elements.
Synthesis of model
The molar reaction stoichiometry is expressed in a generalized way:
Generalized reaction: a CH4 + b CO2 ⇒ x CO + y H2 + z H2 O

There are three likely reaction products at the right hand side, excluding O2 as a po-
tential fourth candidate because the latter is not formed due to unfavorable thermo-
dynamics.
The unknown stoichiometric coefficients {a, b, x, y, z} are constrained by three
element balances:
C balance: a + b − x = 0
H balance: 4a − 2y − 2z = 0
O balance: 2b − x − z = 0
The difference between the stoichiometric coefficients of CO2 and CH4 is introduced
as a convenient, additional variable:
δ=b−a≥0
These coefficients should be:
a ≥ 1 and b≥1
The other coefficients must be nonnegative:
x≥0 y≥0 z≥0
When adopting the conventional form of writing chemical reaction equations, the sto-
ichiometric coefficients take integer values only. From a mathematical point of view
this integer condition is not strictly needed and one can take a = 1 as a pivot in the
above reaction equation.
The objective is to find the reaction equation that maximizes the CO2 consumption
relative to methane consumption:
Max∀ F = b/a
Analysis of model
This model has six variables {a, b, x, y, z, δ} with four algebraic equality relations, re-
sulting in two degrees of freedom. If a problem would have a much larger size, it can
be better numerically solved and optimized by applying linear programming (LP). Be-
cause of the small size of this problem, an analytical approach with graphical solu-
tions will be followed here. The stoichiometric coefficients of the reaction products
{x, y, z} can be expressed as linear functions of a and b:
x = a + b ≥ 0 ⇒ a ≥ 0 and b≥0
y = 3a − b ≥ 0 ⇒ 3a ≥ b
z = −a + b ≥ 0 ⇒ b ≥ a
Now the problem has been reduced to a two dimensional search space {a, b} for opti-
mization.
5 Inequality δ=3 Sector of

constraint feasible solutions
CO2 (b) Infeasible:
4 3a ≥ b High dry reforming:
δ=2 a = 2; b = 5; δ = 3 ; F = 2.5
2 CH4 + 5 CO2 => 7 CO + H2 + 3 H2O
δ=0
3 Super dry reforming:
δ=1 a = 1; b = 3; δ = 2 ; F = 3
CH4 + 3 CO2 => 4 CO + 2 H2O
2 Medium dry reforming:

a = 1; b = 2; δ = 1 ; F = 2
Infeasible: CH4 + 2 CO2 => 3 CO + H2 + H2O
b≥a
1 Dry reforming:
a = 1 ; b = 1; δ = 0 ; F = 1
Inequality CH4 + CO2 => 2 CO + 2 H2
constraint
0
1 2 3 4
CH4 (a)
Feasible domain with three stoichiometric possibilities for CO2 consumption.
The highest consumption (super dry reforming) arises for a=1 and b=3.
Fig. A9.1: Searching for optimized solutions in the stoichiometric reaction space.
Solutions of the model

Figure A9.1 depicts the {a, b} solution space with contour lines of the objective func-
tion for (e.g., δ = 0, 1, 2) and two inequality constraints {3a ≥ b; b ≥ a}. Three solu-
tions are shown in this space:
Dry reforming (a = 1 ; b =1; δ = 0)

Semi dry reforming (a = 2 ; b =3; δ = 1)
Super dry reforming (a = 1 ; b =3; δ = 2)
It is noted that the feasibility condition 3a ≥ b imposes an upper bound on the objec-
tive function:
F = b/a ≤ 3a/a ⇒ F ≤ 3
One could wonder how many feasible solutions exist for δ = 1, 2, 3 and so forth. It
is convenient to take {a, δ} as the two independent variables. The remaining stoichio-
metric coefficients become:
b=a+δ; x = 2a + δ ; y = 2a − δ; z = δ
Critical feasibility condition is:
3a ≥ b ⇒ 2a ≥ δ
Lower and upper bounds on F:

1<F<3
Tab. A9.1: Possible ranges of stoichiometry of dry reforming reactions with CO2 excess.
a b (= a + δ) x (= 2a + δ) y (= 2a − δ) z (= δ) a CH4 + b CO2 ⇒ x CO + y H2 + z H2 O b/a

1 3 4 0 2 1 CH4 + 3 CO2 ⇒ 4 CO + 2 H2 O 3.00
2 5 7 1 3 2 CH4 + 5 CO2 ⇒ 7 CO + 1 H2 + 3 H2 O 2.50
1 2 3 1 1 1 CH4 + 2 CO2 ⇒ 3 CO + 1 H2 + 1 H2 O 2.00
5 9 14 6 4 5 CH4 + 9 CO2 ⇒ 14 CO + 6 H2 + 4 H2 O 1.80
4 7 11 5 3 4 CH4 + 7 CO2 ⇒ 11 CO + 5 H2 + 3 H2 O 1.75
3 5 8 4 2 3 CH4 + 5 CO2 ⇒ 8 CO + 4 H2 + 2 H2 O 1.67
5 8 13 7 3 5 CH4 + 8 CO2 ⇒ 13 CO + 7 H2 + 3 H2 O 1.60
7 11 18 10 4 7 CH4 + 11 CO2 ⇒ 18 CO + 10 H2 + 4 H2 O 1.56
2 3 5 3 1 2 CH4 + 3 CO2 ⇒ 5 CO + 3 H2 + 1 H2 O 1.50
1 1 2 2 0 1 CH4 + 1 CO2 ⇒ 2 CO + 2 H2 1.00
What are the results of a systematic exploration of the {a, δ} space for a = {1, 2, 3,
. . . , 7} with 2a ≥ δ and δ = {0, 1, 2, 3, 4}? In fact, one finds a wide range of reaction
equations. These reactions can be ordered in decreasing ratio of CO2 consumption
(= b/a). Table A9.1 shows the more striking reactions.
The upper one is super dry reforming (b/a = 3); the bottom one (b/a = 1) is
normal-dry reforming. Theoretically, one can construct numerous reaction equations
with a decreasing b/a ratio between 1.5 and 1.0. Being at the lower end of the perfor-
mance spectrum, these reactions are of less interest.
In view of the many reactions in Table A9.1, one starts to wonder how many of
these reactions are actually independent ones from a stoichiometric point of view,
i.e., a reaction within a given set of reactions has an independent stoichiometry if
the latter cannot be construed as a linear combination of the stoichiometry of some
other reactions within this set. In this dry reforming case, the outcome is that there
are, at most, two independent reactions (a mathematical example will be given be-
low). To illustrate this outcome, let us take the dry reforming and super dry reform-
ing reactions as the independent ones. It appears that all other reactions can be writ-
ten as different linear combinations of the two independent ones. For instance, the
medium dry reforming reaction (1 CH4 + 2 CO2 ⇒ 3 CO + 1 H2 + 1 H2 O) is the sum of
the super dry reforming (1 CH4 + 3 CO2 ⇒ 4 CO + 2 H2 O) and dry reforming reaction
(1 CH4 + 1 CO2 ⇒ 2 CO + 2 H2 ) divided by two. In fact, all reactions other than the
super dry and dry reforming reactions are linear combinations of these two.
It is interesting to note that the water-gas-shift reaction [CO + H2 O ⇒ CO2 + H2 ]
is also included in the set of dependent reactions. If one subtracts the super dry re-
forming stoichiometry from the regular dry reforming stoichiometry and divides this
by two, one gets the water-gas-shift reaction.
Can it be mathematically proven there are, at most, two independent reactions?

Independence of reaction stoichiometry has its mathematical basis in linear algebra
and vector spaces. The stoichiometry of a reaction can be treated as a single vector in
a real valued vector space (ℜN ), where N is the number of chemical species between
which a specified set of reactions occurs. N is the dimension of the ‘species space’.
A vector is a dependent one if it is a linear combination of some other vectors in this
space. In (ℜN ), one can always choose a set of N independent vectors which span the
entire space. Any other vector in this space is a linear combination of these N indepen-
dent ones. To determine the number of reactions with an independent stoichiometry
(or ‘independent reactions’), one must also take into account the conservation of ele-
ments in the reactions. The coefficients in the element (C, H, O) balance equations also
make up vectors of dimension N in the species space. Let E be the element species ma-
trix for the dry reforming case:
Elements CH4 CO2 CO H2 H2 O ⇐ Species

C [ 1 1 1 0 0 ]
Matrix E = H [ 4 0 0 2 2 ]
O [ 0 2 1 0 1 ]
Each row in this matrix corresponds to a vector in the five dimensional species space.
This is called an ‘element’ vector. The above matrix E has three linearly independent
vectors in the species space, which can be easily verified; i.e., none of the rows can
be written as a linear combination of the two other rows. Because of the principle of
conservation of elements in reactions, each of these element vectors must be orthog-
onal to the reaction ‘stoichiometric’ vectors. Be reminded that a row in stoichiometric
matrix S is the stoichiometry of a reaction, while a column in S shows the occurrence
of a particular chemical species in the reactions. Due to the conservation principle,
the inner product of an element vector and a stoichiometric vector is always zero. In
matrix notation, this is as follows:
E ∗ ST = O with Nspecies ≥ Nindep.elements + Nindep.reactions
This inequality relation between the number of species and the numbers of indepen-
dent elements and reactions is a useful one in chemical reaction network analysis. In
the case of the dry reforming reactions, there are three element vectors in a five dimen-
sional species space. Only two other independent reaction vectors are left to fully span
the species space. These two vectors can be arbitrarily taken from any set of reactions
with independent stoichiometry. Here, the dry reforming and the super dry reform-
ing reactions are selected as the independent ones. This result completes the proof of
having only two reactions with independent stoichiometry. For the sake of clarity, it
is mentioned that stoichiometric dependence between reactions also implies depen-
dence between their reaction equilibrium equations through the law of mass action.
Evaluation of model and solution

The derived stoichiometric model enables the exploration of potentially interesting
and useful reactions in a given species space. The super dry reforming reaction is the
highest performing reaction for CO2 to CO conversion, corresponding to the (b/a = 3)
entry in Table A9.1. An experimental study of this super dry reforming reaction and its
underlying innovative catalytic scheme was first reported in [1] by Buelens et al. Our
stoichiometric analysis also revealed the high dry (b/a = 2.5) and medium dry reform-
ing reactions (b/a = 2) as interesting side reaction. In principle, these reactions may
occur simultaneously with super dry reforming reactions. The stoichiometric scheme
also implicitly covers the water-gas-shift reaction. Buelens et al., [1] have developed
a novel experimental conversion scheme. The shift reaction is avoided by means of
clever sequencing of catalytic reactions. In the first reaction, the methane is fully oxi-
dized to H2 O and CO2 , using an oxygen carrying solid state medium:
CH4 + 4 [O] ⇒ CO2 + 2 H2 O
All CO2 generated by the reaction and in the feed is chemically adsorbed in another
solid phase, while the gas phase reaction water is flushed out of the reaction system.
In the next step, all adsorbed CO2 is reduced to CO:
4 CO2 ⇒ 4 CO + 4 [O]
The oxygen is adsorbed by a solid medium for subsequent use in the next oxidation
cycle of CH4 , as in a chemical looping fashion. Jointly, these reactions satisfy the sto-
ichiometry of super dry reforming.
As a final step, it is useful to reflect on uncertainties and limitations playing a
role in formulating and solving of a model. In this case the model structure and the
outcome of the analysis are determined by the chosen set of chemical species and re-
actions which codetermine the set of independent reactions. In this case, one could
possibly argue a case to add solid carbon (Csolid ) along with the Boudouard reaction
(2 CO ⇒ CO2 + Csolid ) if the CO2 would have been released as a gas at lower tempera-
tures.
Case A9.2: A conversion–separation–recycle structure using lumped units
A linear algebraic model of a common reactor–separator–recycle processing structure

operating at steady state is derived and solved. This case may serve as an example on
how to tackle such cases in general.
Synthesis of process block diagram and model

The arrangement of the main units in a process and their connecting mass streams is
shown by means of a process block diagram (Figure A9.2), where each main unit is
represented as a block. The process block diagram shows:
(1 - s) 0<s<1 (s)
Recycle Split Purge
S(7) {A,B,C,P}
S(8) {A,B,C,P} S(6){A,B,C,P} Feed recovery
S(1){A} (0 < mi < 1)

Conversion: Separation:
A + B => P (ξ1) mA = 1 (1 - mi)
Fresh mB = 1- εB (< 1)
A + P => Q (ξ2)
feeds Mixing Reactor Crude mC = 1 Product
feed ηA : conversion of A product mP = εP (> 0, small)
σAP: selectivity to P mQ = 0
S(2){B,C} S(3){A,B,C,P} S(4){A,B,C,P,Q} S(5){B,P,Q}
Five components:
A: reactant (in surplus over B) P: main product
B: reactant Q: side product
C: trace inert in feed stream
Fig. A9.2: Block diagram of a conversion–separation–recycle system.
– A ‘conversion’ unit which turns, by two parallel reactions, the reactants A and B
into the main product P (by reaction 1) and a side product Q (by reaction 2).
Reactions: A+B⇒P
and A+P⇒Q
– A ‘separation’ unit recovers unconverted reactants from the products for recycling.
– A ‘split’ unit purges an inert trace component (C) in the feeds from the process so
as to avoid its accumulation through the recycle stream.
– A ‘mixing’ unit combines the recycle stream with fresh feed into a feed to the con-
version unit.
– The connecting ‘streams’ to, in between and from the process units.
This process counts five chemical species {A, B, C, P, Q} occurring in varying amounts
in seven streams {S(2) , . . . , S(7) }. The labeled streams are indicated in the block dia-
gram.
Synthesis of the model

At this stage of design and modeling, each process unit is defined by its steady state
processing functionality (e.g., conversion) rather than as a piece of equipment (e.g.,
reactor). The units in a block diagram will be modeled in an input-output fashion. The
transformations of the incoming streams (or mass) into the outgoing streams will be
captured in a lumped way by means of some parameters, without expanding yet on
detailed mechanistic workings inside. For example, a chemical conversion unit can be
characterized by the degree of conversion of a feed component as well as by selectivity
factors. The function of a separation unit can be simply represented by separation
factors. Each factor indicates how much of a component in the feed to the separator
is going to a particular exit stream of this unit.
The variables and equations for all streams and units will be specified. To avoid
modeling errors, it is important to maintain modeling clarity by keeping the notation
uniform. It is customary to work with a uniform representation of all streams and strive
for similarity in the model representation of the process units. Furthermore, a delib-
erate attempt is made to keep the model equations linear in the variables to facilitate
ease of computation and interpretation of solutions.
Streams
Each stream is characterized by a vector of molar flows (f) of the five species. When a
component is absent in a particular stream, its molar flow is set to zero.
(j)
fi : molar flow of species i [lower index] in stream j [upper index] j; j = 1, . . . , 8
Note that this process model assumes fixed stream connections of the units, i.e., the
process unit models contain the flows of these preassigned stream connections. There
is no separate flow sheet structure model in this case. Such a structure model would
allow for adjustable connections between units by being able to switch stream con-
nections on or off.
Conversion unit
One wants to favor the conversion to main product P over the formation of the side
product Q. The reaction stoichiometry shows that A is consumed by both reactions.
Therefore, it should not be in short supply relative to B, which is involved in the first
reaction only. The molar balances are given in a matrix vector notation by:
f (3) − f (4) + ST ⋅ ξ = 0
S is the stoichiometric matrix: [−1 − 10 + 10] and extents of reactions ξ : [ξ1 ]
[−100 − 1 + 1] [ξ2 ]
The conversion of a component and the selectivity to products can be related to the
extents (ξ ) of reactions in which it participates. Choosing limiting reactant A as a ref-
erence, its degree of conversion of A (ηA ) and the selectivity of reactant A to product P
(σ A P ) are related to the two extents of reaction:
(3)
Conversion: fA ⋅ ηA − (ξ1 + ξ2 ) = 0
Selectivity: (ξ1 − ξ2 ) − σ A P ⋅ (ξ1 + ξ2 ) = 0
At this stage of design, one can use the degree of conversion and the selectivity as
specified model parameters. The required numerical parameter values (or ranges for
these values) can be derived from the literature or experimental conversion data. Hav-
ing specified the parameter values, the two equations remain linear in the variables
(3)
(fA , ξ1 , ξ2 ).
Separation unit
The amounts of chemical species entering the unit are distributed (by diffusional type
of separation) over two exit streams, according to their thermodynamic preferences
for either of the exit streams. These distributions are characterized by means of a sin-
gle distribution parameter per species: 0 ≤ m i ≤ 1. This parameter m i indicates the
fraction of the incoming molar flow going to the recycle stream S(5) , while the remain-
ing fraction (1 − m i ) goes to the product stream S(4) . In this process the components A
and C will go fully to the recycle stream, along with the bulk of B and a small fraction
of P. A bit of B, the bulk of P and all of Q go to the product stream.
(6) (4) 5) (4)
fi − mi ⋅ fi = 0 and f i − (1 − m i ) ⋅ f i =0 i = {A, B, C, P, Q}
The numerical order of magnitude (or the numerical range) of the distribution coeffi-
cients must be derived from literature or experimental data.
Splitting and mixing

The entrance stream is split in two streams of identical composition. All components
are distributed by the same fraction(s) over the two exit streams. In this case:
(7) (6) (8) (6)
fi − s ⋅ fi = 0 and fi − (1 − s) ⋅ f i =0 i = {A, B, C, P, Q}
The fraction(s) is a (specified) design decision parameter.

