P R E S S U R E D R O P C O N S ID E R A T IO N IN P R O C E S S IN T E G R A T IO N

A thesis submitted to the

University o f Manchester

fo r the degree o f

D O C T O R O F P H IL O S O P H Y

in the Faculty o f Technology

by

M O H A M M A D HASSAN PAN JE H SH AH I

under the supervision o f

Dr. Graham T. Pol ley

Department o f Chemical Engineering,

University o f Manchester Institute o f Science and Technology,
Sackville Street, P.O. Box 88,
Manchester M 60 1QD.

February 1992
 D E C L A R A T IO N

N o portion of the work referred to in this thesis has been submitted in support o f

an application for another degree or qualification o f this or any other university or

other institution of learning.

/ / V ^ r * y ê -C* S C—s

Mohammad Hassan Panjeh Shahi

 ACKNO W LED G EM ENTS

1 would like to express my gratitude to all the people inside and outside U M IS T

who helped and encouraged me during the course o f this research.

In particular. I am grateful to my supervisor. Dr. Graham Polley, fo r his invaluable

guidance and help. His striking personality together with his continued confidence in

me always gave me motivation to undertake this research with enthusiasm.

I would like to acknowledge the financial supports from the Islamic Republic o f

Iran and Centre for Process Integration at U M IS T .

Within the group, I wish to thank all my colleagues and friends, past and present,

for their friendship and providing such a stimulating environment. M any thanks are

due to the secretaries. Miss Eileen Boocock and Mrs. O live Ireland, for their

generous assistance during my work in CPI. I would also like to thank Mr. Simon

Perry and M r. Steve D oyle for their help in using the computer facilities.

I am deeply indebted to my parents, brothers and sisters for their love and

encouragements throughout my life and for tolerating the fact that w e have been far

away from each other during these years. Last but not least, the warmest gratitude is

extended to my w ife and two darling sons for their love, understanding and patience

during the period o f time I was studying. Without their dedication this work could

have never reached completion.

 ABSTRACT

Recent years have witnessed a rapid development in the understanding o f heat

exchanger network synthesis. Advances in ’Pinch Tech n ology’ have produced
targeting and design procedures fo r both grass-roots design and retrofit o f heat
exchanger networks. However, all o f the available techniques make use o f assumed
heat transfer coefficient and do not consider stream pressure drop at all. Th is strategy
can have unfortunate consequences such as:

inconsistency between network synthesis and detailed design;

incorrect capital-energy trade-off and network optimisation; and
misleading targeting results during retrofit projects.

This thesis is a response to the above situation. T h e work starts with the
recognition o f the relationship between pressure drop, heat transfer coefficient and
area; and then attempts to introduce a consideration o f stream allowable pressure drop
into both the targeting and synthesis stages o f heat recovery system design.

This research has resulted in:

1. a new pressure drop relationship for the shell-side performance which is based
on Bell-Delaware model;

2. a rapid and accurate heat exchanger design algorithm that makes full use of
specified pressure drop on both sides o f the exchanger;

3. a new algorithm for estimation o f the network minimum area which is based
on stream allowable pressure drop and not heat transfer coefficient;

4. a grassroots design procedure in which the predictions o f targeting and the

results o f synthesis are in a very good agreement w ith final detailed network
design;

5. a retrofit targeting procedure which takes account o f a plant’s existing flow

system and appreciates the dependency o f the stream coefficients on project
scope;

6. new design tools which enable the designer to analyze and im prove the
efficien cy o f the exchangers with respect to pressure drop utilisation;

7. a retrofit design procedure which ensures that the final results are fully
consistent with the targets.
 TAB LE O F CONTENTS

P A R T 1 - PR E S S U R E D R O P A N D N E T W O R K IN G

CH APTER 1 : IM P O R T A N C E O F C O N S ID E R IN G P R E S S U R E D R O P IN
N E T W O R K T A R G E T IN G A N D D E S IG N

1.1 Introduction 3
1.2 Hazards o f Delaying Consideration o f Stream Pressure
Drop during the Design o f N ew Plants 4
1.3 Hazards o f Delaying Consideration o f Stream Pressure
Drop during Retrofit Studies 16
1.4 N ew Procedures needed 19
1.5 Summary 20
Nomenclature 21
References 22

CH APTER 2 : P R E S S U R E D R O P R E L A T I O N S H I P F O R S H E L L -A N D -
T U B E H E A T E X C H A N G E R S (N E W M O D E L )

2.1 Introduction 26
2.2 Shell-side Flow Model 28
2.3 Pressure Drop Relationship 40
2.4 Summary 53
Nomenclature 54
References 57

CH APTER 3 : T H E R M A L D E S IG N O F S H E L L -A N D -T U B E HEAT
E X C H A N G E R S (N E W A L G O R I T H M )

3.1 Introduction 61
3.2 Recommended Design Algorithm 62
3.3 Example Problem and Comparison 73
3.4 Summary 77
Nomenclature 78
References 79
 P A R T 2 - G R A S S -R O O T S D E S IG N

CH APTER 4 : M IN IM U M A R E A O F H E A T E X C H A N G E R N E T W O R K S
(N E W A L G O R IT H M )

4.1 Introduction 82
4.2 Current Area Algorithm 83
4.3 N ew Area Algorithm 88
4.4 Summary 96
Nomenclature 97
References 98

CH APTER 5 : G R A S S -R O O T S D E S IG N O F H E A T EXCHANGER
N E T W O R K S (N E W P R O C E D U R E )

5.1 Introduction 102

5.2 Supertargeting 103
5.3 Network Synthesis 108
5.4 Case Study H2
5.5 Summary 125
Nomenclature 126
References 127

CHAPTER 6 : IN T E R F A C E OF NETW ORK S Y N T H E S IS AND

D E T A IL E D H E A T E X C H A N G E R D E S IG N

6.1 Introduction 130

6.2 Case Study (Aromatics Plant) 131
6.3 N ew Procedure andOptimalNetwork Synthesis 150
6.4 Summary 158
References 159
 P A R T 3 - R E T R O F I T D E S IG N

CHAPTER 7 : R E T R O F IT O F H E A T E X C H A N G E R NETW ORKS

(N E W T A R G E T I N G P R O C E D U R E )

7.1 Introduction 162

7.2 Retrofit Targeting Procedure to Date 163
7.3 Problems with the Existing Targeting Procedures 171
7.4 N e w Retrofit Targeting Procedure 173
7.5 Summary 180
Nomenclature 181
References 182

CH APTER 8 : R E T R O F IT O F H E A T EXCHANGER NETW ORKS

(D E S IG N T O O L S )

8.1 Introduction 185

8.2 Cross-Pinch Exchanger A n alysis 187
8.3 Driving Force Plot 189
8.4 Remaining Problem A nalysis 191
8.5 Exchanger Shifting 195
8.6 Ideal Pressure Drop 199
8.7 Summary 201
Nomenclature 202
References 204

CHAPTER 9 : R E T R O F IT O F H E A T E X C H A N G E R NETW ORKS

(N E W D E S IG N P R O C E D U R E )

9.1 Introduction 207

9.2 A nalyzing Existing Exchangers 208
9.3 Improving the Performance o f Inefficient Exchangers 210
9.4 Placing N ew Exchangers 213
9.5 Evolution 215
9.6 Summary 216
Nomenclature 217
References 218
 P A R T 3 - C O N T IN U E D

C H A P T E R 10 : R E T R O F IT O F HEAT EXCHANGER NETW ORKS

(C A S E S T U D Y )

10.1 Introduction 222

10.2 The Problem 223
10.3 Preliminary Calculation 227
10.4 Targeting 231
10.5 Design 235
10.6 Summary 264
Nomenclature 265

C H A P T E R II : C O N C L U S IO N S A N D F U T U R E W O R K

11.1 Conclusions 269

11.2 Future Work 277

A P P E N D IX A : P U B L IC A T IO N S 278
 F IG U R E S . T A B L E S A N D E Q U A T IO N S

The Figures presented in this thesis have been numbered so that they can easily

be referred to when used elsewhere in the thesis. The first digit to the Chapter

number in which the Figure first appears, follow ed by the Figure number.

The same numbering system applies for the Tables and Equations.

e.g. Figure 3.2 is the second figure in Chapter 3

Table 1.3 is the third table in Chapter 1

Equation 2.1 is the first equation in Chapter 2

PRESSURE DROP

AND NETWORKING
C hapter I Im portance o f Considering Pressure D ro p In N etw o rk .... Page 1

CHAPTER 1

IMPORTANCE O F

CONSIDERING PRESSURE D R O P

IN

NETWORK TARGETING AN D DESIGN

Chapter 1 Im portance o f Considering Pressure D ro p In Network .... P a ge 2

Contents

1.1 Introduction

1.2 Hazards o f Delaying Consideration o f Stream Pressure Drop during the Design

o f N ew Plants

1.2.1 Inconsistency between network synthesis and detailed design

1.2.2 Misleading capital-energy trade-off and optimisation

1.3 Hazards o f Delaying Consideration o f Stream Pressure Drop during Retrofit

Studies

1.4 N ew Procedures needed

1.5 Summary

Nomenclature

References
Chapter I Im portance o f Considering Pressure D rop In N etw ork .... Pa ge 3

1.1 Introduction

The design o f heat recovery networks is generally concluded in a number o f stages.

First, there is a targeting stage where the economics o f heat recovery are evaluated in

order to set the recovery level. This involves the trading-off o f energy and network

capital costs. The second stage involves network synthesis. Here the topography o f the

heat recovery system required to realise the targeted energy recovery is determined.

Finally the detailed design stage is entered. H ere the detailed design o f the exchangers

forming the recovery system is undertaken.

T o date process stream pressure drop has o n ly been considered at the detailed design

stage. It is now appreciated that the delay in considering pressure drop can lead to

serious design mistakes. For instance, it can result in networks which contain tw ice as

much area as the targeting and synthesis calculations would suggest. It can result in

network topographies which are non-optimal. It can result in poor retrofit projects.

In this chapter we demonstrate the hazards o f leaving a consideration o f pressure drop

to the later stages o f the project.