In a mixer, two or more incoming streams are added together to the exit stream.
In this case:
(8) (1) (2) (3)
fi + fi + fi − fi = 0 i = {A, B, C, P, Q}
Analysis and computation of model solution

Having achieved a compact linear form of the model, one should apply a degree of
freedom analysis. How many unknowns do occur in this model and how many equa-
tions have been specified? The difference between the two is the degree of freedom
in the model. Regarding the unknowns, there are five species flows in eight streams
plus two extents of reactions, resulting in 42 variables. These variables can be ordered
into a single vector of unknowns x of dimension 42 in a vector space of real variables.
The number of equations adds up to: 5 + 2 (conversion unit) + 2 ⋅ 5 (separator) + 2 ⋅
5 (splitter) + 5 (mixing) = 32 equations. These equations are linearly independent,
containing no redundancies. Thus, the degree of freedom of this model is 42−32 = 10.
In order to make the model compact for a convenient solution procedure, the coeffi-
cients of the variables in these linear equations are put into a matrix AM of dimension
32 (rows) ⋅ 42 (columns). The right hand sides of the equations can be put into the
right hand side vector bM of dimension 32. In this case all right hand sides are zero
and the vector bM is the null vector. The resulting matrix vector form is nonsquare
[AM (32 ⋅ 42); x (42); bM (32)], falling short by ten degrees of freedom to finding a
solution to:
AM ⋅ x = bM
To fill the gap, one needs an external scenario about how this process is supposed to
run with target feed and/or product flows. Such an external scenario must specify the
remaining ten degrees of freedom.
Specification of an external scenario

The first step is to specify the feed streams. These streams come from elsewhere and
usually one has no direct control over its composition. Having two fresh feed streams
for the two main reactants, A and B, one may readily assume that feed stream (1) is
rich in A and stream (2) is rich in B. The molar compositions of these streams will
(1) (2) (1)
be given by known molar ratios {r i|A , r i|B }. Here r i|A and is the given molar ratio of
(2)
component i relative to main component A in stream (1), while r i|B is the given molar
ratio of component i relative to main component B in stream (2). Knowing these ratios,
one can express the species feed rates relative to the feed rate of the main component:
(1) (1) (1)
fi − r i|A ⋅ fA = 0 i = {B, C, P, Q}
(2) (2) (2)
fi − r i|B ⋅ fB =0 i = {A, C, P, Q}
These specifications of the two feed streams take away eight degrees of freedom, leav-
ing two degrees of freedom to be determined.
Now there can be several alternative scenarios. For instance:
Specify the fresh feed streams:
(1) (1)
fA = b A = given flow specification of component A in stream (1)
(2) (2)
fB = b B = given flow specification of component B in stream (2)
Set targets for the output flows of the main and side products:
(5) (5)
fP = b P = target rate of product P in stream (5)
(5) (5)
fQ = b Q = target rate of side product Q in stream (5).
Set a target rate for the main product with a ratio for the side product rate:
(5) (5)
fP = bP = target rate of product P in stream (5)
(5) (5) (5) (5)
fQ − rQ|P ⋅ fP =0 rQ|P . is molar ratio of Q over P in stream (5)
These scenario specifications can be cast into a matrix vector form as well:
AScenario ⋅ x = bScenario
Dimensions: AScenario : 10 (rows)* 42 and bScenario : 10 (rows)
(∗)
The molar ratios r i|j become elements in the AScenario matrix while the flow specifica-
(∗)
tions b ∗ are put in the right hand side bScenario vector.
Comments on model solution procedure

The 32 equations for the process units are combined with ten equations for the scenar-
ios. Together they form a square problem A ⋅ x = b. The matrix A should be regular. If
it is singular, the most likely cause is a modeling error such as having dependent mass
balances in the set of equations.
A square linear problem A ⋅ x = b can be quickly solved by any numerical linear
equation solver. It may be tempting to solve this model analytically and computing the
variables in a sequential manner. For instance, one can start by eliminating the extents
of reactions from the variable and equation sets and work only with the species flows
as unknowns. Then, one can solve for the flows of species A in the process as well
as for the flows of the other species. However, such model reductions and transfor-
mations are not effective in terms of the engineering work efforts. That is, unless the
derivations from the linear full model are reproduced again as an expensive repeat.
Maintaining the transparency of the original process model is more important to ef-
fective engineering work than sacrificing it for small potential gains in computational
speed. In engineering practice, the application of analytical model transformations to
enhance speed of computation are often counterproductive in the long run.
Evaluation of the solution and the model

The model parameters (degree of conversion, selectivity, separation factors and split
ratio), as well as the external scenario specifications can be assigned numerical val-
ues. These values will be called the reference values. One can solve the linear A ⋅ x = b
model and find the corresponding molar flows of the species (x) in the process. How-
ever, some model parameters may not be very accurate yet. At this stage of design
these may vary over a range. Similarly, the external specifications (e.g., production
rates) may show variability. As such, it makes sense to analyze the model solutions
for a wider range than only the reference values of the model parameters. However,
how does one select the parameter perturbations within their specified ranges of un-
certainties? There are at least three options:
(1) To apply local sensitivity analysis by means of small parameter perturbations

around the reference values.
(2) To apply a nonlocal sensitivity analysis by tracking a trajectory of parameters in
the feasible parameter space.
(3) Nonlocal analysis by statistical sampling in the parameter space, governed by the
probability density functions of the parameters.
Apart from the local sensitivity analysis, the two other methods are too involved to be
discussed here. A suitable reference for further reading is found in [2] and in a reactor
design application [3]. One can easily perform a numerical local sensitivity analysis.
Given a parameter reference case, one can introduce a sequence of small perturba-
tions in model parameters, {p(q) , q = 0, 1, 2, . . . , Q}, where q = 0 is the unperturbed
reference. The model expands into multiple instances of matrix A and vector b:
{A(p(q) ), b(p(q) )} , q = 0, 1, 2, . . . , Q .
One can generate a sequence of dimension free sensitivity matrices:
Ω(q) = [(x(q) − x(0) )/x(0) ]/[(p(q) − p(0) )/p(0) ] q = 1, 2, . . . , Q .
In view of the linear differences between A(q) , A(0) , b(q) and b(0) , one would expect to
have to avail a computational shortcut to easily compute x(q) from x(0) . This is not the
case. For every perturbation (q) the full set A(q) ⋅ x(q) = b(q) must be solved. However,
it is not a burden given the efficiency of numerical solvers for linear equation sets. For
further analysis, one may vertically stack all sensitivity matrices in a main sensitivity
matrix:
Ω(main) = [Ω(1 /Ω(2) / . . . /Ω(q) / . . . Ω(Q) ]
Application of singular value decomposition to Ω(main) yields the dominant singular
value(s). The corresponding right singular vectors indicate which parameters or pa-
rameter combinations are the more influential ones.
Maintaining model linearity when adding an enthalpy balance

(j)
The enthalpy flow of a stream (F H ) is computationally the product of a mass flow and
a temperature. It spoils the desired linear structure of the model equations. Yet, it is
important for process design to get some quantitative understanding of the main en-
thalpy flows and the main enthalpy sources and sinks. By means of some reasonable
assumptions about temperature levels, one can induce a linear structure for the en-
thalpy balance equations. This is done in the following way:
An enthalpy flow is obtained by taking the sum over all species in the stream of
(j)
the product of the specific enthalpy, hi (T) and the flow rate of each component fi :
(j) (j)
FH = ∑ h i (T) ⋅ f i
i=1,...,N
The specific enthalpy h i (T) accounts only for the sensible heat effects to get to the
temperature T from the thermodynamic reference condition. It excludes the enthalpy
of formation of a component from its elements at reference conditions. The product of
(j)
two process variables, temperature (T) and flow rate (f i ) destroys linearity. However,
if one would be able to fix the temperatures of the streams and process units, and so
the specific enthalpy, one can retain the linear form of the enthalpy flow variables. It
is also presumed that the linear mass balance model has been solved and the local
(j)
flows { fI } are all available for the computation of the enthalpy flows. The enthalpy
balance over the conversion unit to which a heat flow is added (Qexch ) for cooling or
heating would then read:
(3) (4)
F H − ∆H R,1 ⋅ ξ1 − ∆H R,2 ⋅ ξ2 − F H + Qexch. = 0
(3) (3)
FH − ∑ h i (T (3) ) ⋅ f i =0
i=1,...,N
(4) (4)
FH − ∑ h i (T (4) ) ⋅ f i =0
i=1,...,N
When accounting for the reaction enthalpies {∆H R,1 , ∆H R,2 } as being given, these
(3) (4)
three equations match with three additional unknowns {F H , F H , Qexch }.
This approach of using preset temperatures is justified when the effects of the en-
thalpy sources and sinks (such as a reaction enthalpy ∆H R ⋅ ξ or a phase change en-
(j)
thalpy) is much larger than the change in a stream enthalpy ∆F H due to a variation in
temperature.
Obviously, this linear approach by means of use of guesstimated temperatures
will break down for an adiabatic operation of a process unit because of the additional
specification Qexch = 0. A new variable needs to be introduced which is the temper-
ature of the unit outlet stream T (4) within this enthalpy balance. By adjusting this
temperature, the right amount of enthalpy is carried away to ensure adiabatic opera-
tion.
Case A9.3: Alternative connectivity in a block diagram by use of a structure model
The units in the preceding block diagram in Figure A9.2 are directly connected by
a stream. The behavior models of the process units include the connecting streams.
There is no distinction between the behavioral models for the units and a structural
model for describing how units are connected by streams. The behavioral and struc-
tural model are lumped together. This is disadvantageous if a process block diagram
needs to be rearranged with rearranged connections between process units. A unit
may have got into different streams. Then, the model equations of a unit involved in
a reshuffle needs to be rewritten because ingoing and outgoing streams are relabeled.
Is it logical that a change in connectivity of units is reflected in having to change the
unit models? No, because unit models are the more complex ones in the overall pro-
Direct stream connections

S(5)
Reactor Conversion: Crude
feed product Separation Product
A + B => P
unit
A + P => Q
S(3) S(4) S(6)
Reactor Reactor Separator
//
inlet outlet inlet Product
Reactor S(Sep-recycle) outlet
feed Conversion:
Separation
// A + B => P // //
S(R-out) S(Sep-in)
unit S(Sep-prod)
S(R-in) A + P => Q
/ / = stream connectivity
Stream connectivity conditions
Fig. A9.3: Stream connectivity conditions between process units.
cess model and such model rewriting is error prone and may involve quite some effort.
It would be better instead to have some changes in the stream connections only. This
preferred way of modeling can be realized at a slight expense of introducing more lin-
ear connectivity equations in the model. The principle of creating flexible connections
is explained in Figure A9.3.
The top part of this diagram shows a direct connection between units. For in-
stance, stream S(4) directly connects the conversion and separation units. The lower
part assigns generic slots to pass inlet and outlet streams to each unit. These slots can
be labeled as pertaining to a particular unit. For instance, the generic entrance slot for
a feed stream to the conversion unit is now labeled as S(R−in) , while its generic outlet
slot is S(R−out) . Similarly, the generic inlet slot of the separator is S(Sep−in) .
The model equations for the conversion unit are written in terms of streams per-
taining to its slots:
Species balances: f (R−in) − f (R−out) + ST ⋅ ξ =0

(R−in)
Conversion: fA ⋅ ηA − (ξ1 + ξ2 ) =0
Selectivity: ξ1 − σ A P ⋅ (ξ1 + ξ2 ) =0
Similarly, the separation equations in the separation unit can be written for all
species (i) as:
(Sep−recycle) (Sep−in)
Recycle stream: fi − mi ⋅ fi =0
(Sep−prod) (Sep−in)
Product stream: fi − (1 − m i ) ⋅ fi =0
The unit models have a generic form now; it is independent of their connectivity to
other units. To connect units, the outlet slot of one unit is connected to the corre-
sponding inlet slot of another unit. Making such a connection must be explicitly done
by means of stream connectivity equations. Here is the example of connecting the
stream through the outlet slot of the conversion unit with the stream through the inlet
slot of the separation unit:
Connectivity: f (R−out) − f (Sep−in) = 0
The set of all connectivity equations forms a structure model of the block diagram.
Is there a price to pay for this convenient generalization by separating behavioral
and structure models? The number of variables related to streams that connects two
process units will double, and for every additional stream variable there is an addi-
tional connectivity equation. Fortunately, the price in terms of additional computing
effort is very low. Modern numerical equation solvers are very efficient in dealing with
such simple linear connectivity equations and variables. They are internally quickly
eliminated by a solver, reducing the problem to its core structure of multi term model
equations. Overall, one can say that keeping a process model transparent to its de-
veloper and users is much more important for efficient work processes than small
amounts of additional computing costs.
Creating models of alternative block diagram structures by use of binary variables

The benefit of having explicit connectivity equations can also be exploited when in-
serting alternative process units. These alternatives exercise the same function but
with different efficiencies and costs. Let us assume that the current separation block
(now labeled as Sep-A) will be put in competition with an alternative separation unit
(labeled Sep-B). It is now possible to flexibly connect the outlet stream of the conver-
sion unit to the inlet stream of either of both separation units. This flexible connection
is realized with the help of binary variables Ysep,A and Ysep,B . A binary variable takes
a value as 0 or 1 only.
Connectivity: f (R−out) − Ysep,A ⋅ f (Sep−A−in) − Ysep,B ⋅ f (Sep−B−in) = 0
Since only one of both types of separation units can be present, this exclusivity con-
dition is expressed as a constraint on the binary variables:
Exclusivity of units: Ysep,A + Ysep,B = 1

For Y sep,A = 1, the conversion unit is connected with separation unit type A. For
Ysep,B = 1, a connection is made with separation unit type B. If the binary variables
would be treated as variables along with continuous variables in the design computa-
tions, the model would become nonlinear due to the product form of a binary with a
continuous variable in terms, such as: Ysep,A ⋅ f (Sep−A−in) . Therefore, the explicit use of
binary design variables will be postponed until the treatment of mixed integer, nonlin-
ear programming approaches. The latter are applied to formulating and solving pro-
cess synthesis and design problems in Chapter 9.4.
Case A9.4: Optimization of process performance by LP
For linear process design, optimization one has to do five things.

(a) To open up some degrees of freedom in the design model for optimization.
(b) To select a linear performance metric.
(c) To introduce linear inequality constraints on process variables to demarcate the
physical domain in which a process design remains physically feasible.
(d) To select a reference production scenario and compute its performance.
(e) To compute the optimum solution and assess its relevance over the reference so-
lution.
This will be applied to the conversion-separator-recycle (CSR) process shown in the

block diagram in Figure A9.2. The species balances give rise to a linear model with
some residual degrees of freedom.
Re (a): The feed rates of species A and B in streams (1) and (2) are set free for opti-
mization. The internal process parameters {degree of conversion, selectivity, sep-
aration fraction and split factors} are kept fixed in order to maintain model linear-
ity. The process performance can be optimized by varying an external production
scenario (i.e., feed rates).
Re (b): The selected performance metric (Eprofit ) is an economic one; the profit rate.
This rate is obtained by deducting the following from the proceeds of the products:
– costs of the feeds per unit time
– a cost term for cleaning of the purge stream
The resulting profit rate expression is:
(5) (5) (1) (2) (7)
Eprofit = fP ⋅ pP + fQ ⋅ pQ − fA ⋅ cA − fB ⋅ cB − fB ⋅ cpurge
The economic parameters {pP , pQ , cA , cB , cpurge } are, respectively, the unit sales
prices of products P and Q, the unit costs of feed A and B and the purge cost of
component B. Since a cost is associated with removing the excess of reactant B
via the purge, one may expect that an optimization will push for reduced use of
reactant B.
Tab. A9.2: Design, economic and feed stream data for CSR process.
Design parameters (symbol) Value (-) Economic parameters (symbol) Value ($/kmol)
Conversion of feed A (ηA ) 0.98 Sale price of product P (p P ) 75.0
Selectivity of A to P (σ AP ) 0.80 Sale price of product Q (p Q ) 50.0
Separation fraction A (m A ) 1 Cost of feed A (c A ) 15.0
Separation fraction B (m B ) 0.95 Cost of feed B (c B ) 30.0
Separation fraction C (m C ) 1 Cost of purging B (c purge ) 5.0
Separation fraction P (m P ) 0.05 Production parameters (kmol/s)
(5)
Separation fraction Q (m Q ) 0.0 Nominal production rate P (fP ) 0.10
Split factor (s) 0.05 Maximum sale of Q (FQ,max ) 0.02
Feeding ratios in stream (1) Ratio (–) Feeding ratios in stream (2) Ratio (–)
(1) (2)
A (rA|A ) 1.0 A (rA|B ) 0.0
(1) (2)
B (rB|A ) 0 B (rB|B ) 1.0
(1) (2)
C (rC|A ) 0 C (rC|B ) 0.02
(1) (2)
P (rP|A ) 0 P (rP|B ) 0.0
(1) (2)
Q (rQ|A ) 0 Q (rQ|B ) 0.0
Re (c): Two inequalities are imposed:

(5)
– a market constraint on the amount of Q that can be sold: FQ,max − fQ ≥ 0
(4)
– the exit flow of B from the reactor cannot go below zero: fP ≥ 0
Re (d): The base case (reference) scenario is:
(1) (2)
– fA = 0.9 ∗ fB
(5)
– fP = 0.1 [kmol/s]
Re (e): Comparing the results of the base case and the optimized case.
Table A9.2 lists numerical values of the design and economic and feed parameters in
the CSR process model.
Given this input data, the ‘mass’ balances and economic performance function
can be computed for the nominal and the optimized case. The mass balances are ex-
pressed by means of the molar flows of the five species in the process streams. The
mass balance results are presented in Tables A9.3 and the key performance data in
Table A9.4, covering both the nominal and optimized case.
The results in Table A9.3 show a substantial increase in the output rate of prod-
uct P, while side product Q is pushed to its maximum level for attainable sale. It is also
remarkable that reactant B is reacted nearly to extinction in the optimum case, thus
reducing the cost of purging excess B. Also the recycle flows (stream 8) have become
much smaller in the optimized case by better balancing the feed intakes, relative to
the base case. In the optimized case there is a slight surplus of intake of A over B in
order to push the production of side product Q to its sales limit.
Tab. A9.3: Molar flows in CSR process at nominal and optimized cases.
Species Stream Stream Stream Stream Stream Stream Stream Stream

1 2 3 4 5 6 7 8
Nominal case
A (kmol/s) 0.12546 0.00000 0.12789 0.00256 0.00000 0.00256 0.00013 0.00243
B (kmol/s) 0.00000 0.13940 0.38562 0.27282 0.01364 0.25918 0.01296 0.24622
C (kmol/s) 0.00000 0.00279 0.05576 0.05576 0.00000 0.05576 0.00279 0.05297
P (kmol/s) 0.00000 0.00000 0.00500 0.10526 0.10000 0.00526 0.00026 0.00500
Q (kmol/s) 0.00000 0.00000 0.00000 0.01253 0.01253 0.00000 0.00000 0.00000
Optimum case
A (kmol/s) 0.20020 0.00000 0.20408 0.00408 0.00000 0.00408 0.00020 0.00388
B (kmol/s) 0.00000 0.18001 0.18007 0.00007 0.00000 0.00007 0.00000 0.00007
C (kmol/s) 0.00000 0.00360 0.07200 0.07200 0.00000 0.07200 0.00360 0.06840
P (kmol/s) 0.00000 0.00000 0.00798 0.16798 0.15958 0.00840 0.00042 0.00798
Q (kmol/s) 0.00000 0.00000 0.00000 0.02000 0.02000 0.00000 0.00000 0.00000
Tab. A9.4: Comparison of nominal and optimum cases in CSR process.
(5) (5) (5)

Profit fP f Q ≤ 0.02 fB ≥ 0 Feed A Feed B ξreaction 1 ξreaction 2
($/s) (kmol/s) (kmol/s) (kmol/s) (kmol/s) (kmol/s) (kmol/s) (kmol/s)
Nominal 1.99810 0.10000 0.01253 0.01364 0.12546 0.13940 0.11280 0.01253
Optimal 4.56521 0.15958 0.02000 0.00000 0.20020 0.18001 0.18000 0.02000
The results in Table A9.4 show that the economic performance has been substantially
improved by optimization. For instance, the profit per unit of product P increased from
19.98 $/kmol P to 28.61 $/kmol. It is also seen that the optimum is indeed bounded by
two active inequality constraints. The production of Q reaches its maximum (sales)
level, while the component B is completely reacted away and has zero flow in the ef-
fluent.
Figure A9.4 shows the shift from the nominal case to the optimum point in the
plane of the two (independent) feed flows, A and B. In the nominal case, the produc-
tion of the maximum sales volume of Q is not reached yet. The optimum shifts along
the profit gradient towards the intersection of two constraints, where profit takes its
achievable maximum value.
Analysis of sensitivities of process performance for parameter changes

Suppose one would like to further improve the performance of the process. The ques-
tion is where to have the most effective of further technological or commercial devel-
opments. To address this question a parametric sensitivity analysis will be made. The
following three parameters have been chosen for further analysis of their critical role
in process performance:
Inlet flow B
Profit isoclines increase
(kmol/s)
towards the opmum
0.2
1.20 2.86
0,1800
= Opmum case:
Exit flow P = 0.1596 (kmol/s)
Exit flow Q = 0.0200 (kmol/s)
0.1394 Exit flow B = 0.0000 (kmol/s)
= Nominal case: Profit = 4.5652 ($/s)
Exit flow P = 0.1000 (kmol/s)
0.1 Exit flow Q = 0.0125 (kmol/s)
Exit flow B = 0.0136 (kmol/s)
Profit = 1.9981 ($/s)
Infeasible region: Infeasible region:

Exit flow B < 0 Exit flow Q > 0.02
0 0.1 0.2 0.3

0.1255 0.2002 Inlet flow A
(kmol/s)
Fig. A9.4: Graph of outcome of optimization of reactor-separator-recycle process.
Technological
Selectivity factor of reactions from A to P (σ A P ) because it will enhance production
of P.
Separation factor of component P (mP ) because it will also enhance output of P.
Commercial
An increase of the maximum sales volume of Q.
Starting from the optimized situation, the values of these three model parameters
are sequentially changed by 2% relative to their nominal values. With each parame-
ter change, the process design is reoptimized for maximum profit. The effect of the
parameter change on the profit is expressed as a scaled sensitivity of the profit with
respect to that parameter (p):
Spar = Pref ∗ {∆Profit/∆p}/Profitref
Selectivity factor of reactions from A to P (σ AP ): 0.11109

Separation factor P (mP ): −0.00009
Larger sales volume of side product Q: 0.01948
The conclusion from these sensitivity data is that further performance enhancement,
from among these three options, should focus on improving the selectivity of the re-
actions. However, this priority may be reversed when the cost of further technological
development well exceeds the benefits of exercising the commercial option.
In a concluding reflection on this optimization case one could argue that it is a hy-
pothetical example with nice results to show. Indeed, it is a hypothetical case with il-
lustrative results. Yet the current way of presenting and analyzing the results between
a base case and an optimized case is representative of practical applications.
Case A9.5: Inverse use of linear models for process targeting and estimation
Consider a linear algebraic process model, A ⋅ u = y, where u is the N dimensional

vector of process inputs and y the M dimensional vector of process outputs. Matrix A
(M ⋅ N) contains the internal process parameters. The matrix is nonsquare, making
direct algebraic inversion impossible. Now there are three different ways to proceed in
using such a model:
(1) input forward: given A and u, compute y. This case is straightforward and will be
skipped
(2) output targeting: given A and targets for y, compute u.
(3) estimation: given a sequence of experimental observations of inputs {u1 , u2 , . . . ,
uK } and of outputs {y , y , . . . , y } find the unknown process parameters which
1 2 K
are the elements of matrix A.
Both the targeting and the estimation are inverse problems. These are hard(er) to solve
if the numbers of inputs and outputs are different. Then one has to deal with a non-
square matrix A (M > N: more inputs than outputs or M < N, more outputs than
inputs). One wants to find solutions match in some ‘best possible’ way with the lin-
earity of the model. The best possible way has to be defined: here it will be done by
means of an additional quadratic function that aims for minimization of deviations
from model linearity. The fitting function is a quadratic function over the model resid-
uals: ε = A ⋅ u − y. The benefit of using quadratic expressions as fitting functions in
both targeting and estimation is that closed from linear analytical expressions result
in answers. By using quadratic functions, linearity returns immediately when the nec-
essary conditions for an extremum (minimum or maximum) are formulated in terms
of the first order derivatives being set equal to zero. These derivative expressions are
linear in all variables.
Targeting
One can set targets for the outputs and ask what are the inputs needed to realize these
targets. When the number of output targets exactly matches the number of adjustable
inputs, this inverse problem can be solved if the matrix A is regular (= nonsingular;
A−1 exists):
u = A−1 ⋅ y with t = target
t
What to do when there are more targets M than adjustable inputs N, (M > N)
Obviously, it will be impossible to exactly match the targets through the few inputs
available. Then a strategy could be to minimize the differences between the targets
and what the inputs can achieve on the outputs. If one takes a quadratic criterion
with a (diagonal) weighting matrix W for scaling the different outputs one obtains
the following minimization problem. It has a closed analytical expression for the best
matching u∗ :
Minimize over {u} the matching function
1
Ω= (Au − yt )T ⋅ W y−1 ⋅ (Au − yt )
2
Result:
u∗ = [AT ⋅ W−1
y ⋅ A]
−1
⋅ (W−1
y ⋅A )⋅y
T
t
The condition for (AT ⋅ W ⋅ A)−1

to exist is that matrix A must have full-row rank, mean-
ing no linear dependencies between the outputs. The residual errors (e ∗ ) of the match-
ing are:
e∗ = A ⋅ u∗ − y
t
What to do when there are more adjustable inputs N than targets M, (M < N)
In such a situation, a number of inputs will remain undetermined by the target out-
puts. That is, unless one follows a policy of minimizing the changes in the input vector
u, taking zero as the reference value. The result is again a mathematical optimization
problem. One may scale the various inputs by means of the diagonal matrix Wu .
Minimize the Lagrangian function over {u}
1 T
L= (u) ⋅ W−1 T
u ⋅ (u) − λ ⋅ (A ⋅ u − yt )
2
subject to:
(A ⋅ u − y ) = 0
t
The outcome is in a closed form for u∗ :
u∗ = W u ⋅ AT ⋅ [A ⋅ W u ⋅ AT ]−1 ⋅ y
t
The Lagrange multipliers are:
λ = [A ⋅ W u ⋅ AT ]−1 ⋅ y (λ has dim. M) .

t
A Lagrangian multiplier indicates the sensitivity of the objective function L with re-
spect to a change in the corresponding output target yt,m : λ m = dL/dy t,m .
The interpretation of the solution is that the optimized inputs take minimized
sizes while jointly meeting the output targets.
Estimation
Let be given a sequence of K experimental observations of inputs {u1 , u2 , . . . , uK } and
outputs {y , y , . . . , y }. Experiment k yields the combined measurements of inputs
1 2 K
and outputs {uk,e , y }. The goal is to obtain statistical estimates of the process pa-
k,e
rameters. These parameters are in this case the elements of matrix A, satisfying the
linear model. The usual approach to obtain parameter estimates is a weighted least
squares fitting to the experimental data. The first step is always to look into the scal-
ing of the output variables. These variables may take very different magnitudes, e.g.,
the temperature of a stream is 300 K, while the molar fraction of a trace component
in a stream can be < 10−4 . To ensure more robust solutions, it is prudent to scale all
output variables to the same order of magnitude of unity. Let the scaling factors(s) be
ordered in a diagonal matrix Dy . Similarly one is advised to scale the inputs by means
of another diagonal scaling matrix Du . Then the scaled outputs and parameter matrix
become:
y󸀠 = D−1 y ⋅y; u󸀠 = D−1
u ⋅u A󸀠 = D−1 +1
y ⋅ A ⋅ Du
The goal is to find the parameters in the linear equation that match the experimental
observations best according to some criterion.
Selecting a suitable criterion

The selection of a proper criterion is related to the nature of the probabilistic distribu-
tion of the measurement errors. Here, a simplified approach is taken by assuming that
the error distribution of the residuals of the model equation using experimental input
and output data is a multivariate, normal distribution with a mean of zero and uncor-
related spreads (i.e., the covariances of the residuals are zero). That is, the error in the
residual of one equation is probabilistic independent of the error in another equation.
Due to the scaling of the variables one can also assume that the spreads of all residuals
are the same. A quadratic fitting criterion can be used, corresponding to the quadratic
argument of a normal error distribution. If there are reasons to treat the measurement
error distributions of outputs and inputs separately, one can use a more refined total
least squares approach (see below for a formulation). For now a weighted least squares
approach is applied to the residuals of the equations, working with scaled quantities.
The objective function is defined as the sum over all experiments of the squared resid-
uals of the equations, using the (scaled) experimental data for the inputs and outputs.
One can also assign a weight per experiment {w k , k = 1, . . . K} if a different degree of
trust is put in the quality of the data per experiment. If one would put w k = 1/K, the
total sum (Ω) is scaled as the averaged squared residual over all experiments.
1 K 󸀠
Error sum: Ω= ∑ {y − A󸀠 ⋅ u󸀠k,e }T ⋅ w k ⋅ {y󸀠k,e − A󸀠 ⋅ u󸀠k,e }
2 k=1 k,e
Minimization of this sum function with respect to the elements of matrix A results in
an analytical closed expression for matrix A. This least squares solution can be written
in a compact form if the measured observations are ordered in data matrices:
Y󸀠K = [y󸀠 , y󸀠 , . . . , y󸀠 ] {M ⋅ K} data matrix

1,e 2,e K,e
U󸀠K = [u1,e , u2,e , . . . , u󸀠K,e ]

󸀠 󸀠
{N ⋅ K} data matrix
W K = diagonal[w k , k = 1, . . . , K] {K ⋅ K} weights matrix
Solution: A󸀠estim = [Y󸀠K ⋅ WK ⋅ U󸀠K T ] ⋅ [U󸀠K ⋅ WK ⋅ UTK ]−1
Matrix dimensions: (M ⋅ N) (M ⋅ N) (N ⋅ N)
This solution can be computed as long as the input information matrix [U󸀠K ⋅ W K ⋅ U󸀠K T ]
is regular; meaning that matrix U󸀠K must have rank N. This requirement translates into
the following practical conditions:
– The measurements of the components of u should be linearly independent, or in
other words: no measurement of a component should be a linear combination of
measurements of some other components.
– There should be at least N experiments (K ≥ N).
– No experiment should be a perfect linear combination of other experiments.
Having stated these requirements, the problem of matrix parameter estimation based
on squaring the residuals of the equations is resolved.
Refined estimation by total least squares

With the conventional approach to estimating the model parameters, it is common to
lump all measurement errors of input and output variables occurring in an equation
into one remaining residual error per model equation. This is conceptually a bit rough,
although computationally convenient. For those liking a more refined statistical ap-
proach, the following improvements are made:
The measurement errors of inputs and outputs are accounted for separately. Nor-
mal distributions with zero mean and covariances Σ u and Σ y are assumed.
A distinction is made between the experimental values of variables {y󸀠 ; u󸀠e } and
e
their theoretical counterparts, {y󸀠m ; u󸀠m }. The latter satisfy the linear model equations
exactly:
y󸀠 − A󸀠 ⋅ u󸀠m = 0 .
m
The relation between the experimental and model variables is given by the presence
of experimental and modeling errors lumped in the gross error vectors ε y and ε u :
y󸀠 = y󸀠 + ε y
e m
󸀠
ue = u󸀠m + εy
Given these relations, one can formulate the total least squares estimation problem in-
volving a minimization of the sum over all experiments of the squared errors between
experimental and model variables, subject to the model equations: Minimize (over A󸀠 )
1 K T
ΩTLS = ⋅ ∑ {ε ⋅ Σ−1 ⋅ ε y,k + εTu,k ⋅ Σ−1
u ⋅ ε u,k }
2 k=1 y,k y
Subject to:
y󸀠 − y󸀠 − ε y,k =0 k = 1, . . . , K
e,k m,k
u󸀠e,k − u󸀠m,k − ε u,k =0 k = 1, . . . , K

y󸀠 − A󸀠 ⋅ u󸀠m,k =0 k = 1, . . . , K
m,k
This problem objective function can be reformulated as a Lagrangian function with

Lagrangian multipliers dealing with each of the associated algebraic equations.
K
L = ΩTLS − ∑ [λ Ty,k ⋅ (y󸀠e,k − y󸀠m,k − ε y,k ) − λTu,k ⋅ (u󸀠e,k − u󸀠m,k − ε u,k ) − λTu,k ⋅ (y󸀠m,k − A󸀠 ⋅ u󸀠m,k )]
k=1
The solution of this (large scale) optimization problem is more quickly determined by
numerical means. A closed form, analytical solution exists but it contains many ma-
trix and vector operations with matrix inversions. This will require numerical support
anyway. It is also possible to find estimates for the covariance matrices of the unknown
quantities, although the expressions become very involved.
In any case, consulting a book on engineering statistics is recommended (e.g., [4])
or an expert statistician on the nature of suitable error distributions, the correspond-
ing choice of a suitable fitting function and aspects of interpretation of the estimation
results to avoid conceptual pitfalls.
Case A9.6: Notes on the use of other types of linear models
(a) Computing dynamic responses of a process (unit) by means of dynamic state

space models
Inputs to process: u(t); internal states of process: x(t); process outputs: y(t)
Model equations:
dx(t)/dt = A ⋅ x(t) + B ⋅ u(t) with x(t0 ) = x0 and t > t0

y(t) = C ⋅ x(t)
The analytical solution can be expressed in a closed form:

t
x(t) = exp{A ⋅ (t − t0 )} ⋅ x0 + ∫ exp{A ⋅ (t − τ)} ⋅ B ⋅ u(τ) ⋅ dτ

t0
In the concept stage, such dynamic modeling will be useful when changes in the feed
supply to the process are expected to occur and its response is of interest.
(b) Distributed process/product dynamics

Inputs to process: u(z, t); internal states of process: x(z, t)
Model equations:
∂x(z, t)/∂t + v ⋅ ∂x(z, t)/∂z = A ⋅ x(z, t) + B ⋅ u(z, t) for t > t0 and z < z0
Initial condition:
x(z, t0 ) = x0 (z) for t = t0 and z > z0
Boundary condition:
x(z0 , t) = xIN (t) for t > t0 and z = z0
This is a Cauchy type of problem and an analytical solution in a closed form exists.
While working with the analytical solution will give more insight, direct numerical
equation solving will be quicker. In the concept stage, this form of dynamic modeling
is useful for describing the change in the mass (or size) distribution of a particulate
product in a process unit (e.g., crystals and polymer). In these cases, vector x becomes
a scalar function x(z, t). Quantity x(z, t) is the number of particulates per unit volume
of the process unit, having mass z at time t. The quantity v is a fixed mass growth rate.
Similarly, matrix A reduces to a scalar and usually it takes a negative value (A ⇒ −a),
representing the mass decay rate of particulates. The final term, B ⋅ u(z, t), also be-
comes a scalar one, representing the distributed mass birth rate of particulates with
mass z at time t. Commonly, only particulates with small mass stand a chance to be
born, e.g., u(z, t) = u(z0 , t) ⋅ e{−λ⋅z/z0} in which u(z0 , t) is the birth rate of particulates
with minimum mass (z0 ) while λ (> 0) is a decline factor for the mass birth rate of big-
ger mass. The boundary condition x(z0 , t) = xIN (t) indicates how many particulates
of minimum mass z0 are introduced from an external source over time t. If there is a
nonzero birth rate term, u(z, t), one usually takes xIN (t) = 0. This is the same the other
way around also.
This model enables the computation of the evolution of the mass size distribution
of the particulates under different scenarios, such as different initial distributions and
the impact of changes in the physical parameters (V, A, B, u(z0 , t), λ) on the mass
distribution. It is of practical importance to know if the mass distribution becomes
a steady one over a long enough time, or that the distribution will show sustained
oscillations.
(c) Discrete time model for the dynamics of a fast moving consumer goods (FMCG)
warehouse inventory
Inputs to process at time k:
uk ; internal states at time k: x k ; Outputs: y

k
Model equations:
xk+1 = A ⋅ xk + B ⋅ uk − C ⋅ y k = 1, 2, 3, . . . , K ⇒ ∞
k
A : N ⋅ N matrix ; B : N ⋅ J matrix ; C : N ⋅ M matrix
Initial condition:
xk = 0 for k = 1
This model can be applied in the concept stage to determine the dynamics of the in-
ventory of a warehouse for FMCG. Knowing the dynamics is important for the design of
the storage capacity of the warehouse. A closed analytical expression for the solution
xk+1 can be derived but it is more cumbersome to use than doing a direct numerical
simulation.
Suppose N types of products are kept in stock in the warehouse. For each type of
product, the stored mass may be different. The mass (x i ) of each type of product n,
{n = 1, . . . , N} that is stored at instant k is accounted for in the mass hold up vector
x K = [x1,k , x2,k , . . . , x n,k , . . . , x N,k ]T . The N products are supplied to the warehouse
in J different types of packages, at every instant k. All packages contain the same N
products; however, the mass fraction of a particular product may vary over the input
packages. Each type of input package {u j } has a fixed composition of the N products.
This composition is stored in the columns of delivery matrix B (= N rows ∗ J columns).
An entry u j {j = 1, . . . , J} in the input vector u represents the mass of a package of type
J. The vector uk = [u 1,k , u 2,k , . . . , u j,k , . . . , u J,k ]T shows the masses of all supplied
packages at instant k.
In the warehouse, the products are rearranged in new packages of different com-
positions for sales purposes. The products are sold in multiple packages: there are M
different sales packages to customers. The composition of sales package m is speci-
fied by the data in column m of matrix C (= N rows ∗ M columns). There are M types
of packages leaving at any instant k. The mass of sales package m at instant k is y m,k .
The column vector y = [y1,k , y2,k , . . . , y m,k , . . . , y M,k ]T is made up of the masses of
k
all sold packages at instant k.
Product storage and handling in the warehouse leads to loss of product quality.
So, a small fraction of the products deteriorates and has to be discarded. The decay
fraction can vary per type of product but it remains constant over time. This product
decay is expressed as a fractional loss in matrix A:
A = IN − D .
IN is the identity matrix of dimension N. The diagonal matrix D (0 < D < IN ) of dimen-
sion N contains the decay fractions of the product mass stored.
It is a design question and is about how big the warehouse should be in order to
store its hold up in a steady state ⟨x⟩, given time averaged masses of the input and
output packages, ⟨u⟩ and ⟨y⟩. The answer can be given in a closed form:
⟨x⟩ = (IN − D) ⋅ ⟨x⟩ + B ⋅ ⟨u⟩ − C ⋅ ⟨y⟩