C hapter 1 Importance o f Considering Pressure D rop In Network .... Pa ge 4

1.2 Hazards o f Delaying Consideration of Stream Pressure P r o p during the

Design of New Plants

A ll o f the techniques currently used for network targeting and synthesis make use o f

assumed film heat transfer coefficients. In contrast, detailed heat exchanger design is

generally performed based on stream allowable pressure drops.

T h e detailed design strategy proceeds through the testing o f a range o f potential

exchanger geometries in order to find those which satisfy three major design objectives;

a) transfer o f required heat duty,

b) tube-side pressure drop below a maximum allowed value,

c ) shell-side pressure drop below a maximum allowed value.

In this strategy film heat transfer coefficients are incidental considerations.

Consequently, there is no guarantee that the values assumed during network synthesis

are those actually achieved in detailed equipment design. This difference between

assumed and achieved values for heat transfer coefficients causes tw o major problems

in new designs:

- inconsistency between network synthesis and detailed design, w ith respect to

exchanger surface area.

- incorrect capital-energy trade-off and network optimisation.

In next tw o sub-sections, w e are going to explain more about these problems through

an example problem.
C hapter I Im portance o f Considering Pressure D rop In Network .... Page 5

1.2.1 Inconsistency between network synthesis and detailed design

Sagli et al [1990] recently presented a detailed study o f the possible solutions to a

simple four stream problem (defined in Table 1.1).

Stream T . CC) T, CC) mCp (kW C C ) h (W/m2 ' O

HI 150 60 20 100

H2 90 60 80 100

Cl 20 125 25 100

C2 25 100 30 100

ST 180 180 100

cw 10 15 100

A T min=20 Q t.=400 kW C O S T = 8600 + 670 A ° "

Q,,= 1075 kW

Table 1.1: Four stream problem data.

The energy recovery level was fixed in original example, in order to consider only

capital cost optimisation. Even for this simple problem, a significant number of

solutions were found (Figure 1.1)

C ha p ter 1 Im portance o f Considering Pressure D ro p In Network .... P a ge 6

Am ongst these solutions, the best solution overall was found to be network 6.1,

having a total area o f 3045.4 m2 and a capital cost o f $716k. Alternatively, the best 5

units network was spotted to be network 5.1, having a total area o f 3146.6 m2 and a

capital cost o f $717.3k. The tw o networks are illustrated in figures 1.2 and 1.3.
C hapter 1 Im portance o f Considering Pressure D ro p In N etw ork .... P a ge 7

In developing their solutions, Sagli et al assumed that each stream had the same heat

transfer coefficient (100 W/m2 *C), givin g an overall heat transfer coefficient o f 50

W/m2 *C. W hile, the true values for the coefficients are only known after detailed

design.
C hapter 1 Im portance o f Considering Pressure D ro p In Network .... Page 8

Let us proceed to detailed exchanger design. In order to do so, w e need information

on fluid physical properties, stream fouling resistances and exchanger allowable

pressure drops. These are not given in the original problem. W e assume the values

presented in Table 1.2.

Stream Q Cp P k R, AP

kg/m3 J /kg'C cPs W/m *C m2 'C/W kPa

HI 800 2600 0.5 0.12 0.00018 20

H2 800 2600 0.5 0.12 0.00018 30

Cl 800 2600 0.5 0.12 0.00018 10

C2 800 2600 0.5 0.12 0.00018 60

Table 1.2 : Data for detailed design (Four stream problem).

The physical properties are those o f a typical organic liquid. The fouling factors are

typical values recommended by T E M A [1978]. The pressure drops, although quite low,

are in fact close to being optimal for the network designs shown.

The individual exchanger designs were conducted using the Bell-Delaware method,

and in such a w ay that all available pressure drop is used (the design algorithm w ill be

described in Chapter 3). W here more than one exchanger is present on a stream, the

overall stream pressure drop is taken to be distributed linearly with respect to predicted

area (it w ill be demonstrated later, in Chapter 2, that actual pressure drop is directly

proportional to installed area). The results are presented in Tables 1.3 and 1.4.
C ha pter I Importance o f Considering Pressure D r o p In Network .... Page 9

Exchanger El E3 E4 C HI H2

Expected Size(m 2) 1175.6 225.6 1209.1 160.0 124.5 150.9

Actual Size (m2) 184.0 27.7 118.3 12.0 10.6 12.5

No. o f Shells 3 1 2 i co 1 1

Shell Diam. (mm) 479.2 676.7 493.6 374.6

No. o f Tubes 206 412 201 124

Tube Passes 2 4 4 1

Tube O.D. (mm) 19.1 19.1 19.1 19.1

Tube I.D. (mm) 15.4 15.4 15.4 15.4

Tube Pitch (mm) 25.4 25.4 25.4 25.4

Tube Length (mm) 4956 1123 4916 1613

Baffle Cut (% ) 23.7 25.8 42.5 39.6

No. o f Baffles 28 3 12 5

h „ (W/m2 -C) 618.9 618.3 1109.1 1679.6

h„ (W/m2 "C) 937.7 1277.2 1073.8 1104.4

U (W/m2 'C ) 372.8 416.6 545.6 667.0

AP, (k Pa) 8.15 1.22 54.12 1.00

AP, (k Pa) 20.00 5.40 22.41 2.19

Table 1.3 : Results o f the Detailed Exchanger Design for Network 6.1,
(Four stream problem).
Chapter 1 Im porta n ce o f Considering Pressure D rop In Network .... P a ge 10

| Exchanger E2 E3 E4 C HI

Expected S ize(m 2) 1099.5 1342.2 256.5 160.0 288.8

Actual Size (m 2) 122.2 180.3 23.3 13.1 26.1

No. o f Shells 3 2 1 10) 1

Shell Diam. (m m) 360.1 670.5 612.5 386.2

No. o f Tubes 104 403 315 133

Tube Passes 2 4 6 1

Tube O.D. (m m ) 19.1 19.1 19.1 19.1

Tube I.D. (m m) 15.4 15.4 15.4 15.4

Tube Pitch (m m ) 25.4 25.4 25.4 25.4

Tube Length (mm) 6492 3735 1233 1639

Baffle Cut (% ) 35.3 27.6 33.3 42.5

No. o f Baffles 29 12 3 5

h*, (W/m2 -C) 1085.7 627.7 1079.7 1626.6

h„ (W/m2 -C) 919.2 1233.3 1150.5 980.3

U (W/m2 -C ) 497.8 416.0 557.0 611.7

AP, (k Pa) 50.60 8.46 9.40 0.90

AP. (k Pa) 20.00 25.22 3.22 1.56

Table 1.4 : Results o f the Detailed Exchanger Design for Network 5.1,
(Four stream problem).
Chapter 1 Im portance o f Considering Pressure D rop In Network .... Page 11

A s can be seen, the detailed exchanger design (given the assumed stream properties

and pressure drops) results in exchanger sizes that are very different to those predicted

during the network synthesis. The total area fo r the networks 6.1 and 5.1 (after detailed

design) are 365.1 m2 and 364.9 m2. respectively. The overall heat transfer coefficients

are by n o means uniform, and the heat transfer coefficients assumed during the network

synthesis were far too conservative.

T h e reason for such large discrepancies is that the area estimated during targeting and

network synthesis is based on assumed ’h ’ values, while the detailed design is

conducted, using allowable pressure drops. T h e assumed ’h* values and the allowable

pressure drops are unlikely to be consistent, because, their values originate from

different information sources.

T h e information flo w between the tw o engineering groups is also interesting. The

only information going forward is network structure and individual exchanger loads and

temperatures. Therefore, the network area determined during the synthesis appears

irrelevant. The actual area o f the network is determined by the detailed design.

Figure 1.4, schematically, shows the current practice in design o f Heat Recovery

N etworks and highlights its drawback.

Chapter 1 Importance o f Considering Pressure D r o p In Network .... Page 12

Figure 1.4 : Current design practice in Heat R ecovery Network.

»•2.2__ Misleading capital-enerpy trade-off and optimisation

As described above, the first stage in the design o f heat exchanger networks, using

Pinch Technology, is targeting. In this stage, a capital-energy trade-off is carried out to

set the targets for area and energy, and to find the minimum total cost. These targets,

subsequently, initialize the second stage which is the network synthesis. During these

two stages, the values fo r the heat transfer coefficients are assumed based on experience

or literature.
Chapter 1 Importance o f Considering Pressure D ro p In Network .... P a ge 13

N ow, let us see how sensitive the optimisation is to heat transfer coefficient

assumptions. In table 1.5. we g iv e the capital costs o f networks 6.1 and 5.1 fo r a range

o f overall heat transfer coefficients, assuming that each exchanger has the same value.

O.H.T.C. C A P IT A L C O S T ($)

(W/m2 -C) N E T W O R K 6.1 N E T W O R K 5.1

25 1232740 1241820

50 716000 717300

75 526160 524640

100 425340 422370

250 226300 220300

500 149900 142750

1000 106900 99100

Table 1.5 : Influence o f O.H.T.C. on Capital Costs.

W e see that assumptions regarding heat transfer coefficients are very critical to the

optimisation. Only with overall coefficients less than about 65 W/m2 *C is the 6.1

network cheaper than 5.1 network. Even at fixed energy consumption, network 6.1 is

no longer optimal at an overall heat transfer coefficient o f 100 W/m2 *C. This

observation is confirmed by the results o f detailed design. W e remind ourselves that

after detailed design the final capital cost o f network 6.1 w ould be $163.6k, whilst that

o f network 5.1 would be $154k (6% lower). However, having applied the current state-

of-the-art technique, network 1.6 was recognised by Sagli et al to be ’optim al’ solution.
C hapter 1 Importance o f Considering Pressure D ro p In Network .... P a g e 14

M ore importantly, as the capital cost changes the trade-off between capital and energy

changes, and w e can expect very different energy consumptions and network structure.

In other words, i f this had been a real problem, the large error between predicted

($716k) and actual ($163.6k) capital costs would have invalidated the previously

determined trade-off between energy and capital casts. Consequently, if the true

optimum network is to be identified, there would have to be an iteration (Figure 1.5)

between network synthesis and detailed exchanger design.

Figure 1.5 : True optimisation would involve impractical iteration.