The steady state hold up is:
⟨x⟩ = D−1 ⋅ {B ⋅ ⟨u⟩ − C ⋅ ⟨y⟩}
Analysis of this solution:

The first question is what happens when a product n does not decay: D nn = 0. An
infinitely large stored mass of product n can only be avoided when the total input of
product n taken over all supply packages is exactly balanced by its total output over
all the sales packages:
B|n ⋅ ⟨u⟩ − C|n ⋅ ⟨y⟩ = 0
When D nn is small, the total supply of product n must slightly exceed the total sales
of this product in order to keep a modest storage ⟨xn ⟩:
B|n ⋅ ⟨u⟩ = C|n ⋅ ⟨y⟩ + D nn ⋅ ⟨xn ⟩
The design question about the size of the warehouse can be addressed now the total
averaged mass hold up ⟨Mwarehouse ⟩ can be derived from the typical supply ⟨u⟩ and
demand ⟨y⟩ pattern:
N
⟨Mwarehouse ⟩ = ∑ ⟨x n ⟩
n=1
The warehouse hold up should be designed larger than the average mass content be-
cause of the dynamic fluctuations in supply and demand. The dynamic model xk+1 =
A ⋅ xk + B ⋅ uk − C ⋅ y can assist in performing scenario analysis for expected varying
k
supply and demand patterns around the prior estimated steady states values ⟨u⟩, ⟨y⟩
and starting from the steady state hold up ⟨x⟩ for k = 0.
One can do further analysis of this design problem with demand and supply
treated as stochastic processes; this is beyond the scope of this section.
(d) Linear integer equations with binary variables

Model equations:
Z⋅Y≤ ω
Symbols:
Z : m ⋅ n matrix of integers ; Y : n-vector of binaries ; ω : m-vector of integers.
This type of algebraic equation relates binary variables to each other. It plays an essen-
tial role in defining structure models that numerically represent superstructure flow
sheets, harboring many alternative specific flow sheets. A small example is given as an
illustration of the use of these equations. Suppose that in flow sheet synthesis one can
apply either reactor type A or reactor type B. It is known that reactor B performs better
than reactor A (by achieving higher yields) but B is also substantially more expensive
than A. Yet reactor type A may still outperform reactor type B by placing two reactors A
in cascade. The superstructure for the conversion section is one reactor B in parallel
to two reactors A in cascade. One of the following structures has to be extracted in the
computer assisted process synthesis:
(1) one reactor B, no reactor(s) A
(2) one reactor A, no reactor B
(3) two reactors A in cascade, no reactor B
(4) exclusion: Reactor A2 cannot be placed in cascade after reactor B
Binary variables are assigned for the presence of a reactor (Y = 1: present) in the flow
sheet:
Y A1 belongs to reactor A1 ; YA2 to reactor A2 ; YB to reactor B.
The following conditions enforce the above selection options:

Y A1 + YB = 1 (either reactor A or B)
YA1 + YA2 ≤ 2 (maximum of two reactors A in cascade)
YA1 + YA2 ≤ 0 (reactor A2 only possible when A1 present)
The optional structures mentioned above are all covered as solutions to this structure
model:
(1) one reactor B, no reactor(s) A: YB = 1, YA1 = 0, YA2 = 0
(2) one reactor A, no reactor B: YB = 0, YA1 = 1, YA2 = 0
(3) two reactors A in cascade, no reactor B: YB = 0, YA1 = 1, YA2 = 1
(4) reactor A2 not in cascade after reactor B: YB = 1, YA1 = 0, YA2 = 1 is not a solution.
It requires a process design optimization effort to find out which of the three options
is the more attractive one in terms of overall process performance.
A9.1.1 Nonlinear models and process simulations for development stage
Case A9.7: Note on ill conditioned constitutive equations in design models
This case is a simple illustration of a locally ill conditioned model by the use of ill-
conceived reaction kinetic equation. It violates the computational requirement of hav-
ing bounded derivatives in the model equations everywhere in the domain of model
application. The model involves a continuous flow, well mixed tank reactor (CSTR)
with a simple isothermal reaction A ⇒ B. The reaction rate law is a commonly occur-
ring type of kinetic expression. The powers of the species concentrations in the rate
expression are allowed to deviate from the stoichiometric coefficients (Law of mass
action) and become empirical ones. Often, such a power is allowed to take any real
value (negative, zero, positive) when fitting the rate parameters to experimental ki-
netic data. Such type of rate law is applied here in the design equation for a CSTR.
Variables: x exit concentration of reactant A x≥0

xf inlet concentration of reactant A xf > 0
r reaction rate r≥0
τ residence time of reacting medium τ>0
k reaction rate constant k>0
a exponent of concentration in rate law 0≤a<1
It is assumed that exponent a takes a value less than one, but remains nonnegative.
The inlet concentration (x f ), the residence time (τ), the rate constant (k) and expo-
nent (a) are all known quantities. The outlet concentration (x) is the unknown to be
determined as a function of the residence time.
Equations:
Reactor design: m ≡ x f − x − τ ⋅ r(x) = 0
Rate law: r(x) = k ⋅ x a
Derivative of reactor equation: ∂m/∂x = −1 − τ ⋅ k ⋅ [a ⋅ x(a−1) ]
For x ⇒ 0: x(a−1) ⇒ ∞ (numerical singularity)
Indeed, if one of the powers of a species concentration in a rate law is below one,
the derivative of the rate law with respect to that species concentration becomes un-
bounded if this species concentration goes to zero. The smaller the concentration x the
smaller the reaction rate r gets, but it slows down disproportionally fast. In fact, the
rate slows down when the concentration approaches zero. The rate itself stops at zero
concentration. The derivative of the reaction rate, with respect to the concentration,
goes to infinity as the concentration goes down to zero. A numerical solution of the
reactor design equation m(x) = 0 will fail when x gets close to zero. Beyond a math-
ematical singularity it is also a physical singularity. This kind of modeling problem
appears at the lower boundary of the physically feasible domain; e.g., the boundaries
of zero concentrations. This case illustrates that constitutive laws must be formulated
with due regard to correct physical behavior in boundary situations. Otherwise, it will
bring process engineering computations in unnecessary trouble.
This particular singularity problem could have been easily avoided by a physically
more realistic modeling of the reaction rate equation:
Refined rate law: r(x) = k ⋅ x/(b + x)(1−a) with a < 1 , k > 0 , b > 0 (small)
Parameter a is the empirical exponent while parameter b is an empirical threshold
concentration (small). When the concentration x gets below this threshold value b
and would go to zero, both the reaction rate and its derivative start to level down to
zero.
Derivative: dr(x)/dx = {k ⋅ [b + x] − k ⋅ x ⋅ (1 − a) ⋅ (b + x)−a }/{(b + x)(2−a) }
It is noted that for b > 0 the derivative remains finite in
lim .x → 0: dr(x)/dx = k ⋅ b (−1+a)
For the upper limit a ⇒ 1 one gets the normal exponential decay rate: dr(x)/dx = k.
Case A9.8: Consistency of design specifications with thermodynamics
It is rather easy to overlook the (implicit) limits imposed by thermodynamics when

setting design targets for process units. If targets are set beyond the limit of thermo-
dynamic equilibrium, the computations with the process model and the thermody-
namically infeasible design specifications will always fail due to this underlying in-
consistency. The failure can show itself by lack of convergence in the computations or
yielding unphysical design parameters for the unit. It seems like the numerical solver
is failing. Actually, an infeasible problem is impossible to solve and any solver will
fail!
A very simple design example illustrates this point. In a continuous flow, well
mixed tank reactor, a single, reversible first order reaction A ⇒ B is carried out at
isothermal conditions. The reactor design question refers to what residence time is
needed in order to meet a target conversion of A. The required residence time will
determine the volume of the reactor. The reactor design model is given by the following
process design equations and process data:
[M] : CA,f − CA − τ ⋅ {k f ⋅ CA − k b ⋅ CB } = 0
CB,f − CB − τ ⋅ {k f ⋅ CA − k b ⋅ CB } = 0
Feed concentrations are known: CA,f = 1 and CB,f = 0
Known reaction rate parameters: kf = 1 and k b = 1/3
The outlet concentrations {CA , CB } and the residence time τ are the unknowns. Apart
from the process model equations [M], another equation is needed for computing
three unknowns. A design target must be added. The design target is set at 90% con-
version of reactant A, leaving 10% unconverted:
[Sperf ] : CA − 0.1CA,f = 0
The process design equations [M] are reworked to render the residence time explicit:
τ = (CA,f − CA )/{k f ⋅ CA − k b ⋅ (C f,A − CA )}
Computing the residence time with the given data yields a negative value:
τ = −0.5
This answer is a bizarre outcome; what went wrong?
Analysis of the design problem

The reaction is reversible; i.e., conversion cannot be pushed beyond reaction equilib-
rium.
The equilibrium condition is:
k f /k b = K = CB,eq /CA.eq with CA,eq + CB,eq = 1

The equilibrium concentration of A:
CA,eq = CA,f ⋅ 1/(1 + K) = 1/(1 + 3) = 0.25
When approaching reaction equilibrium, the denominator {k f ⋅CA −k b ⋅(C f,A −CA )} goes
to zero. As a result, the residence time goes to infinity and it turns negative beyond
the equilibrium condition. This infinity situation may give rise to a declaration of a
singularity in computations with a numerical tool.
The initial target of 90% conversion of A results in an exit concentration for A of
0.1. This is smaller than the equilibrium concentration of 0.25. A more realistic target
for CA at 110% of the equilibrium concentration gives a positive residence time:
CA = 0.275 > 0.25 ⇒ τ = 21.75
Setting a more lax conversion target (160%):
CA = 0.400 > 0.25 ⇒ τ = 3.0
Moving away from the equilibrium boundary reduces the residence time and vessel
volume, although at a loss of having a lower conversion. This analysis illustrates an
inherent tradeoff between size of reaction vessel and conversion, which is in fact quite
general. The resulting heuristic is:
Physical design targets should be located within a thermodynamically feasible do-
main.
The distance between a target and the thermodynamic equilibrium boundary is ne-
gotiable in design.
Case A9.9: Analysis of operability aspects in process design
The operational success of a process plant depends, to a large extent, on the quality
of its design. Therefore, operational characteristics of process should be taken into
consideration in process design. This can be done by both experimental testing and
by model based simulation of the behavior of a process design under various opera-
tional scenarios. The experimental testing of vulnerable equipment parts in a process
often focuses on the effects of active operation over a long time horizon, such as the
tear and wear of equipment over time under a range of processing conditions, as well
as loss of functionality of active agents in a process (e.g., catalysts, membranes and
solvents) by deactivation, fouling or decompositions. Such long term effects play out
on a timescale that is much longer than the total pass through time of feed material
through the entire process. These effects are often hard to observe, model and simu-
late. It involves phenomena having low amplitudes and slow rates occurring below the
threshold level of detail in a process (unit) model. Yet the effects of such phenomena
consistently push in the same direction. By accumulation, they cause deterioration of
process performance over a long time horizon (months to a year).
Model based operability analysis is suitable to study the effect of the dominant
phenomena on the behavior and performance of a process over smaller time scales,
e.g., the time scale of daily production and below. Operability analysis is needed to
assess the effects of deterministic and random changes outside and inside a process.
The changes in the surroundings of the process (e.g., feed quality, production rates,
product quality requirements and environmental conditions) can be a switch from one
steady level to another as well as dynamic fluctuations. Inside a process (model) there
will be some uncertainties in its physical parameters. The outcomes of an operabil-
ity analysis may give reason to change a design so as to make it easier to operate and
control a process plant. Conditions for proper operability are often included as condi-
tions to be satisfied (inequality constraints) in an optimization of process economics.
Without respecting such conditions, it will be hard to realize the computed economic
performance in reality by coping with an unwilling process plant. Safety analysis of a
design always ranks high in any operability analysis of the intended production prac-
tice.
Operability analysis
Operability analysis of a process has become an extensive subfield of process engi-
neering, linking process design, control and operations. It is too extensive to cover it
in any degree of detail in this section. The aim is to create awareness of the main as-
pects of operability while giving some references to pertaining literature for further
study. Marlin [5] reviews eight topics of operability from the perspective of use in pro-
cess design teaching:
Safety
Process units can fail and people operating a process make mistakes from time to time,
resulting in incidents or even accidents. Layers of protection (LOPA) can be designed
to prevent failures or to mitigate the effects of failures.
The operating window of process units and of the entire process

The operating window is the joint ranges of processing conditions (T, P, x, through-
put) for which the process (unit equipment) is supposed to function normally, accord-
ing to specifications. Are there any peculiar features of the operating window? Does it
offer enough room for flexible operation?
Flexibility and controllability of the process

A process is flexible when there is enough room to adjust some manipulated process
variables within their operating window to counteract the effects of external changes
or disturbances to keep the production goals on target. Controllability implies that
enough manipulated inputs (by means of actuators such as valves for flows) are in
place in the process to effectively keep the controlled output variables close to their
target values; e.g., set points for product quality, physical conditions (P, T) and the
rate of production. Often a distinction is made between steady state (looking at steady
state changes) and dynamic controllability (the nature of time responses by outputs
to changes in inputs). A process that is both flexible and controllable is resilient to
external changes and disturbances. Care must be taken that process integration and
intensification do not result in a loss of controllability. Integration often takes away
operational degrees of freedom in the process adversely affecting controllability, while
intensification will reduce time constants and increase speed of response.
Reliability
Process unit equipment will malfunction from time to time and the production rate
of a process and/or the product quality will be adversely affected. One can express
the reliability of a process unit by means of a probability of failure. By putting identi-
cal process units in parallel, one has backups in case of failure and the reliability of
the entire process increases. An increase in reliability of a process with more product
output per year improves income from product sales. However, this requires higher
investment costs due to the extra units, leading to a tradeoff between enhanced pro-
duction from higher reliability and additional investment costs.
Transitions
During process operations there will be transitions from one steady level of opera-
tion to another level. This can involve startup and shutdown, regeneration or clean-
ing and sterilization, switches between two successive steady state production modes
and load following. The process design should address such transitions, making sure
these can be accomplished over a practical time interval and at a reasonable cost.
Dynamic behavior
This involves the time responses of process units from inputs to outputs and aspects
of stability and runaway. These can all be affected by process design variables such
as holdup volumes and surface areas for heat and mass transfers. It also involves an
identification of multiple steady states if these would occur in the process (units) and
making a decision at which state the process should be run.
Efficiency
This includes effective use of resources in a process, including its utilities. Are there
ways to maintain product quality (avoiding production losses) and steering on profit
(avoiding loss of economic opportunity)?
Monitoring and diagnosis

This involves the continuous monitoring of processing and equipment conditions by
means of sensors. The purpose is to offer information to the process operators to as-
sist them in running the process, both short term for rapid analysis and actions and
long term for identifying slow changes in the process, such a fouling, plugging and
deactivation.
In which respects can process models and simulations contribute to operability

analysis?
Let us suppose that a process model covers steady state and dynamic situations within
the specified operating window. Then this model is applicable to all of the above op-
erability cases, except reliability and safety. Reliability is an exception as it requires
probabilistic models for failure rates. Models for reliability engineering are covered
in, among others, [6] and [7]. Safety is another exception because it often involves
process events going beyond the normal operating window. The conventional process
models are rarely tuned to such exceptional conditions beyond the normal operating
window. Separate dedicated models are required to study the physical behavior of a
process unit under abnormal conditions, such as flow reversal in pipelines, a thermal
runaway event in a reactor or flows through pressure relieve valves.
A9.1.2 Nonlinear models and process optimization for feasibility stage
Case A9.10: Constraint analysis to identify options for technological innovations
For reasons of keeping optimization complexity at bay only a case of nonlinear pro-
gramming (NLP) is presented here. If a process design model would contain both
continuous and discrete decision variables, it is assumed the discrete variables are
kept fixed during the optimization over the continuous variables. Prior to presenting a
small chemical engineering example on constraint analysis to identify options for pro-
cess improvements, some necessary optimization theory is presented to set the scene
for the application case.
Necessity conditions for an extremum in optimization

The generic format of the optimization reads as:
Minimize F(x)
over {x}
Subject to g(x) = 0
and h(x) ≥ 0
All functions are algebraic ones, and are supposed to be continuous and differentiable
at least once. When analyzing and solving optimization problems, the Lagrange func-
tion plays a key role; see [8, 9].
Lagrange function: L(x, u, v) = F(x) + g(x)T ⋅ u + h(x)T ⋅ v
u is a vector of Lagrange multipliers, while v is a vector of Kuhn–Tucker multipliers.