In industrial practice such an iteration is rarely practical. The two engineering

activities are usually conducted by different engineers, working at different times for

different managers and sometimes even in different geographical locations. The result

is that mistakes made at the network synthesis stage can rarely be redressed.
Chapter 1 Im portance o f Considering Pressure Drop In Network .... P a ge 15

Now, the work o f Sagli et al is a demonstration o f the current state-of-the-art in

network synthesis and the solution they present is optimal for the problem they posed.

Our argument is that the problem stops too far short to be realistic and this is due to the

dependence o f the synthesis methodology on assumed heat transfer coefficients.

B y adding information on stream pressure drops and physical properties, w e have

changed the problem, but in doing so we have made it closer to industrial reality.

Subsequent examination o f the problem then reveals a further inherent weakness within

current technology: it does not recognise the g u lf that exists between engineering design

activities. Given the impracticality o f iterating between network synthesis and detailed

exchanger design, we should seek to develop a procedure which ensures that the results

o f the two activities are consistent.

C ha pter 1 Im portance o f Considering Pressure D ro p In Network .... Pa ge 16

LA__ Hazards o f Delaying C onsideration of S lrfam Pressure D rop during the

Retrofit Studies

Process integration technology is now w idely used in both energy saving retrofits

[T joe, 1986] and plant debottlenecks [Ahmad and Polley, 1990]. H ow ever, in common

with much o f the developed technology, systematic consideration is only given to the

thermal aspects o f process design. A retrofit project, analogous to a new design,

proceeds in two major stages (i.e. Targeting and Design). In Figure 1.6, a typical

targeting curve fo r the retrofit o f heat exchanger networks is shown.

Figure 1.6 : A typical targeting curve for H.E.N. retrofit.

C hapter 1 Im p orta n ce o f Considering Pressure D r o p In Network .... Pa ge 17

The profile enclosing the shaded area indicates the ’ ideal’ area-energy plot fo r a heat

exchanger network. The position denoted by (* ) represents the existing network, and the

curve originating from this position represents the retrofit targeting curve. B y applying

energy and area cost data, this curve can be translated to Investment-Saving Plot. The

project scope is then set by either payback period or investment limit. Using one o f

these criteria, the designer can make his decision regarding required additional area

(A A ) for a given amount o f energy reduction (AE).

In design stage, the well-placed exchangers are first recognized and remained

unchanged. Subsequently, the inefficient exchangers are examined and corrected.

Finally, the new exchangers are installed to complete the design. Figure 1.7 shows a

typical network before and after retrofit modification.

Figure 1.7 : A typical network before and after retrofit.

C hapter 1 Im portance o f C onsidering Pressure D rop In N etw ork .... Page 18

The n e w exchangers installed (indicated by dark circles) w ill usually result in

additional pressure drop for the streams. H ow ever, in current retrofit strategy, the effect

o f any changes in the heat exchanger network on the network’ s flo w system (pumps,

compressors, gravity) is ignored, and considered only after the new network

arrangement has been determined. Flow considerations do not form part o f targeting or

network design procedures.

In one project known to the author, the use o f an established retrofit targeting

procedure [T joe, 1986] suggested that the annual energy costs o f a plant could be

reduced b y around £250 000, through a retrofit project involving new exchangers having

an installed cost o f £500 000. The project appeared very attractive and a new network,

that confirm ed the targeted predictions, was designed. Unfortunately, when detailed

system design was undertaken, it was found that the new scheme would require the

replacement o f pumps costing around £1 000 000. The economics o f the project then

proved unviable. This discovery was made quite late and after the expenditure o f a lot

o f engineering effort.

Therefore, it is important that the effect o f any changes in the heat exchanger network

on its f lo w system are properly appreciated, and full consideration o f pressure drop is

made during network retrofit.

Chapter l Im portance o f Considering Pressure D ro p In N etw ork .... Page 19

1.4 New procedures needed

It is now clear that stream pressure drop should be considered during network

targeting and synthesis. In order to do this, w e need the follow in g:

1. Algorithms which base network area estimation on stream pressure drop

and not assumed heat transfer coefficients. These need to c o v e r both new

plant design and retrofit.

2. Design tools which ensure that use o f pressure drop is c o n tro lle d during

network synthesis.

However, w e also have a problem when it comes to detailed exchanger design.

Current algorithms do not guarantee that designs which are produced make full use o f

allowable pressure drop. E fficien t use of pressure drop throughout exchanger networks

demands that the individual exchangers fully utilise what is available [Polley et al,

1991]. Thus w e also need

3. A new heat exchanger design algorithm that allows f u l l utilisation o f

allowable pressure d rop o n both sides o f the exchanger.

C h a p te r 1 Importance o f Considering Pressure D r o p In N etw ork.... Page 20

1.5 Summary

A ll o f the techniques currently used for network targeting and synthesis make use o f

assumed film heat transfer coefficients, and leave the stream pressure drops to be

considered at the detailed design stage.

It has been demonstrated (through an example p rob lem ) that, this delay in considering

pressure drop can lead to serious design mistakes. T h e s e mistakes can be summarised

as follow s:

1. inconsistency between network synthesis and detailed design, with respect to

surface area o f the exchangers.

2. incorrect capital-energy trade-off and network optimisation.

3. misleading targeting results during retrofit projects.

It has, therefore, made clear that stream pressure d ro p should be considered during

network targeting and synthesis. In order to do this, w e need new targeting algorithms

(w h ich base network area estimation on stream pressure drop and not assumed heat

transfer coefficients) as w ell as new design tools.

C hapter 1 Importance o f Considering Pressure D ro p In Network .... Page 21

Nomenclature

A heat transfer surface based on outside tube area

Cp specific heat

h heat transfer coefficient

hio tube-side heat transfer coefficient (based on tube O.D.)

h,, shell-side heat transfer coefficient

k heat conductivity

m stream m ass flowrate

Qc cold utility consumption

Qh hot utility consumption

R, fouling resistance

T, stream supply temperature

T, stream target temperature

U overall heat transfer coefficient

AP pressure drop

APS shell-side pressure drop

AP, tube-side pressure drop

A T min minimum temperature approach on composite curves

p stream v iscosity
Chapter 1 Importance o f C onsidering Pressure D rop In N etw ork .... Page 22

References

Ahmad S., P olley G.T.,

’ Debottlenecking o f Heat Exchanger N etw ork s’ ,
J.Heat R ecovery & CHP, 10(4), pp. 369-385, 1990.

Jegede F.O.,
’ Power, Capital and Energy Cost T rad e-offs in Heat Exchanger Networks’ ,
Ph.D. Thesis, U .M .I.S.T., 1990.

Linnhoff B. et al,
’ User Guide on Process Integration fo r the Efficient Use o f Energy’ ,
IChemE, Rugby, 1982.

Linnhoff B. & Ahmad S.,

'Optimal Synthesis o f Energy Management Systems’ ,
T A S M E (J.Eng.Res.Tech.), 111(3), pp. 121-130, 1989.

P olley G.T. & Panjeh Shahi M.H.,

’ Interfacing Heat Exchanger Network Synthesis and Detailed Heat Exchanger Design’ ,
Trans. I.Chem.E. 69A, pp. 445-457, 1991.

P olley G.T. , Panjeh Shahi M.H. and J eged e F.O.,

’ Pressure Drop Considerations in the R etrofit o f Heat Exchanger Networks’ ,
Trans. I.Chem.E. 68A, pp. 211-220, 1990.

P olley G.T. , Panjeh Shahi M.H. and N u n ez M.P.,

’ Rapid Design Algorithms fo r Shell-and-tube and Compact Heat Exchangers’ ,
Trans. I.Chem.E. 69A , pp. 435-444, 1991.

Sagli B., Gundersen T. and Y e e T.,

’Topology Traps in Evolutionary Strategies for Heat Exchanger Network Synthesis’ ,
Computer Application in Chemical Engineering, ed. H.Bussemaker 8c P.Iedema,
Elsevier, Amsterdam, 1990.

TEM A,
’Standards o f Tubular Exchangers Manufacturers Association (6th edn.)’ ,
N ew York, 1978.
Chapter 1 Im portance o f Considering Pressure D rop In N e tw o rk .... P a ge 23

T joe T.N.,
’ Retrofit o f Heat Exchanger Networks’ ,
Ph.D. Thesis, U M IS T , 1986.
Chapter 2 Pressure D ro p Relationship f o r S& T heat exchangers P a ge 24

CHAPTER 2

PRESSURE DROP RELATIONSHIP

FOR

SHELL-AND-TUBE HEAT EXCHANGERS

( NEW M ODEL )
C hapter 2 Pressure D r o p Relationship f o r S& T heat exchangers Page 25

Contents

2.1 Introduction

2.2 Shell-side Flow Model

2.2.1 Survey o f shell-side flo w models based on analytical approach

2.2.2 Selected shell-side model

A . Method structure

B. Shell-side heat transfer

C. shell-side pressure drop

2.3 Pressure Drop Relationship

2.3.1 Background

2.3.2 N ew relationship for the shell-side based on the Bell-Delaware

model

2.3.3 Simplification to Bell-Delaware model

2.3.4 Extension o f the idea to state-of-the-art models

2.4 Summary

Nomenclature

References
Chapter 2 Pressure D r o p Relationship f o r S& T heat exchangers Page 26

2 .1 In tr o d u c tio n

In chapter 1, when w e discussed the importance o f pressure drop in heat

exchanger network design, three significant drawbacks w ere found in existing

procedures. These drawbacks can be listed as follows:

In new design:

1. Targeting and network synthesis are based on assumed stream film

coefficients, whereas, in detailed design stage, decisions are made based on

specified allowable pressure drops. This inconsistency results in the final

network area to be different from that predicted in targeting stage.

2. In current methods, trade-off between energy and capital costs is carried out

using assumed heat transfer coefficients. However, since the true values of

coefficients are only known after detailed design stage, the final capital cost

w ill be quite different to those predicted during targeting or network synthesis.

Therefore, this difference between predicted and actual capital costs invalidates

the previously determined trade-off.

In retrofit design:

3. Installation o f additional area due to energy saving causes more flow

resistance. Therefore, when detailed design is undertaken, it is often found that

the new scheme would require the replacement o f pumps and/or compressors.

These can sometimes cost even more than the whole retrofit project.
Chapter 2 Pressure D ro p Relationship f o r S & T heat exchangers P a ge 2 7

In addition to above drawbacks, another weakness was also mentioned regarding

current methods for heat exchanger detailed design.

In detailed design:

4. Pressure drop fo r both shell-side and tube-side are generally set as two

constraints and not part o f design objectives. A s long as the real pressure drops

are less than the allowable values, the design is acceptable. H owever, since the

smallest exchanger w ill be the one which makes fu ll use o f both pressure drops,

current algorithms cannot guarantee design optimality.

These disadvantages suggest that pressure drop should be considered, right from

the beginning, as a ) an input parameter in networking and b ) part o f design

objectives in detailed design. However, fo r doing this, w e need a simple expression

that relates the allowable pressure drop o f a stream to the surface area and the film

heat transfer coefficient o f the stream.

Such an expression has been developed for tube-side performance by Jegede

[1990]. W e do not yet have an expression o f acceptable accuracy for shell-side

performance. In this chapter, we first present a survey about different shell-side flo w

models and, consequently, an appropriate model is chosen. The selected model is,

then, used in order to develop an accurate relationship fo r the shell-side performance.

Kern model is also revisited and m odified to improve its accuracy. Finally, w e see

that it is possible to derive such a relationship for commercial heat exchanger design

programs.
C hapter 2 Pressure D ro p Relationship f o r S& T heat exchangers Pa ge 28

2.2 Shell-side Flow Model

T h e process o f heat exchanger evaluation consists o f calculations o f heat transfer

coefficient "h" and pressure drop "AP" on both shell side and tube side. H owever,

since the tube side calculations «*•« relatively simple and w ell established in the

literature, only the methods for shell-side flo w are presented here.

2.2.1 Survey o f shell-side flow models based on analytical approach

Early, so-called integral approaches, such as Kern [1950], Gilmour [1952-54] or

Donohue [1955] methods, assumed that the shell-side fluid flow s across the bundle

in pure cross-flow without leakage as in Figure 2.1.

Figure 2.1 : Idealised shell-side model

Chapter 2 Pressure D ro p Relationship f o r S & T heat exchangers Page 29

Palen and Taborek [1969] compared various shell-side design methods against a

large experimental data bank o f H.T.R.I. and concluded that the Donohue and Kern

methods produced predominant error ratios o f such large magnitude that they could

not be considered valid entries for the solution o f the shell-side problem [Saunders,

1988].

In reality, only part o f the fluid takes the desirable path through the tube nest,

whereas a potentially substantial portion flow s through the "leakage" areas (baffle-to-

shell and tube-to-baffle) and through the "bypass" area between tube bundle and the

shell w all. These clearances are inherent to the manufacturing and assembly process

o f shell-and-tube exchangers, and the flo w distribution within the exchanger must be

taken into account when considering pressure drop and heat transfer.

T h e earliest analysis o f shell-side flo w is due to Tinker [1958], w h o suggested a

schematic flo w pattern as shown in Figure 2.2.

Chapter 2 Pressure D r o p Relationship f o r S& T heat exchangers P a ge 30

The shell-side is divided into a number o f individual streams as follows:

- Stream A is the leakage stream in the orifice formed by the clearance between

the baffle tube hole and the tube wall.

- Stream B is the main effective cross-flow stream, which can be related to flow

across ideal tube banks.

- Stream C is the tube bundle bypass stream in the gap between the bundle and

the shell wall.

- Stream E is the leakage stream between the baffle edge and the shell wall.

- Stream F is the bypass stream in flo w channels due to omission o f the tubes in

tube pass partitions. (This stream was added to the original Tin ker model by Palen

and Taborek [1969]). It behaves similarly to stream C, but w ill be present only in

some tube layouts.

Chapter 2 Pressure D ro p Relationship f o r S & T heat exchangers P a ge 31

Tinker’s original analysis was quite com plex and hard to apply by hand.

Therefore, it never became w idely accepted, despite its imaginative approach. It was

not until computers became widely available that the real potential o f Tinker’ s

approach could be put to proper use.

T w o other additional areas o f work contributed much to an understanding o f shell-

side flow .

First, flo w visualization studies [Kopp & et al, 1947 and Gupta & Katz, 1957]

dramatically demonstrated the widely varying effectiveness o f the various streams.

Hence, it became obvious that effectiveness factors must be introduced to properly

adjust the heat transfer performance.

Secondly, the results o f extensive research at the University o f Delaware [B ell,

1963] supplied (a) values for the A and E stream orifice resistance coefficients

(indispensable to proper calculation o f these flo w fractions), (b) a massive data set

on small-scale baffled exchangers with reasonably systematic variation o f many

parameters; and, (c) data on ideal tube banks in laminar and transition flow.

The first direct result o f this progress was the formulation o f the Delaware method

by Bell, Bergelin, ColbUm and co-workers [1960 & 1963]. Searching for a non-

reiterative type o f method suitable for manual calculations, they could not take full

advantage o f Tinker’ s flo w model. Instead, a new type o f method emerged that may

be best termed "semi-analytical". It is based on ideal tube bank flo w and respects the
Chapter 2 Pressure D rop Relationship f o r S& T heat exchangers Page 32

effects o f leakage and bypass streams, but not the interaction between them. Other

calculational refinements include a consideration o f the effects o f baffle cut and

adverse temperature gradient build-up in laminar flow . The method has become very

popular and n ow appears in standard textbooks .

T h e next development was the rigorous application o f Tinker’ s flo w principles in

a reiterative system o f calculations. This was made possible through both availability

o f digital computers and the data o f the Delaware project. These methods are known

as "stream analysis methods", and are documented by Palen and Taborek [1969].

In addition to the Delaware data, a very extensive data set on commercial-size

exchangers w as necessary for the formulation o f these methods. Consequently, the

methods are the confidential property o f commercial companies providing heat

exchanger technology (H TR I, HTFS).

2.2.2 Selected shell-side model

The Bell-Delaware method is considered by Taborek [1983] ’ to be the best

available and the most suitable fo r general engineering applications’ . It is also named

in literature as ’ a method with a sound basis such that the results are realistic’

[Saunders, 1988] and [Coulson et al., 1989]. Consequently, the method has been

selected as a basis for the development o f the new design algorithm presented in this

thesis.
C hapter 2 Pressure D r o p Relationship f o r S& T heat exchangers Page 33

A. Method structure |Taborek, 19831

The method uses ideal tube bank ^ and fj factors and then corrects directly the

resulting hj and APj for deviations caused by the various split streams.

The jj and fj factors are defined as follows:

Ji - . ^ ----- ( P r s ) 2/3 ( <p„)~r - f (.R e ,. t u b e l a y o u t ) (2 .1 )

' CP •3 mB

f ± = 1 0 3 — ^ P± — ( 0 S) ~z - f{ R e a, tu b e la y o u t) (2 .2 )
2 ( mB) 2 N c

where

h, : heat transfer coefficien t for an ideal tube bank,

APj : pressure drop fo r an ideal tube bank,

m, : shell-side mass velocity,

q, : shell-side fluid density at average temperature,

(Cp), : shell-side fluid specific heat at average temperature,

4>s : viscosity correction factor for shell-side fluid,

Prf : shell-side Prandtl number,

Re, : shell-side Reynolds number,

Nc : total number o f tube r ow s crossed in the exchanger or any number o f tube

rows crossed.
C hapter 2 Pressure D ro p Relationship f o r S& T heat exchangers Page 34

Given overall throughput and geometry the ideal Reynolds number (Re.) can be

calculated. T h e Reynolds number in this model is based on the minimum cross-

sectional flo w area at the shell diameter and the full shell-side stream. This is

contrary to Tin k er’s method and the stream analysis-type approach, where the

magnitude o f the individual stream is determined.

Then, the Colburn j factor (j() and friction factor (fs) can either be read directly

from the graphs [Taborek, 1983] or calculated using follow in g correlations:

a and b in above correlations are:

a
1 + 0 . 1 4 ( R e , ) **

¿3
b - (2 . 6 )
1 + 0 .1 4 (R e .)

and the correlational coefficients are listed in table 2.1.

Chapter 2 Pressure D ro p Relationship f o r S & T heat exchangers Page 35

Table 2.1 : Correlational coefficients for jj and f ; .

B. Shell-side heat transfer [Taborek, 19831

The effective average shell-side heat transfer coefficient is calculated from the

ideal value as follows:

hB = c J b J a <JZ) = h i J tot (2 .7 )

where:

Jc is the correction factor for the baffle cut. This factor takes into account the heat

transfer in the window and calculates the overall average for the entire heat

exchanger. This correction factor is essentially a function o f the fraction o f the total

tubes in the heat exchanger that are in cross flow (i.e.: located between the b a ffle tips

o f adjacent baffles).
C h a p ter 2 Pressure D ro p Relationship f o r S& T heat exchangers P a ge 36

J, is the correction factor for baffle leakage effects, including both shell-to-baffle

and tube-to-baffle leakage (A and E streams). This correction factor is a function of

the ratio o f total leakage area per baffle to the cross-flow area between adjacent

baffles and o f the ratio o f the shell-to-baffle leakage area to the tube-to-baffle leakage

area.

Jb is the correction factor for the bundle bypass flo w (C and F streams). This

factor accounts for differences in construction:

For the relatively small clearance between the outermost tubes and the shell fo r fixed

tube-sheet construction, Jb = 0.9, whereas for the much larger clearances required by

pull-through floating-head construction, Jb « 0.7.

J, is the correction factor for non-equal baffle spacing in the inlet and outlet

sections. N o zzle dimensions, when baffle spacing is relatively small, require that the

nozzles be located some distance from the tube-sheets and this may lead to an

increase in the spacing o f the inlet and outlet sections. This correction factor takes

into account the effect o f that non-ideality.