In a NLP optimization, one searches for the optimum point(s) in the solution space {x}
where:
(a) the gradient of the Lagrange function vanishes:
∇x L∗ = ∇x F(x∗ ) + ∇x g(x∗ )T ⋅ u∗ + ∇x h(x∗ )T ⋅ v∗ = 0 ,
(b) the equality and inequality equations are satisfied:
g(x∗) = 0 and h(x∗) ≥ 0 ,
(c) the complementary slack conditions apply:
h(x∗ )T ⋅ v∗ = 0 and v∗ ≥ 0 ,
(d) there are no linear dependencies between the equality and the active inequality
constraints.
These four conditions are jointly called the Karush–Kuhn–Tucker necessity condi-
tions [9]. These conditions need to be satisfied in solution point {x∗ , u∗ , v∗ }. Vector
u contains the Lagrange multipliers associated with the equalities g(x) = 0. Vector
v contains the Kuhn–Tucker multipliers associated with the inequality constraints
h(x∗ ) ≥ 0. A Kuhn–Tucker multiplier can be interpreted as the derivative of the ob-
jective function F with respect to active inequality constraint h i :
v∗i = ∂L/∂h i for x = x∗
The complementary slack conditions state that (a) if an inequality constraint is not
active (> 0), its corresponding Kuhn–Tucker multiplier v i is zero. For example:
h i (x∗ ) > 0 and vi = 0 .
And (b) the other way around: active inequality constraints have positive Kuhn–
Tucker multipliers:
h i (x∗ ) = 0 and v∗i > 0 .
In addition to these necessity conditions for an extremum, there exist sufficiency con-
ditions for the existence of a (local) minimum. For detailed, fundamental explana-
tions, see [9].
Constraint analysis
Suppose one has performed an optimization and a minimum is obtained:
{x∗ , u∗ , v∗ } and {F ∗ , g(x∗ ) = 0, h(x∗ ) ≥ 0}
A subset of the inequality constraints will be active in the optimum and can codeter-
mine the position of the minimum. The associated Kuhn–Tucker multipliers indicate
how much the objective function F ∗ would change if a pertaining inequality constraint
would shift a bit: h(x) ≥ 0 ⇒ h(x) − ε ≥ 0. Why is it relevant to know the effects of
inequality constraints on the objective function? Realistic process models will reflect
current limitations of the technologies applied in the process. Such limitations are
often expressed as inequality constraints in the optimization problem. These inequal-
ities delineate the domain in which a technology can trustfully be applied. The tech-
nologies that are constraining an optimum by means of active inequality constraints
can be identified by means of their associated Kuhn–Tucker multipliers. When ana-
lyzing the results from a successfully completed optimization, one can put the Kuhn–
Tucker multipliers in order of decreasing magnitude. In that way, one obtains the or-
der by which active inequality constraints are restraining the objective function. That
is, the active inequality constraint with the largest Kuhn–Tucker multiplier has the
biggest influence on the objective function of all active inequality constraints. If one
wants to enhance future performance of a process, one may decide to push away some
active inequalities and particularly those with a high impact on the performance func-
tion. One may want to improve those technologies that have one or more constraining
inequalities with a high impact on the performance. One may also widen the domain
of application and so relax tight inequality constraints and increase the performance.
The message is that an analysis of the active inequality constraints and their Kuhn–
Tucker multipliers in an optimization outcome can help to find incentives and clues
for further technology development and process innovations.
Interpretation of the results at a computed optimum of a simple reactor design case

Chemical engineering textbooks on optimization mainly present small scale, mathe-
matically motivated examples also dealing with multiplier values. However, no fur-
ther use is made of such multipliers for a posterior analysis, probably because these
mathematical examples have no application context. Therefore, a small scale chemi-
cal engineering example is presented here to illustrate the use of constraint analysis.
The example involves the optimization of a continuous flow well mixed reactor with an
isothermal reaction A ⇒ B. The density of the reaction mixture is assumed constant.
The reactor has two free design variables, both with an upper bound: reaction volume
and temperature. Therefore, there are two inequality constraints (upper bounds). The
rate of production of B is fixed. The objective of the process optimization is to maxi-
mize the profit rate. The profit rate accounts for the proceeds from product sales, minus
the cost of feed A, the cost of waste treatment of unconverted A and a capital charge
term related to the volume dependent capital investment of the reactor. The capital
investment is assumed to scale linearly with reactor volume. The reactor optimization
problem will be cast in the standard NLP format assuming the objective function will
be minimized. This can be achieved by changing the sign of the profit rate.
Optimization problem (NLP1):

0bjective function (F(x)):
Minimize F = −P (= maximize profit rate P)
Equality constraints (g(x) = 0):

Profit rate expression: P = pB ⋅ fB,out − cA ⋅ fA,feed − cwaste ⋅ fA,out
−cc.Invref ⋅ (Vreac /Vref )/(θ∗ 3600.0)
Reactor design equations: fA,feed − fA,out − V ⋅ r =0
fB,feed − fB,out + V ⋅ r =0
r − k ⋅ xA,out =0
k − k T=300 ⋅ exp{−γ ⋅ [(Tref /T) − 1]} = 0
fA,feed − {Φ ⋅ xA,feed } =0
fA,out − {Φ ⋅ xA,out } =0
fB,feed − {Φ ⋅ xB,feed } =0
fB,out − {Φ ⋅ xB,out } =0
fB,out − Bproduction =0
Inequality constraints (h(x) ≥ 0):

Reactor temperature: 310.0 – T ≥0
Reactor volume: 500.0 – V ≥0
Variables (x):
fA,feed Flow rate of A to reactor [kg s−1 ]
fA,out Flow rate of A out of reactor [kg s−1 ]
fB,feed Flow rate of B to reactor [kg s−1 ]
fB,out Flow rate of B out of reactor [kg s−1 ]
k First order reaction rate constant [s−1 ]
r Reaction rate [kg m−3 s−1 ]
P Profit rate [€ s−1 ]:
T Reactor temperature [K]
xA,out Concentration of A in effluent [kg A m−3 ]
xB,out Concentration of B in effluent [kg B m−3 ]
V Reactor design volume [m3 ]
Φ Volume flow (in and out) [m3 s−1 ]
Parameters with numerical values:

cB Price of product B 1.200 [€/kg B]
cA Cost of fresh feed A 1.000 [€/kg A]
cwaste Cost of waste processing 0.500 [€/kg A]
cc Capital charge factor 0.250 [€/(€ year)]
Bprod Production rate of B 2.0 [kg B s−1 ]
Invref Reference capital investment reactor 6.000 ⋅ 10+6 [€]
k Tref Reaction kinetic constant at Tref 2.000 ⋅ 10−3 [s−1 ]
Tref Reference temperature for kinetics 300.0 [K]
Vref Reference reactor volume 260.0 [m3 ]
xA,feed Concentration of A in feed 10.0 [kg A m−3 ]
xB,feed Concentration of B in feed 0.0 [kg B m−3 ]
γ Kinetic parameter {= Eact /(Rgas ⋅ Tref )} 50.0 [–]
Θ Hours of production per year 8000.0 [h/y]
This optimization problem has twelve variables and ten equalities, including one eco-
nomic function and nine reactor design equations. The difference between number of
variables and number of equalities leaves two degrees of freedom. These two degrees
of freedom are reactor volume and temperature and both have an upper bound.
Results of optimization
With two degrees of freedom, one can take values for T and V and easily compute all
reactor variables as well as the profit rate P. The outcome of an optimization in the
{T, V} plane is that the profit rate is constrained by the upper bounds of the reactor
temperature and volume. The numerical results are:
P∗ Profit rate 0.175 [€ s−1 ]
∗
fA,feed Flow rate of A to reactor 2.083 [kg s−1 ]
∗
fA,out Flow rate of A out of reactor 0.083 [kg s−1 ]
∗
fB,feed Flow rate of B to reactor 0.0 [kg s−1 ]
∗
fB,out Flow rate of B out of reactor 2.000 [kg s−1 ]
k ∗ First order reaction rate constant 1.003 ⋅ 10 −2 [s-1]
r∗ Reaction rate 4.012 ⋅ 10−2 [kg m−3 s−1 ]
T∗ Reactor temperature 310.0 [K]
x∗A Concentration of A in effluent 0.4 [kg A m−3 ]
x∗B Concentration of B in effluent 9.6 [kg B m−3 ]
V∗ Reactor design volume 500.0 [m3 ]
Φ ∗ Volume flow (in and out) 0.208 [m3 s−1 ]
The profit rate is higher in the infeasible area; e.g., T = 310 [K]; V = 562 [m3 ]; P =
0.177 [€ s−1 ]
The corresponding Kuhn–Tucker multipliers in this constrained optimum are:

v∗T Temperature multiplier in optimum 0.0175 [€ K−1 ]
v∗V Volume multiplier in optimum 0.0001 [€ m−1 ]
This result suggests one should aim for an increase of the upper bound of the temper-
ature.
The constraint analysis for this small scale optimization example can be carried
over to large scale optimization problems that are solved by means of numerical math-
ematical programming algorithms. Upon extracting the Kuhn–Tucker multipliers of
the active inequality constraints from the computed solution and putting these in
decreasing order of magnitude, one can identify the corresponding dominating in-
equality constraints. Relaxation of the bounds of the more dominating inequalities
will improve performance. These dominating inequalities often express underlying
process technological restrictions. Such restrictions may be removed by focused im-
provements in technology development, enhancing process performance.
Case A9.11: An example of a Pareto tradeoff between two performance metrics
A simple chemical engineering example is given of a Pareto set of design solutions

for two performance criteria. Commonly, two criteria in the realms of economy and
ecology are often used. The latter may also cover efficiency of transfer of a critical
chemical element from reactants to products.
Example
The chemical reactor problem of case A9.10 is used as a vehicle to demonstrate the
effects of optimizing two criteria simultaneously. The two criteria are the profit rate (P)
and the mass efficiency (ME) of the process. The latter indicates which fraction of the
intake mass is usefully transferred to the product. The reactor temperature and volume
are the two design variables. The results are shown in Figure A9.5. There is a region
to the left in the diagram (where the profit rate rises to its peak) where both the mass
efficiency and the profit rate get better. To the right of the peak starts the Pareto set
where an improvement in the mass efficiency is met by a decrease of the profit rate.
Case A9.12: Optimization of designs of industrial processes – gas-to-liquid (GTL)

example
Due to production scale and complexity, GTL processes are prime candidates for de-
sign optimization.
The main conversion steps in the GTL process are:
(1) Syngas manufacturing using steam (by reforming) or pure oxygen by catalytic par-
tial oxidation (CPOx) to turn methane into syngas (CO: n H2 ; 2 < n < 4).
Profit rate Mass efficiency Profit rate

( €/s ) 0.19
0.9231 0.0979
0.18 0.9474 0.1572
0.9600 0.1748
0.17
0.9677 0.1758
0.16 0.9730 0.1684
0.9767 0.1563
0.15
0.9796 0.1412
0.14 0.9818 0.1241
0.13
0.12 Region of
Pareto trade off
0.11
Region of
mutual gain
0.10
0.90 0.91 0.92 0.93 0.94 0.95 0.96 0.97 0.98 0.99
Mass efficiency ( –)
Fig. A9.5: Pareto tradeoff between profit and product mass efficiency of process.
(2) Fischer–Tropsch (FT) synthesis of wax (paraffinic hydrocarbons with some olefins
and alcohols).
Reaction stoichiometry: n CO + (2n + 1)H2 ⇒ Cn H2n+2 + n H2 O n : 1 ⇒∼ 300 .
Low temperature (200–240°C) conversion with cobalt catalyst is used for NG con-
version.
(3) Hydrocracking of the wax into a slate of transportation liquids.
The FT reactors cannot fully convert the syngas due to technological limitations of cat-
alyst and reactor equipment. The question arises of what to do with the unconverted
syngas, being enriched with light hydrocarbons (≤ C4 ) formed by FT synthesis. This is
jointly called off gas. Four possibilities exist for off gas routing:
– Short recycle over the FT reactors to control the hydrogen to carbon ratio in reactor
feeds.
– Long recycle to the syngas manufacturing section for reuse of carbon in CO and
CO2 .
– Fuel to the utility system for steam and power generation.
– Stack (a fraction of the off gas must be flared to get rid of inerts in the feeds).
The design complexity is mainly due to the interactions between the syngas manu-
facturing, the FT conversion due to the syngas recycle structure as well as the heat
exchanges between SGMU and FT sections and the utility system. The design of the
hydrocracking is similar to existing refinery designs.
The list of process synthesis decision variables for syngas manufacturing units
(SGMU) and FT sections is a long one:
– Type of SGMU, with choices between steam methane reforming (SMR), CPOx and
autothermal reforming (ATR).
– Use of multiple SGMU technologies in parallel to control syngas composition.
– Oxygen to carbon ratio in feeds to SGMU’s and their operating temperatures.
– Type of FT reactors (e.g., fixed bed, slurry and structured reactors) with cobalt
catalyst.
– Amount of catalyst in a standardized FT reactor type.
– Number of sequential stages for FT reactors and number of parallel reactors per
stage.
– Distribution of syngas from the SGMU’s over the FT stages.
– Mixing ratio of fresh and recycle syngas to control H2 : CO ratio in the combined
feed to a FT reactor stage (short recycle).
– FT reactor temperature and pressure, uniform per stage.
– Recycle fraction of off gas to the SGMU’s with distribution over these units (long
recycle).
– Fraction of off gas to the utility system for steam and power generation.
– Transfer rates of high quality thermal energy (high T) to lower quality levels in
utility system (the energy quality levels have preassigned temperature and pres-
sures).
– Production rate of liquid wax from FT section. If feed driven, intake of NG to SGMU
section.
– Recycling of FT reaction water to SMR.
In addition, there will be multiple performance metrics:

– profitability, often expressed as net present value (NPV)
– carbon efficiency, expressed as amount of carbon in product over amount of car-
bon in feeds
– energy efficiency
In view of the complicated processing structure, it is also an option to place a lower

bound on the availability and reliability of the process to assure a high enough annual
product output.
An industrial case study has been published [10] on model based, optimization
driven synthesis of a future GTL process, using profitability and carbon efficiency as
performance measures. The process units are modeled in a parsimonious way, focus-
ing on their main features of behavior. Each unit model makes use of one or more
constitutive equations (kinetics and phase equilibria). For each unit, the domain of
applicability of the constitutive equations has been outlined by means of inequality
constraints. The resulting outcome of a GTL case study shows a design region in which
both performance metrics can be improved simultaneously. However, after the peak
of the profitability there is a Pareto curve with a tradeoff between the profitability and
carbon efficiency. In a separate study [11], the operational reliability of a GTL process
was modeled and optimized. One can go even one step further by jointly optimizing
the design of a process along with planning of production and process maintenance.
This has been successfully done for a multipurpose process [12].
Case A9.13: Process performance optimization by model and experiments
Data driven process model development is sometimes cheaper and quicker than devel-
oping a (semi)mechanistic model. However, it is required that one can indeed measure
all input and output variables of a product/process, relevant to the characterization of
its behavior and performance. This case study will focus on experimental process op-
timization, using a response surface approach. The procedure that is followed, applies
equally well to the optimization of product performance.
The inputs to a process model must comprise the feed and key operating condi-
tions of the process. The inputs may also include some quantitative geometric design
decision variables as long as these can be implemented and adjusted in an experi-
mental setup; e.g., by adjusting a volume or a contact surface between two different
phases. The relevant measurable outputs are those that characterize the behavior and
performance of a process and which can be measured. It is a key condition for data
driven black box modeling that process inputs can be adjusted at will in a wide enough
region to be able to collect information rich output data. Making a design of experi-
ments will be instrumental to assure that measured output data is spread over the en-
tire output space, rather than located in a subspace. Such a broad coverage is needed
for good quality estimation of the unknown parameters in the black box model, which
in turn will determine the descriptive accuracy of the model. Now, first a sequential,
multistep procedure, represented in Figure A9.6, will be explained.
This work procedure combines modeling and experimentation in an optimization
of the process performance. After having explained the procedure, the mathematical
features of a black box model will be outlined. The objective of this work procedure
is to closely approximate the point of optimum performance of the process through
a sequence of model computations and experimentation. It is a strong feature of this
approach that the point of optimum performance point is experimentally confirmed.
The procedure gradually moves the performance of the process to the optimum in an
alternating play between:
Start: select measured inputs, outputs with ranges; define performance
Make a design of experiments (DoE).
Collect experimental process input and output data by DoE

Iniaon
Use these experimental data for stascal esmaon of model parameters
Apply model based opmizaon of performance for best inputs & outputs
Perform experiment at model predicted opmum input condions
Compare experimental and computed outputs & performance Validaon
If experimental & computed results match, declare opmum => ready
Update model parameters using most recent set of experimental data
Reapply opmizaon of performance for beer inputs & outputs

Correcon
Compare current and preceding opmizaon results
If significant differences, repeat validaon step; otherwise no progress: stop
Fig. A9.6: A work procedure for an experimental approach to optimum process performance.
(a) Prediction: let the current model ‘predict’ at which inputs the optimum outputs
are obtained.
(b) Validation: an experiment is performed at the model predicted optimum condi-
tions. If the measured and predicted performances are close enough, the optimum
is found and the search stops.
(c) Correction: if there is too much of a discrepancy between model and experimen-
tal performances the model needs adjustment. The newly generated experimental
data under (b) will be used to improve the statistical estimates of the model pa-
rameter.
The sequence (prediction, validation, correction) is needed because any model, and
specifically a black box one, remains an approximation to the real process with its
unknown features. The structure of the procedure is shown in Figure A9.6.
The work procedure and its three phases are explained below.
Initiation phase:
Step 1: Define process boundaries with measurable inputs and outputs and perfor-
mance. The selected inputs and outputs must be relevant for performance.
Step 2: Make a (statistical) design of experiments (DoE).
The objective of the DoE is to create a diverse, rich enough dataset which allows
for statistical parameter estimation with good accuracy.
Step 3: Perform experiments according to DoE plan and collect measured input and
output data.
Step 4: Apply a statistical estimation procedure to derive the first set of parameter es-
timates.
Check for poorly estimated parameters and collinearity of outputs.
Step 5: Use the model with new parameter estimates to compute the optimum condi-
tions.
This computation is in fact a model based NLP optimization, requiring a model
(available), bounds on the ranges of all inputs and outputs (given as inequality
constraints) and a performance function in input and output variables.
Validation phase:
Step 6: Perform an additional experiment with the computed (predicted) optimum in-
put conditions.
Step 7: Compare the experimental and computational data {inputs, outputs and perfor-
mance}.
This can be done, for instance, by looking at the difference per variable over all
variables.
Step 8: Check how close the computed and experimental optimum outputs and perfor-
mance are.
If all outcomes are close (some quantitative criterion may be needed), then one
reasonably assumes that the computed and the experimental results approxi-
mate the true but unknown optimum. The search is successfully finished.
If not close enough yet, a next corrective phase is entered.
Corrective phase:
The corrective phase acts on the difference between the model and the real process.
Step 9: Update the model parameters.
Use the most recent experimental data to update the model parameters by
means of statistical estimation procedure updates, preferably in a Bayesian
fashion: prior ⇒ posterior info.
Step 10: Compute the next (new) optimum conditions, using the updated parameters.
Step 11: Check if the difference between the new and preceding optimum is significant.
If so, start the next validation phase. If not so, further experimentation is likely
not going to improve the optimization results. It is advised to terminate the
procedure and reanalyze the model and the experimental approach.
It is likely that this sequence of validation corrections needs to be repeated a couple