Jr is the correction factor for adverse temperature gradient build-up in laminar

flo w . It is w ell known that, in laminar flo w , the heat transfer coefficient decreases

with increasing distance from the start o f heating. The necessary correction has been

w orked out mathematically for flo w in well-defined geometries such as inside round

tubes, but is also found experimentally to exist during flo w across tube banks.
Chapter 2 Pressure D r o p Relationship f o r S & T heat exchangers P a ge 37

The combined effect o f all o f these correction factors fo r a well-designed shell-

and-tube heat exchanger is typically in the order o f 0.6; that is, the effective mean

shell-side heat transfer coefficient for the exchanger is 60% o f that calculated if the

entire flo w took place across an ideal tube bank corresponding in geometry to one

cross-flow section.

C. Shell-side pressure drop (Taborek, 19831

The shell-side pressure drop is calculated as a summation o f the pressure drops

for the inlet and outlet sections, the internal cross flo w sections, and the window

sections. Again, the method uses the ideal tube bank friction factor f, as a basis, as

defined in Equation 2.2. The correction factors (analogous to those fo r heat transfer)

are defined as follows:

R, is the correction factor fo r baffle leakage effects (A and E streams). This

correction factor is different in magnitude from J, but depends on the same ratios o f

total baffle leakage area to cross-flow area.

Rt, is the correction factor fo r bypass flow (C and F streams). It is also different

in magnitude from Jb, but sim ilar in form.

R, is the correction factor fo r the entrance and exit sections having a different

baffle spacing than the internal sections.

Chapter 2 Pressure D ro p Relationship f o r S & T h eat exchangers Page 38

Next, let us define the pressure drop in an equivalent ideal tube bank for the

cross-flow section as APbi and for the baffle window section as A P wi. These two ideal

pressure drops are calculated in (kPa) as follows:

A P b i = 2 ( I O ’ 3) f il ^Nttc c (2.8)

where N lcc is the number o f effective rows crossed in one cross-flow section (that is,

between the ba ffle tips).

For turbulent flo w , Re, > 100 :

(2 .9 )

For laminar flo w , Re, < 1 0 0 :

where N lcw is the number o f effective rows crossed in one baffle window section.

For an exchanger that has both bypass and leakage streams, the three pressure

drop components are calculated as follows:

1. The pressure drop in the interior cross-flow sections is affected by both bypass

and leakage. Therefore, the combined pressure drop o f all the interior cross-flow

sections is

A P c » A P bl ( Nh -1 ) R j Rb (2 . 11)

where N b is number o f baffles.

 C ha p ter 2 Pressure D ro p Relationship f o r S & T heat exchangers P a g e 39

2. The pressure drop in the windows is affected by leakage but not by bypass.

Therefore, the combined pressure drop o f all the window sections is given by

A P w - A P wiN h R 2 (2 .1 2 )

3. The pressure drop in the entrance and exit sections is affected by bypass but

not by leakage. Additionally, there is an effect due to possibly larger baffle spacing.

Therefore, the combined pressure drop fo r the entrance and exit is

AP. * ( 2 - 13)

w here N lcc and N lcw are the number o f tube rows crossed between baffle tips o f one

ba ffle compartment and in one baffle w indow , respectively.

The total pressure drop over the exchanger shell-side is then

A Pa = A P c + A P „ + A P e (2 .1 4 )

This does not include the pressure drop in the nozzles, which is usually included

in the specification.

Although each o f the correction factors can vary over quite wide ranges,

depending on the configuration o f the heat exchanger, the total shell-side pressure

drop o f a typical shell-and-tube exchanger is o f the order o f 20-30% o f the pressure

drop that would be calculated for flo w through the corresponding heat exchanger

without baffle leakage and without tube bundle bypass effects. (It is here that most

o f the earlier methods for heat exchanger design, such as Kern’ s or Donohue’ s, failed
C ha p ter 2 P ressure D ro p Relationship f o r S & T heat exchangers Page 40

most seriously, and therefore it was not uncommon for some o f them to overestimate

the shell-side pressure drop by a factor o f 2 or even 10).

2.3 Pressure Drop Relationship

2.3.1 Background

A first step in the e ffective consideration o f allowable pressure drop is to establish

a quantitative relationship between the heat transfer coefficient, the surface area o f

the exchanger and the actual pressure drop o f the stream.

Jegede [1990] has shown that, for any heat exchanger system, a general

relationship, o f the form bel ow, can be obtained.

A P - /J (2.15)

where

KP : a constant and function o f fluid properties and the heat exchanger system,

V* : volumetric throughput,

n : Reynolds exponent in friction factor equation and

m : Reynolds exponent in heat transfer equation.

T h e above equation provided a general guideline fo r establishing tw o pressure

drop relationships for shell-and-tube heat exchangers. These relationships are:

 C hapter 2 Pressure D r o p Relationship f o r S & T heat exc/uingers Page 41

For the tube side:

APt - kc A h i 5 (2 . 16)

where

A : heat exchanger area,

AP, : tube-side pressure drop,

h, : tube-side heat transfer coefficient,

k, : tube-side constant, that is a function o f

physical properties o f the tube-side stream,

volumetric flowrate o f the tube-side stream,

tube inside diameter.

and for the shell side:

AP. (2 .1 7 )

where

A : heat exchanger area,

AP, : shell-side pressure drop,

h, : shell-side heat transfer coefficient.

k, shell-side constant, that is a function o f

physical properties o f the shell-side stream,

volumetric flowrate o f the shell-side stream,

tube bundle equivalent diameter.

Chapter 2 Pressure D ro p Relationship f o r S & T heat exchangers Pa ge 42

In deriving the shell-side relationship, Jegede [1990] assumed a fixed baffle cut

(%25), which was the condition at which K ern ’ s equations applied.

Equations 2.16 and 2.17 show that, for both shell side and tube side, the pressure

drop o f a stream increases with the area o f the exchanger and the velocity, (in form

o f heat transfer coefficient), o f the stream.

It is also seen that, design variables such as tube length, tube count, shell diameter

and number o f tube passes have been eliminated from the final expressions.

The above relationships are significant, as they could be applied in tw o important

areas (detailed heat exchanger design and pow er cost targeting fo r the heat exchanger

networks). However, the shell-side relationship suffers from a major drawback, that

is its dependency on Kern's equations for heat transfer coefficient and pressure drop.

These equations are not o f sufficient accuracy for modem design needs. T o be of

practical use, the approach must, at the very least, be compatible with on e o f the

analytical approaches, such as Bell-Delaware method for shell-side performance.

In next sub-section we aim to derive a sim ple relationship between heat transfer

coefficient, pressure drop and exchanger area, using Bell-Delaware method.

Chapter 2 Pressure D ro p Relationship f o r S& T heat exchangers Page 43

2 .3 .2 N e w r e la t io n s h ip f o r t h e s h e ll- s id e b a s e d o n t h e B e l l - D e l a w a r e m o d e l

Here, w e show how the various equations constituting the Bell-Delaware method

can be combined and rearranged to give a relationship between film heat transfer

coefficient, shell-side pressure drop and exchanger surface area. T o do this, heat

transfer coefficient and pressure drop should first be expressed as functions o f the

velocity, and then related to each other through elimination o f velo city between them.

A. H e a t T r a n s f e r C o e ffic ie n t a s a fu n c tio n o f V e lo c it y

W e start from the very basic equation, that is heat transfer coefficient in an ideal

tube bank. Substituting for m5 in Equation 2.1, using m, = p, U,, and rearranging

gives

hl - ( Cp ) a ( P r m) - 2' 3 {<f>3) z ] Ua (2 .1 8 )

where U, is the shell-side fluid velocity, and j* assumed to be an independent design

parameter.

A s it is expected from the Bell-Delaware model, this ideal coefficient should now

be corrected b y various heat transfer correction factors, using Equation 2.7. The result

is

hm = k x Ua (2 .1 9 )

where

< Cp)apa (Pra) ~2/3( 0 . ) ' Jtot (2 .2 0 )

Chapter 2 P ress u re D ro p Relationship f o r S& T heat exchangers Pa ge 44

B. Pressure D rop as a function of Area and Velocity

This time, w e start w ith pressure drop in an equivalent ideai tube bank fo r the

cross-flow section (APbl) and fo r the baffle window section (APwi). Having substituted

m, with Qt U , in Equations 2.8 and 2.9, and assumed that fj is an independent design

parameter, these tw o ideal pressure drops can be expressed as functions o f the

velocity.

A PW - k 2 ( U m) 2 (2.21)

where

Jc2 = 2 (1 0 * 3) N ccc p , (* ,)-* ( 2 . 22)

and

<2 * 23>

where

^3 = (1 + 0 . 3 ( 1 0 _3) p s (2.24 )

N ow , the three pressure drop components can be related to velocity as follows:

1. Substituting fo r A P bi in Equation 2.11, using Equation 2.21 gives cross flo w

pressure drop as

A P C - k 4 (N b - l ) ( Um) 2 (2.25)

where
Chapter 2 Pressure D rop Relationship f o r S& T heat exchangers Page 45

k t = k 2 J?j Rb (2.26)

2. Substituting for A P wi in Equation 2.12, using Equation 2.23 gives pressure drop

in the baffle windows as

A P w = k 5 Nb (U m) 2 (2.27)

where

k s = k3 R2 (2.28)

3. Substituting for A P bi in Equation 2.13, using Equation 2.21 gives pressure drop

in the end zones as

A P 9 = k 6 ( Um) 2 (2.29)

where

k‘ = k* (1+f e ) Rb R° <2'301

The total shell-side pressure drop is given by Equation 2.14. Substituting for APC,

A Pe and A PWin this equation, using Equations 2.25, 2.27 and 2.29, respectively, and

rearranging gives

A P , = [ ( N b* 1 ) ( k 4+ks) +k6- k 5-2 k < ] ( U . ) 2 (2 .3 1 )

The heat transfer surface area is calculated by

Chapter 2 Pressure D ro p Relationship f o r S& T heat exchangers Page 46

A = N ct n D t L t (2 .3 2 )

where

N„ : total number o f tubes,

D, : tube outside diameter,

L, : tube length.

The tube length is

Lc - U V D Ltc (2 .3 3 )

where

Nb : number o f baffles

Lb«. : central baffle spacing

Substituting for L, in Equation 2.32, using Equation 2.33, and rearranging gives

Substituting for N b+1 in Equation 2.31, using Equation 2.34, gives

A P B = ( k 7 A + k B) ( Ua) 2 (2 .3 5 )

where

(2 .3 6 )

and

k B » k6- k 5- 2 k 4 (2 .3 7 )
C ha p ter 2 Pressure D rop Relationship f o r S & T heat exchangers P a ge 47

C. Relationship between AP, h and A.

Heat transfer coefficient and pressure drop are now both functions o f the velocity.

Therefore, they can be related to each other, if we sim ply eliminate the velocity

between Equations 2.19 and 2.35. T h e result is

AP. (2 .3 8 )

or

A P . - [k m2A + k m2] h i (2 -3 9 )

where

(2 .4 0 )

an d

(2 .4 1 )

Equation 2.39, which is termed Pressure Drop Relationship for the shell side, is

n ow based on full Bell-Delaware method. Both o f the constants, in this equation, are

com plex functions involving shell-side flowrate and physical properties, ideal heat

transfer j-factor, ideal friction factor and shell geometry.

Chapter 2 P ressure D ro p Relationship f o r S& T heat exchangers Pa ge 48

In the derivation o f the relationship the following assumptions were made:

1. End zone spacing is made equal to baffle spacing.

2. The flo w regim e in shell side is assumed to be turbulent (i.e. Res £ 100).

3. The baffle cut within the exchanger is selected such that the ’ window’ and

’cross-flow’ areas are equal.

This last assumption is a key one. It is also a very valid one. Designs in which

these two flo w areas are equalised are likely to be good designs. This is because one

objective in design should be to make the best possible use o f available pressure drop

in the promotion o f heat transfer. Designs which provide a smooth flo w path, result

in the major component o f pressure drop being frictional loss, and it is the frictional

loss that is being used to promote the heat transfer. Geometries in which the flo w is

being successively accelerated then decelerated waste pressure drop.

2.3.3 Simplification to Bell-Delaware model

In sub-section 2.3.1, w e saw that the shell side performance follow ed a simple

relationship (Equation 2.17), which had been derived from the heat transfer and

pressure drop correlations suggested by Kern [1950]. Although, the relationship was

quite simple the m odel was not accurate enough for modem heat exchanger design

purposes. So, a better m odel had to be used in order to make the relationship more

accurate and reliable.

C hapter 2 Pressure D rop Relationship f o r S & T heat exchangers Page 49

It was also demonstrated, in sub-section 2.3.2, how a new relationship (Equation

2.39) could be derived fo r shell side performance, using Bell-Delaware model. This

new relationship is significantly more accurate, and hence could be used in heat

exchanger design and optimisation. H owever, the relationship is still too complex to

be used in process synthesis [Polley and Panjeh Shahi, 1991]. For process synthesis

w e need a methodology which is as independent o f exchanger detail as possible. W e

also need a procedure which is non-iterative.

In an attempt to derive a simpler relationship, a typical exchanger geometry was

rated (using the full Bell-Delaware method) o v e r a range o f throughputs and for a

range o f fluids. It was found that the shell-side performance o f the exchanger could

be correlated by an equation o f the form:

A P„ = k a A h ? (2.42)

H ow ever, whilst different exchanger geometries each gave relationships o f this

form , it was found that the value o f the exponent "m " varied from geometry to

geometry. Despite this d ifficulty w e can make a useful deduction. The result implies

that a simple relationship exists between pressure drop, exchanger area and heat

transfer coefficient for exchangers that exhibit geometrical similarity.

This conclusion is supported by Kern, fo r his equation is reported to apply to

exchangers having geometrical similarity (25% baffle cut, and baffle spacing equal

to shell diameter).
C hapter 2 Pressure D r o p Relationship f o r S& T heat exchangers Page 50

Unfortunately, the above form o f similarity is restrictive. What is needed is a form

which places the least possible restriction on design. One possible w ay o f achieving

this would be setting the design criterion that the cross-flow area o f the exchanger

should equal the window flo w area. This gives both geometrical similarity and

freedom in design. This proposal was examined by rating exchangers having this

similarity, using the full Bell-Delaware method. It was found that their performances

were well correlated by:

S P , = k , A h i 412 (2-«3)

It is expected that similar relationship could be developed, using other modem

detailed performance models.

The pressure drop referred to in this equation is that usefully em ployed in the

promotion o f heat transfer. N o allowance is made fo r pressure losses associated with

flo w through exchanger headers and nozzles. Thus, when the relationship is used in

network synthesis a slight adjustment (say 10%) on the full allowable pressure drop

for a stream should be made.

2,3.4 Extension o f the idea to state-of-the-art models

Nowadays, sophisticated "stream analysis" models form the core o f m odem heat

exchanger design methodologies. However, the amount o f computation needed in

these models is, to a large extent, preventing their use in true exchanger design

optimisation. What is perhaps o f even greater significance, is that it pushes the

Chapter 2 Pressure D ro p Relationship f o r S& T heat exchangers Page 51

detailed consideration o f heat exchangers "downstream" in the overall design process.

Rather than being considered by process engineers in the initial design stages, it

becomes part o f the later detailed design activity and restricted to the field of

specialist engineers. This can result in serious lost opportunities [Polley, Panjeh

Shahi, 1991].

Therefore, there is a need for rapid but accurate design procedures that can be

used for exchanger optimisation and in overall process synthesis. Such procedures do

not replace the sophisticated models. For these are still needed for thorough design

analysis.

T o develop such a rapid design procedure, again w e need a sim ple relationship

for the shell-side performance.

So far, w e found that the shell side performance o f exchangers with geometrical

similarity fo llo w ed a simple relationship o f the form shown in Equation 2.42. The

relationship was confirmed by Kern (as an "integral") and Bell-Delaware (as an

"analytical") models.

This finding has significance for the extension o f the approach to current

commercial shell-side performance models. By applying this similarity concept with

these models, it may be possible to derive simple relationships o f the above form,

that can be used in network synthesis as well as heat exchanger design algorithm.
Chapter 2 Pressure D ro p Relationship fo r S & T h e a t exchangers P a ge 52

The resulting relationship could then be applied to develop:

1. A rapid algorithm which makes f u ll use o f stream pressure drop and

produces designs which are consistent with modern shell-side

perform ance models.

2. An a lgorith m f o r network area targeting, that gives consistent result

with detailed exchanger design conducted using those com m ercial

programs.
C hapter 2 Pressure D r o p Relationship f o r S & T heat exchangers Page 53

2.4 Summary

1. There is a need for a simple expression that relates the allowable pressure drop

o f a stream to the surface area and the film heat transfer coefficient o f the stream.

2. Various integral and analytical models fo r the shell-side performance have been

surveyed and amongst them the B ell-D elaw are method has been selected (due to

accuracy and reliability o f the results) as a basis for the development o f a new

pressure drop relationship.

3. It has been shown how various equations constituting Bell-Delaware method

can be combined and rearranged to g iv e a new pressure drop relationship.

4. It has been found that a simple relationship exists between pressure drop,

exchanger area and heat transfer coefficient for exchangers that exhibit geometrical

similarity. This conclusion has been supported by Kern and confirmed by Bell-

Delaware model, and could subsequently be extended to modem and commercial heat

exchanger design methodologies.

C ha p ter 2 Pressure D ro p Relationship f o r S& T heat exchangers Pa ge 54

Nomenclature

A heat transfer surface based on outside tube area

Cp fluid specific heat at average temperature

D, tube outside diameter

Dw hydraulic diameter o f baffle window

fj friction factor for an ideal tube bank

Ft correction factor to the A T LM for non-counter flo w system

h heat transfer coefficient

h4 heat transfer coefficient for an ideal tube bank

h, shell-side heat transfer coefficient

h, tube-side heat transfer coefficient

jj Colburn’s heat transfer factor for ideal tube bank

Jb bundle bypass correction factor for heat transfer

Jc segmental baffle window correction factor for heat transfer

J, baffle leakage correction factor for heat transfer

Jr laminar flo w correction factor for heat transfer

J, unequal baffle spacing correction factor for heat transfer

k, shell-side constant for A P relationship (Kern’ s method)

k,, shell-side constant for A P relationship (B ell’s method)

k .2 shell-side constant for A P relationship (B ell’s method)

k, tube-side constant for A P relationship

L,,,. central baffle spacing

L, tube length
Chapter 2 Pressure D ro p Relationship fo r S& T h e a t exchangers P a g e 55

L lp tube layout pitch

m mass velocity

mw shell-side flo w mass velocity through segmental baffle window

Nb number o f baffles

Nc total number o f tube rows crossed in the exchanger or any number o f tube

rows crossed

N ICC number o f tube rows crossed between baffle tips o f one baffle compartment

N tcw number o f tube rows crossed in one baffle window

N lt total number o f tubes

Pr Prandtl number

Q heat duty o f the exchanger

Rb bundle bypass correction factor for pressure drop

Re Reynolds number

Re, shell-side Reynolds number

R, baffle leakage correction factor for pressure drop

R, baffle end zones correction factor for pressure drop

U overall heat transfer coefficient based on outside tube area

U, shell-side fluid velocity

A P bj cross-flow pressure drop in an equivalent ideal tube bank

APC pressure drop in cross flow between baffle tips

AP. pressure drop in the end zones

AP, shell-side pressure drop

AP, tube-side pressure drop

APW pressure drop in window o f a segmentally baffled exchanger

Chapter 2 Pressure D r o p R elationship f o r S & T heat exchangers Page 56

A Pwi baffle w in dow pressure drop in an equivalent ideal tube bank

A T lm logarithmic mean temperature difference

m shell-side fluid viscosity at average temperature

Q fluid density at average temperature

Q, shell-side flu id density at average temperature

<}>s viscosity correction factor for shell-side fluid

C hapter 2 Pressure D ro p Relationship f o r S& T heat exchangers Page 57

References

Bell K J.,
’ Exchangers design Based on the Delaware Research Program’ ,
Petroleum Eng., voi. 32, no. 11, pp. C26-36. C40a-C40c, 1960.

B ell K J .,
’ Final Report o f the Cooperative Research Program on Shell-and-tube Heat
Exchangers’ ,
University o f Delaware Eng. Exp. Sta. Bull. 5, 1963.

Coulson J.M., Richardson J.F., Bachhurst J.R. and Harker J.H.,

’ A n Introduction to Chemical Engineering D esign ’ ,
’Chem. Eng. Ser., voi. 6, 1989.