of times.
The outcome of this procedure is a set of input conditions for which the perfor-
mance of the process is optimal and experimentally confirmed. As a side effect, a
process model is obtained, having parameter estimates that are accurate enough to
predict the outputs and the performance metric. The strength of this successive ap-
proach of gradual refinement of parameter estimates is that model predictions and
experimental data should become consistently close in the region of optimum perfor-
mance. Thus, a key benefit of this multistep approach is that it refines the accuracy of
the computed model outcomes, particularly in a region that surrounds the optimum
performance point of the process.
The process model in general will be a nonlinear multidimensional one, Y(k) =
f(U(k) ; p). Here, k is an index of an experimental situation. The structure of the equa-
tions is case dependent and should reflect as much as possible the basic understand-
ing of the how the process works. In spite of its black box nature, the model (equa-
tions) should at least respect conservation principles of mass, the chemical elements
and energy. Regarding accuracy of measurements, the conventional assumption that
inputs (U) are set with 100% accuracy is often not realistic. There are always some fluc-
tuations around the intended target settings. The assumption here is that multiple in-
puts (U) can be set with small known adjustment errors. These errors are characterized
by means of an error covariance matrix (V U ). Usually the errors in the adjustments of
the inputs are smaller than the measurement errors of the outputs (Y).
The experimental data and the associated model variables will be fitted by means
of a quadratic norm (a weighted least squares approach), consistent with the com-
mon assumption of normally distributed measurement errors. The input variables in
the model (U) are to be jointly fitted to the measured inputs as well as to model. Sim-
ilarly, the process outputs (Y) in the model must be jointly fitted to measured data
(experimental outputs) and obey the model equations. That is, the model variables
must obey the model equations exactly, while the corresponding experimental data
do not due to measurement errors and differences between model and the real pro-
cess. While inputs and outputs are constrained by the model equations and their link
to the experimental data, the remaining degree of freedoms in the fitting are the ad-
justable model parameters (p).
This approach falls into the category of response surface methods [13]. There are
two extensions to the above chosen approach:
– The usual assumption in statistical textbooks that inputs have no errors is unre-
alistic for most chemical processes and must be relaxed. The result is a practical
generalization of the least squares solution, commonly called total least squares
or ‘error in variables approach’.
– Application of a multistep sequential procedure to move to the optimum perfor-
mance of the process in an alternating play between (a) doing an experiment at
model predicted optimum conditions and (b) using the new experimental data to
correct the model parameter estimates. The purpose of the multistep approach is
to refine the accuracy of the model equations, particularly in a region that sur-
rounds the optimum performance point of the process.
The final outcome is a process model having parameter estimates that are accurate
enough to predict the optimum performance point. This point is also confirmed by ex-
periments. The strength of this successive approach of gradual refinement of parame-
ter estimates is that model outputs and experimental data should become consistent
in the region of optimum performance.
List of symbols, subscripts and abbreviations:
Symbols
A real (non)square matrix (real values as entries)
b real vector
B real (non)square matrix, often mapping process inputs to states
c cost related parameters in economic model expressions
c molar concentration
d symbol for design parameters (continuous) in a process model
E set of quantitative performance metrics for a process design
E matrix of elemental composition of a set of chemical species
E reaction activation energy
f algebraic function
f molar flow rate
F objective function in an optimization problem
g set of equality functions in an optimization problem
h mole based specific enthalpy of a component
h set of inequality functions in an optimization problem
∆HR molar reaction enthalpy
i generic index; e.g., for chemical species
j generic index; e.g., for chemical reactions
k index for discrete time stepping
k reaction rate constant
K reaction equilibrium constant
L Lagrangian function
m phase distribution factor
N dimension of a vector
p generic symbol for a set of physical parameters in a process model
pP proceeds from product sales, in economic model equations
P performance function such as a profit rate
Q heat exchange rate
r reaction rate
Rgas universal gas constant
s split factor, distributing an incoming stream into two outgoing

streams
S stoichiometric matrix
S sensitivity factor
T temperature
t time coordinate in dynamic models
t target value for a design quantity or performance function in process
design
T function expressing a metric (function) related to a target
u generic symbol for inputs to a process model
v Kuhn–Tucker multipliers in an optimization problem
V volume
w disturbance variables in a process model
W matrix of weight factor for scaling of process output variables
x generic symbol for physical states in a process model
x molar fraction of a chemical species
y set of measured outputs of a process (model)
Y generic symbol for structural design parameters (discrete) in a
process model
z generic symbol for a set of adjustable inputs to a process model
z spatial coordinate in distributed models
Z matrix taking integer values only
{α,β,γ} indices for thermodynamic phases in models
γ scaled reaction activation energy
ε deviation between computed and experimental value of a process
output variable
η degree of conversion of a feed component in a reactor
ξ extent of reaction
θ number of production hours per year
λ Lagrangian multiplier
Ω sum of squared deviations between model and experimental outputs
τ residence time
Φ volume flow
σ selectivity factor of a reaction network relative to a particular reaction
product
ℜN mathematical space of dimension N for real variables
∇ spatial derivative
Subscripts
average averaged value of a quantity
design design related
ecol ecology related
ecom economy related
oper operations related
perf performance related
ref reference value of a quantity when making a comparison
tech (process) technology related
Abbreviations
ATR Autothermal reforming
CPOx Catalytic partial oxidation (of natural gas)
CSR Conversion separation recycle (process)
DAE Differential algebraic equation set
GTL Gas to liquid
FT Fischer–Tropsch
LP Linear programming
MILP Mixed integer linear programming
MINLP Mixed integer nonlinear programming
MTBE Methyl tertiary butyl ether
NG Natural gas
NLP Nonlinear programming
SG Syngas
SGMU Syngas manufacturing units
SMR Steam methane reforming
XTL X = {biomass, coal, natural gas} to liquid fuels
[1] Buelens LC, Galvita VV, Poelman H, Detavernier C, Marin GB. Super-dry reforming of methane
intensifies CO2 utilization via Le Chatelier’s principle, Science 354(6311), 2016 449–452.
[2] Varma A, Morbidelli M, Wu H. Parametric sensitivity in Chemical Systems, Cambridge Univer-
sity Press, 1999.
[3] Haaker MPR, Verheijen PJT. Local and global sensitivity analysis for a reactor design with pa-
rameter uncertainty, Chemical Engineering Research & Design, 82, 2004, 591–598.
[4] Ogunnaike BA. Random Phenomena, fundamentals and engineering applications of probability
and statistics, CRC Press, 2010.
[5] Marlin TE. Teaching operability in undergraduate chemical engineering design education,
Computers & Chemical Engineering, 34, 2010, 1421–1431.
[6] S. Mannan. Lees process safety essentials. Hazard identification, assessment and control,
Ch.5, Elsevier, 2014.
[7] Ogunnaike BA. Random phenomena, fundamentals and engineering applications of probability
and statistics, Ch.21, CRC Press, 2010.
[8] Biegler LT. Grossmann IE, Westerberg AW. Systematic methods of chemical process design,
McGraw Hill, 1997.
[9] Biegler LT. Nonlinear programming. Concepts, algorithms and applications to chemical pro-
cesses, Ch. 4, SIAM, 2010.
[10] Ellepola JE, Thijssen N, Grievink J, Baak G, Avhale A, van Schijndel J. Development of a synthe-
sis tool for Gas-To-Liquid complexes, Computers and Chemical Engineering, 42, 2012, 2–14.
[11] Blanco H, Grievink J, Thijssen N, Herder PM. Reliability integration to process synthesis applied
to GTL processes, Computer-Aided-Process-Engineering, 33, 2014, 79–84.
[12] Goel HD, Grievink J, Weijnen MC. Integrated optimal reliable design, production, and mainte-
nance planning for multipurpose process plants, Computers & Chemical Engineering, 27, 2013,
1543–1555.
[13] Ogunnaike BA. Random Phenomena, fundamentals and engineering applications of probability
and statistics, Ch.19, CRC Press, 2010.
[14] Franceschini G, Macchietto S. Model-based design of experiments for parameter precision:
State of the art, Chemical Engineering Science, 63, 2008, 4846–4872.
[15] Klebanov N, Georgakis C. Dynamic response surface models: A data driven approach for the
analysis of time-varying process outputs, Industrial & Engineering Chemistry Research, 55,
2016, 4022–4034.
A13 Appendix to Chapter 13: Communicating
A13.1 Activity report, example
Design project: ‘name & code’

confidential
Activity Report title: Reference

AR title: ‘AR – PQ–PF/Knowledge – PSw’ ‘PQ–PF/Kn – PSw/20171015’
Version Author Date Description

V1.0 Name 15/10/2017 Text
Summary
Text
Contents
– Objectives
– Approach, methods, tools
– Results & discussion
– Conclusions & recommendations
– References
– Appendices
Objectives
Text
Approach, methods, tools

Text
Results & discussion

Text
Conclusions & recommendations

Text
References
Appendices
https://doi.org/10.1515/9783110467741-018
412 | A13 Appendix to Chapter 13: Communicating
A13.2 Stream specification (passing battery limit)

Table A13.1 shows the stream specification of the process stream passing the battery
limit.
Tab. A13.1: Stream specification (passing battery limit) [2].
Stream name: Propylene/propane

Component Units Specification Additional information
Available Used in the Design Notes (also ref. note numbers)
Propylene %wt 50–70 69.0–70.0 (1) (1) Values taken in consultation
Propane %wt 50–30 29.0–30.0 (1) with principal.
Light ends %wt <1.0 (2) (2) As ‘worst case’ scenario,
Heavy ends %wt <1.0 (2) LEs and HEs are taken as
Sulfur ppm wt <20 (3) ethane and butane
respectively.
Metals ppm wt <4 (3) (3) Contaminants not harmful for
Ethane %wt <1.0 (2) the process. Compounds not
Butane %wt <1.0 (2) included in mass balance.
Total 100.0 (4) Not yet known. Will be
provided by principal later.
Process conditions and price
Temp. °C 25 Basis: Mixture of values for
Press. Bara 8 C3=/PG and C3o.
Phase V/L/S V Delivery per pipeline/rail car/
Price Eu/ton tbd (4) Tank truck/1000 l
container/drum: indicate!
A13.3 Concepts/criteria matrix

Table A13.2 shows an example of ‘concepts/criteria matrix’, or Pugh matrix [1]. This
type of matrices support the design team’s discussion on the design alternatives, stim-
ulates creativity to further improve on the concepts and stimulates the decision mak-
ing process. It should not be used as a quantitative decision making tool. Careful se-
lection of the specifications (to avoid interrelations/overlap) and suppression of the
tendency to quantify the subjective scores to a single score for each concept is needed.
Inferior and superior concepts can be easily identified, and the remaining con-
cepts should be further developed and improved to get closer or surpass the superior
concept(s) that fulfill and/or exceed the specifications.
Tab. A13.2: Concepts/criteria matrix.
Concepts/criteria matrix Concept A Concept B Concept C

Specification 1 + o –
Specification 2 – o +
Specification 3 + – +
Specification 4 o + –
Specification 5 + – +
A13.5 Agenda and minutes of meeting (MOM) templates | 413
A13.4 Pure components properties table
Table A13.3 shows how to report pure component properties.
A13.5 Agenda and minutes of meeting (MOM) templates
Agenda template
Agenda
– Meeting information
– Meeting title:
– Project:
– Date:
– Starting time:
– End time:
– Location:
– Meeting preparation:
– Please read:
– Please bring:
– Invited attendees:
– Names, affiliation:
– Agenda items
– Attendance and introductions
– Arrangement of agenda
– Confirmation of MOM previous meeting
– Matters arising from the MOM (action points)
– Meeting topics
– Topic 1, owner, allocated time
– Topic 2, owner, allocated time
– Etc.
– Any other business
– Next meeting
Minutes of meeting (MOM) template

Minutes of meeting
– Meeting information
– Meeting title:
– Project:
– Date:
– Starting time:
– End time:
– Location:
PURE COMPONENT PROPERTIES
Component name Technological data Health & safety data
Design Systematic Formula Mol. Phase Boiling Melting Flash Liquid Vapour Auto-ignit Flammable Lower Upper LC 50 MAC LD50
weight point point point density density temp. limits Explosion Explosion in air/ v alue oral Chemical reactivity Notes
[1] [1] [1] [2] [3] [1] % by vol Limit (LEL ) Limit (UEL) water [4]
o o o o
g/mol C C C kg/m3 kg/m3 C in air % % mg/m3 mg/m3 g
Propylene Propylene C3H6 42.08 V -48.0 -185.2 -108 609.0 1.50 497.0 ? 2.0 11.1 n.a. n.a. ? Extremely flamable, (2),(5)
Propane Propane C3H8 44.09 V -42.2 -187.6 -104 585.0 1.56 450.0 ? 2.1 9.5 n.a. 1800 ? Explosive with air (2),(6)
Tab. A13.3: Pure components properties table [2].
Notes:
[1] At 101.3 kPa
[2] Density at 25 o C, unless specified otherwise Project ID number: GDP-PED-2017-1/1
[3] At 0 Co Completion date: October 1st 2017
[4] Oral ingestion in (g) for a male of 70 kg weight
[5] Density at -47 o C from H2O at 4 o C *Converting mg/m3 -->ppm & vice versa: mg/m3 to ppm calculator
[6] Density at -45 o C from H2O at 4 o C
A13.6 FOOFI list for design project presentations | 415
– Agenda items
– Attendance and introductions
– Attendees: names, affiliations
– Absentees: names, affiliations
– Arrangement of agenda
– Confirmation of MOM previous meeting
– Matters arising from the MOM (action points)
– Meeting topics
– Topic 1, owner, decisions, action points and timing
– Topic 2, owner, decisions, action points and timing
– Etc.
– Any other business
– Next meeting
– Reports, handouts presented, discussed during the meeting
– List of decisions, action points and timing
A13.6 Frequently Occurring Opportunities For Improvement

(FOOFI) list for design project presentations
This FOOFI list is composed in the form that frequently occurring errors/mistakes are
indicated, based on [2].
General
– Page numbering missing
– Slides not readable (too small font size); minimum 16 pt
– University logo missing (use standard TU Delft or company format)
– Very few pictures, tables or figures are used: this makes the presentation boring
– Use of long sentences instead of keywords (use maximum seven lines of seven
words)
– Presenter reads information from screen (mainly because information is in ‘sen-
tence form’)
– Spelling mistakes (use spell checker)
– Presentation too long (aim at 1–1.5 slides/per minute)
– Too many colors used
– (Black) text invisible on colored background. Be very selective of colors; only use
black text on white (or very light) backgrounds. White text works well on black/
dark gray or (dark) blue backgrounds.
Title page
– Project principal (owner/customer) not mentioned
– Logo of Customer (in case of external project) missing
– Names of project team members missing
– Presentation date and location missing
Table of contents
– Missing table of contents
– Present presentation chapters at slide bottom, highlighting the current chapter
Introduction
– References to data, figures, tables missing
Project objectives
– Presentation has too much detail, without highlighting objectives.
– Conclusions & recommendations are not linked to objectives.
Basis of design (BOD)

– BOD section including all aspects agreed with customer is missing, e.g.:
– Product performance specifications
– Sales volume
– Plant capacity (t/annum)
– Number of operating hours/year (including shift pattern)
– Location
– Battery limit of system to be designed
– Specifications (composition, delivery form & conditions, prices) at battery
limit of:
– products
– byproducts
– raw materials
– available utilities
– waste streams handling
– Pure component properties table (including major SHE aspects)
A13.6 FOOFI list for design project presentations | 417
Process block scheme (PBS)

– Missing
– No indication of conversion and split factors for separation blocks
– No indication of t/a, t/t values
– No indication of stream names
Process flow scheme (PFS)

– Standard symbols not used
– Not all raw materials enter at the left
– Not all products, byproducts and waste streams leave at the right
– See separate information on making process flow diagrams
– PFS not readable: split into smaller (enlarged) parts
Mass balances
– No overall mass balance given (e.g., in the form of an input/output diagram show-
ing all input and output streams, utilities and verification of total in = total out).
Tables
– Too many decimals behind decimal point
– Comma is used instead of decimal point
– Wrong units are specified: too many significant numbers are presented (copied
from computing software)
– Numbers are not adjusted ‘right’ or at the decimal point (numbers should be easily
viewed/compared)
– Units are not mentioned in the column title
– No SI-units are used (‘kcal’, ‘j’ instead of J, ‘kJ/h’ instead of ‘kW’ or ‘MW’, etc.)
Comparison of alternatives (conversion routes, separation methods, unit

operations, etc.)
– Inappropriate comparison methods can be improved easily:
– When comparing two different product and/or process design options the
comparison criteria should be mentioned in the first column, and the perfor-
mance of the alternatives should be scored for each criterion.
– The criteria should be formulated in such a way that a score of ‘+’ or ‘++’
means that the performance for that criteria is good or very good.
– The criteria should be ordered in order of importance.
– Use a 5-level scale: ‘– –’, ‘–’, ‘0’, ‘+’, ‘++’. Alternatively present actual perfor-
mance data in the cells and use a 5-scale color code: ‘– –’: red, ‘–’: orange,
‘0’: white, ‘+’: light green, ‘++’: dark green.
– This ensures that all criteria will be evaluated for each alternative and missing
information is immediately evident.
Equipment design
– Design specifications per product component or equipment unit are not clear.
– The design variables to reach the design specifications for the product or process
component are not made clear.
– It is unclear how the design variables are choosen/fixed to achieve the design
specifications of each product or process component and the whole product or
process.
Safety, health, environment, economy, technology, social (SHEETS) evaluation

– General: not a complete SHEETS evaluation given (mainly focus on economics).
– Focus on all SHEETS aspects is required by stakeholders and evaluate the de-
sign accordingly.
– On economics:
– Gross Income is not presented, or only at the very end; mention this upfront.
– No distinction made between operating costs and investment costs
– No conclusions formulated on the economical data presented and the objec-
tives
– No sensitivity analysis carried out: what are the financial risks for the project?
Conclusions and recommendations

– No reference is made to the project objectives: have they all been met?
– No presentation of the design’s strengths and weaknesses: how can it be im-
proved?
– What would be possible next steps?
– Difference between project objectives and achieved results should feed into
the recommendations).
A13.7 Equipment summary and specification sheets
Tab. A13.4: Equipment summary sheet: reactors, columns & vessels – summary [2].
Equipment Nr.: C-01 V-01

Name: P/P splitter C-01 reflux
accumulator
– Tray column Horizontal
Pressure [bara]: 20.9 / 21.5 20.8
Temp. [°C]: 50.0 / 57.0 50.0
Volume [m3 ]: 105.1 (1)
Diameter [m]: 4.5 3.5
L or H [m]: 69.0 11.0
Internals
– Tray type: Sieve Trays n.a.
– Tray number: 130 n.a.
– Fixed packing
Type: n.a. n.a.
Shape: n.a. n.a.
– Catalyst
Type: n.a. n.a.
Shape: n.a. n.a.
Number
– Series: 1 1
– Parallel: – –
Materials of Trays: SS314 CS
construction (2): Column: CS
Other:
Remarks:
(1) V-01: effective volume = 68.8 m3 for residence time of 5 minutes.
(2) SS = Stainless steel; CS = Carbon steel; n.a.: not applicable
Designers: P.P. Kolom W. Wisselaar Project ID-Number: CDP3201

H.P. Pomp R.G. Klep Date: January 1st 1998
Tab. A13.5: Equipment summary sheet: Heat exchangers & furnaces – Summary [2].
Equipment Nr.: E-01 E-02 E-03 E-04

Name: C-01 Ovhd C-01 propylene propane
condenser reboiler cooler cooler
Finned tubes Thermosyphon Fixed tube sheet Fixed tube sheet

Air cooled
Substance
– Tubes: C3 = /n − C3 n − C3 /C3 = C3 = /n − C3 n − C3 /C3 =
– Shell: n.a. L.P. Steam Cooling Water Cooling Water
Duty [kW]: 91,450 91,450 344 378
Heat Exchange
area [m2 ]: 12,760 (1) 2,850 75 62
Number
– Series: – – 1 1
– Parallel: – 6 – –
Pressure [bara]
– Tubes: 20.9 21.7 26.0 26.0
– Shell: Atm. 4.0 4.0 4.0
Temperature
In / Out [°C]
– Tubes: 50.0 / 50.0 57.0 / 57.0 50.0 / 30.0 57.0 / 30.0
– Shell: 25.0 / 41.0 190.0 / 133.0 20.0 / 40.0 20.0 / 40.0
Special Materials of Tubes: CS Tubes: CS Tubes: CS Tubes: CS
Construction (2): Shell: n.a. Shell: CS Shell: Al-Br Shell: Al-Br
Other: Plot Area:
40 × 40 m