Donohue D .A .,
Petroleum R efiner, Part 1, August 1955; Part 2, October 1955; Part 3, November
1955.

Gilmour C.H.,
’Short Cut to Heat Exchanger Design’ ,
Chem. Eng., Parts 1 through 7, 1952-54.

Gupta R .K. and Katz D.L.,

’ Flow Patterns fo r Predicting Shell-side Heat Transfer Coefficients fo r Baffled Shell-
and-tube Exchangers’ ,
Ind. Eng. Chem ., vol. 49, no. 6, pp. 998-999, 1957.

Jegede F.O.,
’ Power, Capital and Energy Costs Trade-offs in Heat Exchanger Netw orks’ ,
Ph.D. Thesis, U M IS T , Manchester, 1990.

Kern D.Q.,
’Process Heat Transfer’ ,
M cG raw -H ill, 1950.

K opp S., Sennstrom H.R. and Gunter A .Y .,

’ A Study o f F lo w Patterns in Baffled Heat Exchangers’ ,
A S M E Paper 47-A-103, 1947.
Chapter 2 Pressure D r o p Relationship f o r S & T heat exchangers Page 58

Palen J.W. and Taborek J.,

’ Solution o f Shell-side Flow Pressure Drop and Heat Transfer by Stream Analysis
M ethod’ ,
Chem. Eng. Prog. Symp. Ser., vol. 65. no. 92, 1969.

P olley G .T., Panjeh Shahi M .H.,

’ Interfacing Heat Exchanger N etw ork Synthesis and Detailed Exchanger Design’ ,
Trans. I.Chem.E. 69A. pp. 445-457. 1991.

Polley G .T., Panjeh Shahi M.H. and Jegede F.O.,

’ Pressure Drop Considerations in the Retrofit of Heat Exchanger N etw ork s’ ,
Trans. I.Chem.E. 68A, pp. 211-220, 1990.

Saunders E.A.D.,
’ Heat Exchangers - Selection, Design and Construction’ ,
Longman Science & Technical. 1988.

Taborek J.,
’ Shell-and-tube Heat Exchangers’ ,
Section 3.3 Heat Exchanger Design Handbook, Hemisphere, 1983.

Tinker T.,
J. Heat Transfer, vol. 80, pp. 36-52, 1958.
Chapter 3 Thermal Design o f S & T Heat Exchangers P a ge 59

C M A Pin m s

THERMAL DESIGN

OF

SHELL-AND-TUBE HEAT EXCHANGERS

( NEW ALGORITHM )
C hapter 3 Thermal Design o f S& T Heat Exchangers Pa ge 60

Contents

3.1 Introduction

3.2 Recommended Design Algorithm

3.2.1 Current practice in shell & tube heat exchanger design

3.2.2 Quick algorithm for shell-and-tube exchanger design

3.2.3 N ew algorithm fo r shell-and-tube heat exchanger design

3.3 Example Problem and Comparison

3.3.1 Example problem

3.3.2 Comparison o f the results

3.4 Summary

Nomenclature

References
C h a p ter 3 Thermal Design o f S& T Heat Exchangers Pa ge 61

3.1 Introduction

The shell-and-tube heat exchanger is the most common o f the various types of

unfired heat transfer equipment used in industry. Although it is not especially

compact, it is robust and its shape makes it w ell suited to pressure operation. It is

versatile and it can be designed to suit almost any application. Except for the special

purpo.se air cooled heat exchanger, it is usually the only type which can be

considered for large surface areas having pressures greater than 30 bar and

temperatures greater than 260*C [Saunders. 1988].

T h e shell-and-tube heat exchanger first appeared in the 1900s and was designed

to accommodate the demands o f the emerging oil industry. O il heaters and coolers,

reboilers, and condensers for a variety o f crude oil fractions and related organic

liquids were required fo r rugged outdoor service, often handling "dirty" fluids, and

involving high temperatures and pressures. Ease o f cleaning and field repairs was

unconditionally required [Taborek, 1983].

Chapter 3 Therm al Design o f S& T H ea t Exchangers P a ge 62

3.2 Recom m ended Design Algorithm

If we are going to design networks, in which stream pressure drops are to be

specified in targeting stage, w e need an appropriate exchanger design algorithm to

ensure that the exchangers use the requisite amount o f pressure drop.

Existing exchanger algorithm do not do this. They treat allowable pressure drop

as a constraint rather than as an objective. A s a consequence, any design which

achieves the required heat load with both tube-side and shell-side pressure drops less

than the specified value are considered to be acceptable design. These algorithms do

not, therefore, result in optimum designs for these are likely to be ones which make

full effective use o f both pressure drops.

Fortunately, the new pressure drop relationships, presented in Chapter 2, allow us

to see a way through this problem. W e have equations for pressure drop for both

tube-side and shell-side performance. These equations can be solved simultaneously

with the basic exchanger design equation to yield the area o f the exchanger and both

film heat transfer coefficients.

In this section w e first explain the conventional methods o f heat exchanger design

and associated drawbacks. Then, a quick design algorithm (based on Kern model) is

presented, to show how it is basically possible to utilise both available pressure drops.

Finally, a new design procedure that is accurate and ensures both specified pressure

drops are fully exploited w ill be described.

Chapter 3 Thermal Design o f S & T H e a t Exchangers Page 63

3.2.1 C u rre n t practice in shell & tube heat exchanger design

The process o f designing shell-and-tube heat exchangers, differs from performance

evaluation (or, rating calculations) in the fo llo w in g respects:

1. T h e process conditions are fully specified, but limits are imposed on

m aximum allowable pressure drop and/or flo w velocities, for both shell

side and tube side.

2. A s a consequence o f (1), a geometric configuration must be found to

accomplish the thermal change on the streams within these allowable

pressure drops in a most economic way.

3. It is then up to the exchanger designer (a) to select the basic construc­

tion elements such as shell type, tube size and pitch, tube bundle layout,

baffle cut and baffle spacing, number o f tube passes, etc., and (b) to find

the ultimate design dimensions o f shell diameter and tube length that will

result in complete thermo-hydraulic exchanger specification.

These comments suggest that, given a minimum o f restriction, a large number of

potential solutions is possible but only a fe w w ill be truly eligible, and ultimately

only one w ill emerge as best overall.

Chapter 3 T h erm al D esig n o f S& T Heat Exchangers Page 64

Nowadays, computer aided design o f heat exchangers has reached a high level of

sophistication. For instance, in the case of shell-and-tube units state-of-the-art

programs typically contain m odels which attempt to model the llo w through the

exchanger shell, bundle vibration analysis, and even important elements of

mechanical design. This sophistication is bought at a price. The designer must make

the subjective decisions, m entioned above, before he or she is able to use such

programs.

In next sub-section a new algorithm for the rapid sizing o f shell-and-tube heat

exchangers is presented, by w hich the designer can make those important decisions

on a much more objective basis. T h is algorithm is based on full use o f the allowable

pressure drops on both tube side and shell side o f the exchanger.

3.2.2 Q uick algorithm fo r shell-and-tube heat exchanger design

In a recent paper [Polley et al, 1990] it has been demonstrated how simple

relationships between fluid exchanger pressure drop, exchanger area and film heat

transfer coefficient can be used to rapidly design shell-and-tube heat exchangers.

It was shown that for tube side performance the follow ing relationship exists:

APt - k eA h î 5 (3-D

where the constant k, is dependent on volumetric flow-rate, the physical properties

o f the stream and the diameter o f the tube.

C hapter 3 Thermal Design o f S & T Heat Exchangers Page 65

It was also shown that for shell side performance the follow in g equation could be

derived from the correlations proposed by Kern [1950]:

A P a = k a A h i ml (3 .2)

where k, is again dependent on volumetric flow-rate, flu id physical properties and a

single characteristic dimensions (this time the bundle equivalent diameter).

These two equations can be solved together with the basic design equation:

Q = U A FT A T^ (3 .3)

to yield the tw o heat transfer coefficients and the required exchanger area. The details

o f the exchanger geometry can then be determined fro m these values as follows:

The tube-side heat transfer coefficient can be d irectly related to tube-side

velocity and hence to exchanger tube-count.

From tube-count and total surface area the tube length can be determined.

The shell diameter can then be computed from the tube-count.

Finally, with the shell diameter known and the shell-side velocity being

calculated from the shell-side coefficient, the number o f baffles and baffle

spacing needed within the exchanger can be determined.

C hapter 3 Thermal Design o f S & T Heat Exchangers Page 66

Here we have a very rapid design algorithm which avoids the searching o f many

potential geom etries and at the same time ensures the full use o f both allowable

pressure drops.

However, the method is restricted. The first restriction is that the pressure drop

referred to in the ab ove equations is that associated with flo w through the exchanger

bundle. N o account is taken o f any nozzle or header pressure drops. A llo w a n c e for

these must be made ahead o f design and checked after design. This restriction is not

considered to be a serious impediment.

The second restriction is the use o f the Kern correlations which (as discussed

above) are too inaccurate for use in modem heat exchanger design. T h e work

presented in next sub-section shows how Bell-Delaware model can be incorporated

into the basic design philosophy.

3.2.3 N ew a lg o rith m fo r shell-and-tube heat exchanger design

It was demonstrated above that the simple relationships o f the Kern ty p e allow

rapid exchanger sizing. The modem sophisticated models do not. What w e need are

simple models that g iv e consistent results with the predictions o f the better, more

accurate, and now w e ll established, flo w models.

C hapter 3 Thermal Design o f S & T Heat Exchangers Page 67

In chapter 2, it was shown how a sim ple relationship between shell-side fluid

pressure drop, exchanger area and shell-side heat transfer coefficient can be obtained,

using full Bell-Delaware model. The relationship is

A P, = ( k sl A * k , 2) h .2 <3.0

Where, both o f the constants are com plex functions involving physical properties,

shell geometry, ideal friction factor and ideal heat transfer j-factor.

The objective o f this work is a design algorithm in which exchanger geometry is

determined from required performance. A simultaneous solution o f Equations 3.1, 3.3

and 3.4 provides the core o f such an algorithm. Given the complex nature o f the

constants appearing in Equation 3.4, som e iteration with detailed exchanger geometry

becomes necessary. H owever, experience shows that the convergence is rapid.