Tab. A13.6: Equipment summary sheet: Pumps, blowers & compressors – Summary [2].
EQUIPMENT Nr.: P-01 P-02 A/B P-03 A/B

NAME: C-01 Reflux: C-01 Bottom: Propylene
Transfer:
Type: Centrifugal Centrifugal Centrifugal

Number: 2 2 2
Medium transferred: C3 = /n − C3 n − C3 /C3 = C3 = /n − C3
Capacity
[kg/s]
[m3 /s]: 0.239 0.011 0.015
Density [kg/m3 ]: 450 450 450
– Pressure [bara]
Suct. / Disch.: 20.8 / 23.0 21.7 / 26.0 20.8 / 26.0
Temperature
In / Out [°C]: 50.0 / 50.0 57.0 / 57.0 50.0 / 30.0
Power [kW]
– Theor.: 53 5 8
– Actual: 75 8 13
Number
– Theor.:
– Actual: 2 (1) 2 (1) 2 (1)
Special Materials of
Construction: MS casing MS casing MS casing
Other: Double mechanical Double mechanical Double mechanical
seals seals seals
Remarks:
(1) Tray numbering from top to bottom

Tab. A13.7: Equipment specification sheet: Distillation column [2].
EQUIPMENT NUMBER: C-01

NAME: Propylene/Propane Splitter
General Data
Service: – distillation / extraction / absorption /
Column Type: – packed / tray / spray /
Tray Type: – cap / sieve / valve /
Tray Number (1)
– Theoretical: 90
– Actual: 130
– Feed (actual): 85
Tray Distance (HETP) [m]: 0.500 Tray Material: SS314 (2)
Column Diameter [m]: 4.500 Column Material: CS (2)
Column Height [m]: 69.000
Heating: none / open steam / reboiler / (3)
Process Conditions
Stream Details Feed Top Bottom Reflux / Extractant /
Absorbent side stream
Temp. [°C] : 53 : 50 : 57 : 50
Pressure [bara] : 21.5 : 20.8 : 21.7 : 20.9
Density [kg/m3 ] : 450 : 450 : 450 : 450
Mass Flow [kg/s] : 11.9 : 6.88 : 5.02 : 107.74
Composition mol% wt% mol% wt% mol% wt% mol% wt% mol% wt%
Propylene 60.0 58.8 95.0 94.8 10.0 9.6 95.0 94.8
Propane 40.0 41.2 5.0 5.2 90.0 90.4 5.0 5.2
Column Internals (4)
Trays (5) Packing: Not Applicable
Number of Type:
caps / sieve holes / : ... Material:
Active Tray Area [m2 ]: ... Volume [m3 ]:
Weir Length [mm]: ... Length [m]:
Diameter of Width [m]:
chute pipe / hole / [mm]: ... Height [m]:
Remarks:
(1) Tray numbering from top to bottom.
(2) SS = Stainless Steel; CS = Carbon Steel.
(3) Reboiler is E-01; operates with LP steam.
(4) Sketch & measures of Column & Tray layout should have been provided.
(5) Tray layout valid for whole column.

A13.8 FOOFI list for design project reports | 423
A13.8 FOOFI (Frequently Occurring Opportunities For

Improvement) list for design project reports
FOOFI list for Design Project Reports (Concept Stage, Feasibility Stage, Final Report)
This FOOFI list is composed in the form that frequently occurring errors/mistakes
are indicated, and is based on [2].
General
– Page numbering missing
– Text not readable (too small font size); minimum 10 pt.
– University logo missing (use standard TU Delft or company format)
– Very few pictures, tables or figures are used: this makes the report very boring
– Use of very long sentences
– Spelling mistakes (use spell checker)
– Report main text too long (aim at 60 pages main text; remainder in appendices
– (Too) many colors used in figures
– (Black) text invisible on colored background: be very selective of colors; only use
black text on white (or VERY light) backgrounds; white text works well on black/
dark gray or (dark) blue backgrounds.
Front page
– Design project stage (concept stage, feasibility stage, final report) omitted from
front page
– Project principal (owner/customer) not mentioned
– Logo of customer (in case of external project) missing
– Names of project team members missing
– Report issue date missing
– Insufficient or nondescriptive key words
– Date on report review meeting missing on front page
Page numbering, table of contents, appendix table of content

– No page numbering before Chapter 1. (it should be: i, ii, iii, etc.)
– No page numbering for appendices (should be Appendix – 1 –, etc.)
– No page numbering of appendix table of contents, (it should be: Appendix – i –,
Appendix – ii –, etc.)
– No information on appendices in (main) table of contents: list the detailed con-
tents and pages of the appendices
– No ‘Appendix Table of Contents’ before Appendix 1
Chapters
– No symbol list and no list of abbreviations. These lists should be positioned before
chapter references.
– Abbreviations are not explained the first time they appear.
– No chapter references present
Tables and figures

– Tables and figures do not have titles (above table, below figure).
– Tables and figures cannot be understood without the main text. Insufficient in-
formation is provided in the title and/or figure to be able to read/understand the
table/figure.
Presentation of numbers in tables

– Too many decimals are present behind decimal point.
– Comma is used instead of decimal point.
– Wrong units are specified: too many significant numbers are presented (copied
from computing software).
– Numbers are not adjusted ‘right’ or at the decimal point (numbers should be easily
viewed/compared).
– Units are not mentioned in the column title.
– No SI-units are used (‘kcal’, ‘j’ instead of J, ‘kJ/h’ instead of ‘kW’ or ‘MW’ etc.).
Units
– SI units not used
Pure component properties table

– Table omitted from report
– Some data categories missing
Block flow diagram (BFD)

– Not all streams enter at the left and leave at the righthand side of the block scheme.
– No descriptive title of each individual stream. The name should already indicate
whether this is a liquid or a solid. Indication of the phase and whether it is a so-
lution or dispersion or solids is necessary, especially when dealing with multiple
phases.
– Individual streams do not have numbers: <1>, etc. (For referral in process stream
summary (PSS)).
– No use of bold lines for flows containing main raw materials/products
– Lines do cross; they shouldn’t, use interrupted lines.
A13.8 FOOFI list for design project reports | 425
– No mass flow rate (t/a), no mass flow rate relative to main product or feed (t/t) for
each individual stream present
– No indication of phase, T, P of entering and exiting streams (for completeness and
later use for battery limit) indication
– Missing data on total in (t/a, t/t), total out (t/a, t/t)
– Blocks represent equipment instead of tasks (heat, cool, separate, recover).
– No indication of pressure, temperature, pH of the task blocks
– The block scheme should have a 1:1 relationship to the stream summary table.
Batch cycle diagram

– Batch cycle diagram is not included (for batch process).
Process stream summary, overall component mass balance and stream heat balance
– The process stream summary is not according to the prescribed format (listing
for each stream: stream number, stream name, components names, incl. molec-
ular weights (kg/kmol), enthalpy (kW), phase (V/L/S), pressure (bara), tempera-
ture (K), mass flow rate (kg/s) and moles flow rate (kmol/s). Show overall mass
balance and stream heat balance.
– Pressure, temperature, phase and enthalpy have been forgotten.
– The stream summary table for the process block scheme or process flow scheme
should have a 1:1 relationship to these schemes: they do not.
– No overall mass and energy balance is given, (e.g., in the form of an input/output
diagram showing all input and output streams, utilities and verification of to-
tal in = total out.
– New formats are created/invented: easiest way is to download provided template
files.
(Reaction) equations
– (Reaction) equations do not have numbers (no making it difficult to refer to). (E.g.,
per chapter: (Eq. 1.1), (Eq. 2.10).
– No phases (s), (l), (g), (aq.) indicated with species in reaction equations
– For complex molecular structure formulas, no component names are given.
– For complex molecules no molecular structure formula is given (stoichiometry
cannot be verified).
Description of the design and design approach

– Principal not mentioned and what his/her interest in this design project is
– In what form will the product be delivered to what type of customers? What is/will
be the supply pattern of the feedstocks and delivery of the product?
– No information given about competitor activities and patent situation.
– No mention of the main challenges you have/will need to overcome during the
design: availability of data, uncertainties. How will you go about solving these
issues?
Basic assumptions
– No clear definition of the battery limit (e.g., By drawing envelope in block scheme).
Is storage and waste treatment included?
– No clear definition of streams entering and exiting the battery limit using the cor-
rect format (Table A13.1)
– No conclusions drawn on thermodynamic and kinetic literature data: region of
validity, accuracy, demonstrated with experimental data.
Product and process designs

– Hierarchical decomposition method from Delft Design Map (DDM) not used. You
should first view the overall supply chain, stakeholder and consumer needs and
main conversion routes, and define rather broad tasks for the process blocks.
– Design choices (at different design levels) not clear or not clearly justified.
– No clear justification for the choice of the product composition & structure, pro-
cess options. No use is made of comparing alternatives against predefined selec-
tion criteria.
– A lot of qualitative statements are made, without referring to the sources. Try to
be as quantitative as possible to accept/reject choices, effectively using your esti-
mation skills.
Comparison of alternatives (conversion routes, separation methods, unit

operations, etc.)
– Inappropriate comparison methods are applied:
– When comparing two different product and/or process design options, the
comparison criteria should be mentioned in the first column, and the perfor-
mance of the alternatives should be scored for each criterion.
– The criteria should be formulated in such a way that a score of ‘+’ or ‘++’
means that the performance for that criteria is good or very good.
– The criteria should be organized in order of importance.
– Use a five level scale: ‘−−’, ‘−’, ‘0’, ‘+’, ‘++’. Alternatively, present actual per-
formance data in the cells and use a five scale color code: ‘−−’: red, ‘−’: orange,
‘0’: white, ‘+’: light green, ‘++’: dark green.
– This ensures that all criteria will be evaluated for each alternative and missing
information is immediately evident.
Equipment design
– Design specifications per product component or equipment units are not clear.
– The design variables to reach the design specifications for the product or process
components are not made clear.
– It is unclear how the design variables are chosen/fixed to achieve the design speci-
fications of each product or process component and the whole product or process.
Safety, health, environment, economy, technology, social (SHEETS) evaluation

– General: not a complete SHEETS evaluation given (mainly focus on economics).
– Focus on all SHEETS aspects required by stakeholders and evaluate the de-
sign accordingly.
– On economics:
– Gross income is not presented, or only at the very end; mention this upfront.
– No distinction made between operating costs and investment costs.
– No conclusions formulated on the economic data presented and the objec-
tives.
– No sensitivity analysis carried out: what are the financial risks for the project?
Conclusions & recommendations

– No reference is made to the project objectives: have they all been met?
– No presentation of the design’s strengths and weaknesses: how can it be im-
proved?
– What would be possible for the next steps?
– Difference between project objectives and achieved results should feed into
the recommendations.
References
– Findings/assumptions/data, etc., are mentioned without referring to the sources.
This makes it difficult for the future reader to check the validity and is considered
plagiarism.
[1] Pugh S. Creating innovative products using total design, New York, NY, USA, Addison-Wesley-
Longman, 1996.
CH3843 – from idea to design. Department of Chemical Engineering, Delft University of Tech-
nology, Delft, Netherlands, 2016.
Index
3D plant layout plan 65 biomass 11

3P 97 biotic aspect 96
6P-model 83 birth rate of particulates 386
black box 227, 404, 406
A black box model 403
abduction 75 Block Diagram 85
action blocks 80 block diagram 376
active inequality constraints 379, 397 block flow diagram (BFD) 14, 53, 152, 424
active related 94 BoD report 348
activity report 312, 411 Bologna declaration 331
actuators 207 Bologna third cycle 332
adjacent 34 BOSCARD 83
affordable optimization procedure 185 Boston learning curves 21
agenda 413 boundaries 220
agents 207 brain writing 315
agrochemicals 349 breakthrough innovation 21
alternative options 313 breakthrough technology 21
alternative process units 376 brown field 13
analyze 82 BSc 31
annual sustainability reports 96 BSc final research project 332
anticipating sequential 88 business case 50
anticipating sequential approach 89 business risk 66
approaches 88 business skills 338
artefact 162 business strategy 18
Artificial Neural Networks [ANN] 174 business to business product 67
Aspen Plus® 340
atom efficiency 299 C
availability 67, 81, 86, 103, 133, 208, 255, 402 capacity exponent 248
capital cost 247
B – estimate 247
back-mixed 63 – order of magnitude 247
‘back-of-the-envelope’ calculations 173 carbon cycle 10, 362
backups 394 carbon efficiency 226
base case solution 214 cash flow 232
basic design data 37, 54 catalyst lifetime 61
basis of design (BOD) 316, 347, 416 catalyst particle 16
batch cycle diagram 425 catalysts 179
batch process 62, 209 Cauchy type 386
batch processing 145 causal flow 221
behavior 198, 200, 201, 361 CEPCI index 249
behavioral models 180 characteristic times 170
belief 96 check implications of design decisions (C) 90
bill of materials (BOM) 241 chemical conversion 202
binary variables 388 chemical looping 367
biodiesel 179 chemical plant 14
https://doi.org/10.1515/9783110467741-019
430 | Index
chemical product 5, 19, 21, 35, 85, 112, 161, 162, contract 67
165, 166, 179, 182, 202, 205, 299, 329, 330, Contract Research Organization (CRO) 31
334, 338, 339, 344–346, 348, 349, control set points 207
356–358, 362 controllability 393, 394
chemical product design 179, 331, 344, 348, cooperation agreement 346
349, 357, 358 copy of an existing process 28
chemical reactions 175 core 34
circular economy 22, 108 corrective phase 405
circumstances 51 corrosion 59
cleaning 212, 394 corrosion rate 62
clear design structure 90 corrosion tests 102
coating process 227 cosmetics 182, 349
cold flow model 63 cost estimation 231
collection 161 covariance matrix 213, 216
commissioning and startup instruction manual cradle-to-cradle 109
65 creating flexible connections 375
communicating 311 creating value 51
communication tool 165 creation of structure 219
company 51 creative activity 77
compartments 207 creativity methods and tools 82
complementary slack conditions 396 criteria 77, 151
composition 166 critical success factors 52, 70
computational fluid dynamics (CFD) 63, 350 critical to quality variables 52
computer-aided molecular design (CAMD) 184 crude source 61
concept 199 crystallization 179
concept stage 4, 51, 385–387 cumulative cash flow 234
concept stage gate 316 cumulative discounted cash flow 234
concept stage report 316 customer evaluations 58
concepts/criteria matrix 412 customer value proposition 52
conceptual design 339
concurrent product-process design 98 D
connecting streams 374 data 128
connectivity 203, 374 – basic 128
consistent models 161 – chemical kinetics 130
consortia 337 – economic 130
constrained optimum 400 – physical-chemistry 130
constraint analysis 224, 397 – Safety, Health and Environmental 131
construction 64 data driven 206, 226, 227, 230, 403, 410
construction materials 59 data driven black box modeling 403
consumer ⇒ property ⇒ process 167 data driven black box models 224
consumer (or quality) function 166 data mining techniques 174
consumer perception requirements 53 database 129
consumer product 14, 67 DDM 160, 340, 426
contingency 245 decision matrix 313
continuous processes 209 decision ranking 94
continuous processing 145 default design activity (D) 90
continuous product variables 184 degree of conversion 370
continuous production 212 degree of freedom 211
continuously extensive 174, 175 degree of freedom analysis 370
Index | 431
degrees of freedom 167 direct costs 237

Delft Design Map (DDM) 37, 78, 80, 83, 88–90, discounted cash flow 234
94, 160, 188, 311, 340, 355, 358, 359, 426 discovery stage 18, 48
Delft Process Technology Institute (DPTI) 333 discovery stage gate 50
Delft skyline debates 24 discovery, concept, and feasibility 38
demonstration process 66 discrete (binary) variables 220
depreciation 233, 240 discrete components 175
derivative 390 dispersed 175
design 75 dispersion of a second phase 63
– boundaries 124 distributed process unit 209
– constraint 125 distribution coefficient 370
– context 124 distribution parameter 370
– goal 123 distributive properties 180
– specifications 125 disturbances 207, 213
design action 155 DOE (Design of Experiments) 174
Design Activity Space (DAS) 79, 80, 88 domain of validity 220
design alternatives 136 DOW Fire & Explosion Index 87
design approaches 77 downscaled 59, 60
design concepts 77 downscaled microplant 55
design cotargets 220 down-scaled pilot plant 102
design decisions 166 DPTI 333
design driven innovation 39 drawing 97
design for six sigma 121 drawing to scale 172
design level 79, 88 drawings 313
design methods 31 drawings/sketches 165
design models 52 droplet 16
design optimization problem 220 dry reforming 202, 203, 362, 364–367
design question 387 dynamic modeling 385
design scenario specifications 214 dynamic process 206, 218, 227
design situation 211 dynamic process modeling 218
design steps 78, 88, 136 dynamic process system 210
design targets 214, 220 dynamic scenario 213
design thinking 52, 75, 78 dynamic simulation models 61
design variables 211 dynamics of the inventory 387
detailed design 102
detailed design stage 64 E
detailed engineering 42 early innovation stages 42
detailed process design 64 ecological metrics 210
detailed process engineering stage 64 economic and ecological metrics 216
deterministic 209, 216 economic indicators 235
develop 50 economic metrics 210
development 199 economic performance function 378
development stage 18, 56 economic potential 51, 235
development stage gate evaluation 64 economics 35
device 174 educators 3
DF&EI – Dow Fire & Explosion Index 210 effective and inventive use 227
diagrams 315 eigenvalues 173
differentiable 214 element balances 363
diffusion of species 175 element species matrix 366
432 | Index
elementary system 16 feasibility 54, 199

Ellen MacArthur Foundation 108 feasibility stage 18, 54, 219
embryonic projects 33 fed-batch 147
emulsification 179 feedstock cost 153
emulsions 175 FEL-2 65
end-of-life stage 44 FEL-3 65
Energy Efficiency Ratio (EER) 21 Fermi problems 173
Engineering Procurement Construction (EPC) fictitious optimization result 222
contractor 19 fields of information 94
engineering statistics 385 final substages 83
Engineering, Procurement, and Construction fine chemicals 56
(EPC) 62 fingerprint DDM tables 88
enthalpy balance equation 373 fingerprint tables 37, 90
enthalpy flow 373 first principles product models 187
entire solution space 219 Fischer–Tropsch synthesis 202
entry barrier 36 fishbone 315
environmental, social, and governance (ESG) 23 fitting function 381
EPC contractor 64 five different mathematical forms of linear
equipment companies 32 models 361
equipment cost 253 flexibility 81, 87, 393
equipment detailed specifications 65 flow chemistry 56
Equipment Specification sheets 87 flowsheet simulators 186
Erasmus program 331 fluxes 186
estimated parameters 223 FMCG 387
estimation 205, 206, 216, 381, 383–385, foams 175
403–405 FOOFI list 316, 415, 423
evaluate 82 framing 83
evaluating 157 front-end loading 1 (FEL-1) 18, 65
evaluation and interpretation 221 front-end loading 2 (FEL-2) 19
exclusivity condition 376 front-end loading 3 (FEL-3) 18
exergy losses 210 fuels 182
existing product 27 fugitive emissions 303
experience 99 function integration 15
experienced developer 97 functional (task) block 53
experimental optimization 206, 226 functional unit 250, 300
experimental process optimization 403
experimental validation 216 G
experimentation 94 gas to liquid 224, 226, 229, 400, 409
Explosion Index 272 gelatin 179
Exposure Index 272 gels 175
extents of reaction 370 general cost 240
exterior surface of a product 178 geometric design variables 212
external disturbance 217 global system analysis 217, 219, 230
external scenario 211 global warming 8–10, 298, 302
external world 186 Go/No Go 88
gPROMS 181
F gPROMS-FORMULATE 182
failure rates 40, 395 gradual refinement 407
fats, oils and waxes 179 graduate school 334
Index | 433
gray box 206, 227 innovation partners 17

green chemical engineering 345 innovation platform 29
green chemistry 345 innovation project 3, 43
green field 13 innovation strategy 34
grossly underestimated 68 input information matrix 384
group design project (GDP) 344 input specifications 213
GTL process 226, 400 input-output 369
guidelines 4 Input-Output Diagram 85
instability 173
H instable behavior 221
HAZOP 165, 268 integrated approach 4
heat transfer 63 integrated product 54
heuristics 4, 99, 137, 143, 144 integrated product and process modeling 208
higher capacity 28 intellectual property 48, 156, 337
honors program 332 intellectual property contracts 30
House of Quality (HoQ) 53, 84, 140, 166, 313 interaction forces 179
human factors 119 interface 175
human factors engineering (HFE) 119 intermediate product 14
hydrodynamics 62 intermediate report 348
hypothetical product 184 internal physical state 186
internal rate of return (IRR) 236
I intuition 39, 100
I/O structures 80 intuition barrier 100
idea storage and retrieval system 51 in-use processes 167
ideation stage 35 inverse problem 211, 222, 381
IDP 346 invert the model 221
ill conditioned model 389 investments 38
imagine 97 investors 66
implementation stages 42 IPA database 41
indirect costs 240 ISBL (inside battery limits) 243
individual design project (IDP) 335, 346
J
industrial ecology 108
JACOBIAN 181
industrial researchers 3
Jacobian matrix 173, 214
industrial symbiosis 12, 13, 109, 343
inequality constraints 204 K
infeasible design specifications 391 Karush–Kuhn–Tucker necessity 396
infinity 390 Kepner and Tregoe 152
information flows 207 key design focus (K) 90
inherently safer design (ISD) 345 Key Performance Indicators (KPIs) 212
inhibition factor 223 kick off 347
in-house innovation 29 kinematic aspect 96
Initialization 223 knowledge 82
initiation phase 404 knowledge gaps 51
innovation 20 knowledge of designing 77
innovation ambition matrix 34 Kuhn–Tucker multipliers 397, 400
innovation class 25
innovation Class 3 34 L
innovation Class 4 35 lab-bench scale research program 53
innovation Classes 5 and 6 36 lab-on-a-chip 56
434 | Index
laboratory testing 101 mass and heat transfer 170

Lagrangian function 221, 382, 385 mass balance 378
Lagrangian multiplier 382, 385 mass efficiency (ME) 299
Lang factors 255 mass efficiency of the stoichiometric reaction
large amount of product 59 (MEST) 299
laundry powder 179 mass size distribution 386
layer of protection analysis (LOPA) 345 mass transfer 63
layers of protection (LOPA) 393 mass transfer units 59
LCA 299 material suppliers 55
learning curves 21 mathematical models 160, 162
Leiden University 331 mathematical singularity 390
length scales 172 mathematical uncertainty analysis 217
life cycle assessment (LCA) 23, 44, 343 matrix parameter estimation 384
life cycle phases 170 matter 174
Life Science & Technology (LST) 331 maximum projected loss 39
limitations 367 mechanically stirred reactors 57
linear algebraic model 203 meeting rooms 314
linear black box 206 methods 4
linear combination 365 microelement 16
linear dependency 173 microfluidics 56
linear deterministic model 361 microheat exchangers 56
linear models 199 micromixers 56
linear process design 377 micromixing 63
linear programming 204, 229, 363, 409 microplant 55
– mixed integer 409 microprocess products 5
linear structure 373 microreactors 56, 146
linguistic model aspect 96 microstructure 166
local sensitivity analyses 211 milestone meetings 315
location factors 250 milestones 347
logical aspect 96 mindmap 315
logistics 67 miniplant 58, 59
long period 59 minutes of meeting (MOM) 312, 413
long time horizon 392 mixed integer, nonlinear optimization problem
loss of controllability 394 (MINLP) 221
loss of functionality 392 mixing 170
loss of product quality 387 mockup 63
loss prevention in process design 345 modal aspect 305
lubricants 182 modal aspects of reality 95
lumped units 203, 367 model applicability 222
model based operability analysis 393
M model residual 381
maintenance costs 238 model supported operability analysis 217
management of innovation 17 model validity 222
managers 3 model-based optimization 183
manufacturing process 181 modeling methods 31
Manufacturing Readiness Level (MRL) 46 modeling of freezing ice cream 186
market launch tests 60 modeling platforms 223
market planning 67 modeling work procedure 199
mass (or size) distribution 386 models 315
Index | 435
models for process innovation and optimization O

227 objective function 220
modes of operations 212 open innovation 29, 46
modified process 27 operability 218
module cost 252 operability analysis 218, 393
molar balances 369 operate continuously 59
molar flows 369 operating labor 238
molar reaction stoichiometry 362 operating window 218, 393, 395
operation manual 70
Molecular Science & Technology (MST) 331
operational conditions 212
momentum transfer 63
operational success 392
monitoring 81, 87, 345, 395
optimism 47
MSc 31
optimization 197, 219
MTBE synthesis 226
optimization features 219
multicultural differences 342
optimization of processes 223
multilevel distribution 175 optional design focus (o) 90
multiple performance metrics 402 order of magnitude estimations 172
multiple solutions 215 OSBL (outside battery limits) 244
multiple steady states 394 oscillations 386
multiplicity 215 oscillatory behavior 173
overdesign factors 217
N P
nature 77 paints 182
need 50 panel 67
net present value (NPV) 236 panel test 67
network 7, 13, 82, 110, 117, 119, 198, 201, 207, parameter estimates 407
208, 322, 325, 340, 341, 350, 351, 366, 408 parameter estimation 213
network structure 200 parameters 213
new catalyst 69 Pareto 87, 220, 224, 225, 302, 400, 401
new competences 37 Pareto curve 403
new customers 35 Pareto frontier 226
new market development 66 Pareto set 400
new markets 35 Pareto tradeoffs 224
particles 179
nodes 207
particulate product 386
nondisclosure agreement 346
passive related 94
nonlinear models 199, 215
pastes 179
nonlinear process model simulations:
patents 48
contributions to a development stage 209
pay back time 235
nonlinear programming (NLP) 225, 229, 377,
PDEng 31, 329
395, 409
PDEng programs 334
– mixed integer 229, 356, 377, 409 People 97
– multi objective 225, 229 performance 198–201, 204, 205, 210–222,
nonlinearity 221, 223 224–229, 361, 365, 377–380, 389, 392,
nontechnical skills 347 393, 397, 400, 403–409, 416
normal-dry reforming 365 performance criteria 400
novel technologies 46 performance functions 99
novelty is not recognized 69 performance metrics 180, 201
436 | Index
peroral products 179 process flow diagrams/schemes (PFD/PFS) 163

personal care products 349 process flow scheme (PFS) 417
personal development plan 334 process flow sheet (PFS) 315
personal improvement plan 342 process flowsheet simulator 355
personal skills 338 process function 166
pessimism 48 process innovation 19, 397
pharmaceuticals 349 process innovation failures 41
PhD 31 process instrumentation 56
physical (or iconic) models 161 process integration and intensification 394
physical and (bio) chemical phenomena 207 process intensification 142, 333, 350
physical and chemical aspects 96 process model [M] 208
physical behavior 219 process modeling 19, 99, 149, 162, 180, 181,
physical model parameters 211 186–188, 190, 197–199, 203, 206–208,
physical singularity 390 217, 361
pilot plant 38, 56, 58, 59, 134 process optimization 199
piping and instrumentation diagram (P&ID) 65, process simulation 199
163 process simulation models 61
Planet 97 process simulator 181
planning and cost 54 process startups 70
plant commissioning 102 process step 14
plug flow 63 process stream summary (PSS) 315, 425
portfolio management 17, 33, 40 process subsystem 15
posterior analysis 397 process synthesis cases 209
practical considerations 77 process system [PS] 209
precaution principle 343 process technology 83
precommission period 70 process topology 183
preconditioning 69 processing functionality 369
prediction, validation, correction 404 process-product industries 4
prelaunch 67 procurement 64
preliminary solutions 137 product 14
presentations 315 – performance features 139
probabilistic models 395 – performance requirements 139
probability density function 373 product (quality) performance metrics 84
problem decomposition 184 product application/use processes 171
ProCAFD 349 product attributes 166
ProCAPD 182, 349 product centered sustainable process design
process alternative 53 349
process analysis cases 209 product centric process design 348
process behavior 200, 209, 211 product concept 54
process block diagram 161, 188, 202–204, 318, product design 317, 320
319, 321, 322, 362, 367, 374 product formation 170
process block scheme (PBS) 163, 315, 417 product functionality 50
process concept 54 product functions 98, 140
process concept design 54 product launch planning 67
process control 62, 216 product market introduction 67
process costs 153 product packaging 68
process design 319, 322 product packaging method 56
process engineering 83 product performance 349
Process Flow Diagram (PFD) 87 product performance in use 99
Index | 437
product pricing 64 reliability 27, 87, 103, 218, 394, 395, 402, 403
product qualities 349 reliability engineering 395
product quality 99 Reliability, Availability, Maintainability (RAM)
product storage and handling 387 87
production 67 renewable energy sources 111
production costs 237 report 83
profitability 231 reporting 158, 311, 314
Project Brief 84 Research School Process Technology 348
project management 17, 42 residence time distribution 59, 63
proof of principle 50 resource exchange networks 86
property function 166 response surface 227, 403
Prosperity 97 response surface methods 406
prototype 54 retrofit 220
prototyping 38 return on investment (ROI) 236
public/private partnerships 329 rheological properties 182
pure component properties 313, 413, 414, 416, risk adjusted value for innovation management
424 38
risk management 259, 260
Q risk of failure 76
QS World University Ranking 332, 334 risks 35
quality checks 316 roadmapping 37
Quality Function Deployment (QFD) 53, 84, 140, robustness 155
166, 167 Royal Dutch Society of Engineers 335
quantitative data 77 runaway reaction 146
R S
radically new process 27 safety, health, and environment 34
radically novel process 66 sales volume 380
ranking 34, 150, 152 scaled sensitivities 211
rapid analysis 148 scale-up by scale-down 170
rapid LCA 300 scale-up financial risk 36
rapid LCA method 149 scale-up information 58
rate expressions 223 scale-up methods 57
rate-limiting mechanism 171 scale-up process challenges 61
raw material streams 213 scale-up strategy 56
REACH 262 scaling 223
reaction stoichiometry 362 scaling factor 383
reactive distillation 27, 144 scaling of the output variables 383
reactor-separator-recycle 203, 367, 380 scenario 155, 156, 198–201, 203, 204, 207, 209,
realism 48 210, 212, 214–216, 220, 222, 272, 292,
rearranged connections 374 294, 295, 344, 371, 372, 377, 378, 386,
recognize 68 388, 392, 412
recycle structure 55 scenario specifications 372
reference case 127, 157 schematic models 161
refined rate law 390 schematics 313
reframing 76 scientific research thinking 78
refrigerants 182 scope 82
regeneration 394 scoring 151
regime analysis 171 search space 220, 363
438 | Index
see bridges where others saw holes 28 – feasibility 131

select 83 – ideation 129
selectivity factor 380 stage gate 158
sense aspect 96 Stage GateTM 79
sensitivity analysis 197, 199, 205, 216, 309, stage-gate approach 17
373, 379, 418, 427 stage-gate approach to industrial innovation 44
sensors 207, 395 stages and gates approach 43
separation factor 380 stakeholders 311, 347
separation synthesis 145 starch 179
serendipity 28 starting values 223
set of inequalities 222 startup 69, 102, 212, 394
Shanghai Ranking 332 startup and operation manuals 76
shape of a product 96 start-up entrepreneurs 3
shaping aspect 96 startup innovators 31
SHEETS 35, 82, 97, 125 startup manual 70
SHEETS criteria 53 startup time 58
SHEETS evaluation 418, 427 startup time correlation 60, 71
shutdown 212, 394 state 198
statistical properties 223
signals 207
statistical sampling 373
silicas 179
steering group 346
simple process models 219
sterilization 394
simulation 197
stiffness 173
simultaneous 88
stoichiometric coefficients 363
singularity 214
stoichiometric space 202, 362
slow convergence 214
stoichiometry 163, 189, 202, 203, 317–322,
small and medium enterprises 337
346, 362, 365–367, 369, 401, 425
Small Medium Enterprise (SME) 31
storyline 315
social (market) acceptance 35
strategic fit 37
social acceptance 96
strategy 50, 51
social innovation 12, 29
stream connectivity 375
society 51, 77 stream connectivity equations 376
solids processing step 61 stream specification 313, 412
solution space 364 streams 207
solvents 182 strictly sequential 88
solvents, refrigerants, lubricants 349 structure model 203, 204, 220, 369, 374, 376,
spatial aspect 96 388, 389
special design approaches 107 structured products 5, 138
species space 366 student engineers 3
specifications 126 successful product launch 58
– process 127 sufficient material for testing 58
– product 126 sugars 179
specified targets for some key physical outputs super dry reforming 365
of the process 201 superstructure flow sheets 388
square linear problem 372 SuperTargetTM Software 341
stability 221 supply chain 11, 68
stage supply chain imbedding 83
– concept 130 surface area 170
– development 134 suspensions 179
Index | 439
Sustainable Development Goals 307 triple P (3P) 97

Swanson’s law 22 TU Delft 331
SWOT 88
syngas 202 U
synthesis heuristics 82 uncertainty 18, 45, 87, 155, 209, 211, 216, 223,
synthesize 82 304, 367
system 198 uncertainty ranges 155
system level 124 uniform nomenclature 44
system performance 219 Unit Block Diagram 86
system splits 168 unit operation 15, 143, 207
systemic time constants 173 United Nations Global Grand Challenges 3
universities 46
T utility streams 213
tacit knowledge 4, 99
target setting 314 V
targeting 36, 81, 86, 205, 206, 381 validate market expectations 67
Task Block Diagram 86 validate the product 58
task integration 53 validating designs 94
tasks 207 validation 61
team members 311 validation phase 405
technical difficulty 68 valorization center 337
technical feasibility 35 valuation 151
technical metrics 210, 216 value 50
technical process performance 212 value chains 23
technical requirements 52, 53 variables and equations 223
technological or commercial developments 379 verbal models 162
technology development 397 verbal, schematic, and physical models 160
technology providers 32, 46 very costly 64
Technology Readiness Level 31 voice of customer 166
textile composition and structure 169 volume fractions 169
theoretical methods and models 77
thermodynamic constitutive equations 223 W
thermodynamic equilibrium 391 WACC 234
thermodynamic phases 175 warehouse 387
Thomson Reuters ASSET4 24 water-gas-shift reaction 365
time constants 170 wear of equipment 392
time value of money 210 weight factors 151
tolerable composition levels 55 weighted least squares approach 383
tolling manufacturers 33 weighting matrix 382
top innovators 20 well-documented 312
total least squares 383, 384, 406 working capital 232
total least squares estimation 384 working environment health risk assessment
trace amounts 60 120
tradeoff 220
train operating personnel 61 X
transformational 34 XTL process 226
transitions 212
transport phenomena 62 Y
transport phenomena models 168 yarn density 169

Product and Process Design Driving Innovation

Uploaded by

Copyright:

Available Formats

Product and Process Design Driving Innovation

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Product and Process Design Driving Innovation

Uploaded by

Copyright:

Available Formats

Jan Harmsen, André B. de Haan, Pieter L. J.

Chemical Process Synthesis.

Process Synthesis and Process Intensiﬁcation.

prof. dr. ir. André B. de Haan

ir. Pieter L. J. Swinkels

Library of Congress Control Number: 2018934236

Bibliographic information published by the Deutsche Nationalbibliothek

© 2018 Walter de Gruyter GmbH, Berlin/Boston

Authors’ biographies | XIX

Part A: Innovation and industry

1 Goal, scope, and structure | 3

2 Systems relevant to design for innovation | 7

3.2.2 Business motives for innovation | 20

3.6.2 Project management by stage-gate system | 43

3.10.3.1 Miniplant purpose and design | 59

4 Designing for innovation | 75

Part B: Design generation

5 Special design approaches | 107

5.9 Design for human factors | 119

6 Scoping the design | 123

7 Executing designs | 136

7.3.2 Qualitative analysis of interim solutions | 148

8 Product modeling and optimization | 160

8.5 Relations between product and process modeling | 180

Part C: Design optimization

9 Process modeling and optimization | 197

9.3.2 Cases for development stage modeling | 217

10 Evaluating economic performance | 231

11 Evaluating for safety and health | 259

11.10 Codes, standards and designing ALARP | 288

12 Evaluating for environmental, social and sustainable development

A3 Appendix to Chapter 3 | 352

A4 Appendix to Chapter 4 | 355

A13 Appendix to Chapter 13: Communicating | 411

ir. Jan Harmsen

Jan Harmsen is currently a consultant for

prof. dr. ir. André B. de Haan

Delft University of Technology (TU Delft)

Prof. André de Haan currently combines

ir. Pieter L. J. Swinkels

Delft University of Technology (TU Delft)

ir. Pieter Swinkels is director of the post-

prof. em. ir. Johan Grievink

Delft University of Technology (TU Delft)

Johan Grievink is a chemical engineering

ir. Ido E. Pat-El, MSc.

ir. Ido Pat-El graduated from the Delft

dr. ir. Jos L. B. van Reisen, PDEng

Jos van Reisen holds the position of se-

– To help companies organize and communicate an integrated approach to R&D,

1.2.1 Products classiﬁcation and description

1.2.2 Description of process industries

and gas (petroleum), chemicals, plastics, rubber, pharmaceuticals, food, beverages,

1.3 Book structure

References and further reading

System Level Size (m) Description; example

Earth 40 × 106 World system of nature and society

2.2 Earth system level

biomass Fauna Society

Soil and Water