A n outline o f the new algorithm, that has recently been published [Polley et al,

1991], is given in Figure 3.1.

C hapter 3 Thermal Design o f S & T Heat Exchangers P a ge 68

Figure 3.1.a : N ew design algorithm for shell-and-tube heat exchangers.

C hapter 3 Thermal Design o f S & T Heat Exchangers P a g e 69

Figure 3.1.C : N e w design algorithm for S & T heat exchangers (continued).

Chapter 3 Thermal Design o f S& T H ea t Exchangers P a ge 70

Figure 3.1.C : N ew design algorithm for S & T heat exchangers (continued).

C hapter 3 Therm al Design o f S & T Heat Exchangers Page 71

In the new algorithm, shell-side friction factor, heat transfer j-factor, baffle cut,

and shell diameter are all set at standard initial values. In case o f shell diameter, the

initial value is set at a value proportional to shell-side flow-rate for quicker

convergence. This allows initial estimates o f tube-count and baffle spacing, and

subsequently leakage and bypass areas (either based on T E M A [1978] standards or

specified clearances) and correction factors to be made for shell-side heat transfer

coefficient and pressure drop. T h e constants in Equation 3.4 can now be calculated.

T h e three simultaneous equations are then solved to yield heat transfer coefficients

and heat exchanger area. This in turn allows the calculation o f tube length, number

o f baffles, tube-side Reynolds number and velocity, tube-count and shell diameter,

respectively. Here, the calculated value for shell diameter can be tested against the

initial guess and updated if necessary.

N ex t step is to calculate shell-side pressure drop from the obtained geom etry and

compare that with specified value. If any discrepancy exists, b affle cut should be

changed within the permissible range, given by Bell [1963], (i.e. 15%-45%) and AP,

recalculated. T o do this more efficiently, one o f the standard search routines, such as

Brent method [Press et al., 1988], can be applied.

From the detailed geometry actual friction factors and j-factors are estimated and

the initial assumptions tested and updated if necessary. Once, initial assumptions and

detailed geom etry coincide, the design has been successfully accomplished.
C hapter 3 Thermal Design o f S& T Heat Exchangers P a ge 72

The input data required by the method is listed in Table 3.1.

Geometrical Data Process Data

tube i.d. mass flow-rates

tube o.d. tube-side A P
tube pitch shell-side A P
tube bundle layout inlet & outlet temperatures
no. o f tube passes fouling resistances
clearances Physical Properties

Table 3.1 : Input data for shell-and-tube exchanger design.

As it is seen, the input information is significantly less than is required by conven­

tional methods.

The new procedure has been used for the detailed design o f three Heat Exchanger

Network case studies (more than 60 units, all together) in this thesis, and proved to

be accurate and efficient.

C ha p ter 3 Thermal Design o f S& T Heat Exchangers P a ge 73

3.3 Exam ple Problem and Com parison

3.3.1 Exam ple problem

A typical example o f designing a shell-and-tube heat exchanger for a fully

specified process w ill be demonstrated. The example used here is an adaptation of

the one used by Taborek [1983] to demonstrate the Bell-Delaware method.

T h e original example involved a light oil that was to be heated from 89.8 to

147.3*C, flow ing on the shell side. The heat medium was isothermal steam on the

tube side at 178*C, with an assumed heat transfer coefficient o f 6000 W/m2 *C. The

heat duty required from process specifications (shell side) was 3100 kW.

In order to observe the effect o f oversizing on both shell-side and tube-side heat

transfer coefficient, however, it has been assumed that water is flow ing through the

tube-side o f the exchanger at a flow-rate that gives a film heat transfer coefficient of

6000 W/m2 *C. Both shell-side and Tube-side temperatures have been adjusted, in

such a way that the E ffective Mean Temperature Difference and the heat duty remain

the same.

N o w , with the geometry set by Taborek, the tube-side and shell-side pressure

drops fo r this situation are 11.66 and 13.7 kPa, respectively. These are the allowable

A P ’ s subsequently used in our design. A full specification o f the design problem is

shown in Table 3.2.

C h a p te r 3 Thermal Design o f S & T H eat Exchangers P a ge 74

----------------------- 1
Process Data Units Shell side Tube side

Flow-rate kg/s 2 2 .4 7 7 .9 6

F lu id Density kg/m3 7 4 0 .0 1 0 0 0 .0

H ea t Capacity J/kg -C 2 4 0 7 .0 4 1 8 7 .0

V iscosity cps. 0 .4 9 4 1 .0 0

Therm al Cond. W/m *C 0 .1 0 5 0 .6 1

In let Temp. •c 1 0 0 .0 7 .1

O utlet Temp. *c 4 2 .5 16.6

A llo w a b le A P kPa 1 3.7 1 1 .6 6

Fou lin g Factor •c m 7 w 0.0 0.0

W a ll Resistance •C m2/W 0 .0 0 0 0 3

H eat Duty kW 3 1 0 0 .0

Geometrical Data

T u b e O.D. mm 1 6.0

T u b e I.D. mm 13.5

T u b e Layout deg 3 0 .0

T u b e Pitch mm 2 0 .8

Bundle/shell cl. mm 1 0.0

Baffle/shell cl. mm 5 .5

Baffle/tube cl. mm 0 .5

Table 3.2 : Problem specification for S & T heat exchanger design.

Chapter 3 Thermal Design o f S& T H eat Exchangers Page 75

3.3.2 C om parison o f the results

The heat exchanger design derived using the new algorithm is compared with that

recommended by Taborek in Table 3.3. The new design makes full use o f the

available pressure drop and achieves an overall heat transfer coefficient o f 1088

W/m2 *C, and the required area is 52.3 m2. The Taborek design has an overall heat

transfer coefficient o f 971 W/mx *C, with a required area o f 58.9 m2. (The

performance o f both exchangers have been confirmed using a rating program).

Because o f the w ay in which the problem is set up both designs make full use o f

AP. However, the Taborek design is oversized having an installed area o f 74.6 m2

(giving a ratio o f required/installed area o f 0.79). W e see here a danger inherent in

exchanger oversizing. T h e additional surface area gives rise to an additional flo w

resistance and a need to reduce velocities and film heat transfer coefficients, in order

to meet the pressure drop constraint. The more area we add the less efficiently we

use the area we really need.

Chapter 3 Thermal Design o f S& T H e a t Exchangers P a ge 76

N e w Algorithm Taborek

Geometry

Shell diameter mm 563.0 591.0

Tube length mm 1815.0 2276.0

Baffle cut % 29.3 24.0

Baffle spacing mm 253.0 250.0

No. o f baffles 6 8

No. o f tubes 574 652

No. o f passes 2 2

Installed Area m2 52.3 74.6

Required area m2 52.3 58.9

Performance

Shell-side Re 21398 20672

Shell-side A P kPa 13.7 13.7

Tube-side A P kPa 11.69 11.66

Shell-side coef. W/m2 'C 1406 1241

Tube-side coef. W/m2 'C 6641 6000

i

O.H.T.C. 1088 971

Table 3.3 : Comparison o f new algorithm and Taborek design method.

C hapter 3 Thermal Design o f S & T Heat Exchangers P a ge 77

3.4 S u m m a ry

1. N e w rapid design algorithm has been developed fo r shell-and-tube heat

exchangers. This is based on the solution o f a small set o f simultaneous equations

and ensures that full use is made o f the allowable pressure drops o f both o f the

streams.

2. T h ere is a need for this algorithm. It appears to be the first one which makes

fu ll use o f both allowable pressure drops and thereby identifies the smallest

exchanger fo r a given duty.

3. It has been shown how the Bell-Delaware analytical method can, in spite o f its

com plexity, be incorporated into the design algorithm to obtain m ore accurate results

than integral models.

4. Because o f application o f the Bell-Delaware model, the algorithm appreciates

different b a ffle leakage and bundle bypass streams, and hence ensures that the shell-

side pressure drop is not overestimated. It should be noted that, since the design is

often based on a permissible value o f pressure drop, overprediction o f pressure drop

may seriously affect the heat transfer design o f the exchanger and may result in an

oversized unit.
C hapter 3 T h erm a l Design o f S & T H e a t Exchangers P a g e 78

N om en clatu re

A heat transfer surface based on outside tube area

Ft correction factor to the A T LM for non-counter flo w system

h, shell-side heat transfer coefficient

h, tube-side heat transfer coefficient

k, shell-side constant fo r A P relationship (K en t's method)

k,, shell-side constant fo r A P relationship (B e ll’s method)

k ,2 shell-side constant fo r A P relationship (B e ll’s method)

k, tube-side constant fo r A P relationship

Q heat duty o f the exchanger

U overall heat transfer coefficient

AP, shell-side pressure drop

AP, tube-side pressure drop

A T lm logarithmic mean temperature difference

Chapter 3 Therm al Design o f Stft T H ea t Exchangers P a ge 79

R eferences

Bell K J .,
’ Final Report o f the Cooperative Research Program on Shell-and-tube Heat
Exchangers’ ,
University o f Delaware Eng. Exp. Sta. Bull. 5, 1963.

Kern D.Q.,
’ Process Heat Transfer’ ,
M cG raw -H ill, 1950.

P olley G.T., Panjeh Shahi M .H . and Jegede F.O.,

’Pressure Drop Considerations in the Retrofit o f Heat Exchanger Networks’ ,
Trans. I.Chem.E. 68A , pp. 211-220, 1990.

P olley G.T., Panjeh Shahi M .H . and Nunez M .P.,

’ Rapid Design Algorithms fo r Shell-and-tube and Compact Heat Exchangers’ ,
Trans. I.Chem.E. 69A, pp. 435-444, 1991.

Press W .H., Flannery B.P., Teukolsky S.A. and Vetterling W .T.,

’ Numerical Recipes - The Art o f Scientific Com puting’ ,
Cambridge University Press, 1988.

Saunders E.A.D.,
’ Heat Exchangers - Selection, Design and Construction’ ,
Longman Science & Technical, 1988.

Taborek J.,
’ Shell-and-tube Heat Exchangers’ ,
Section 3.3 Heat Exchanger Design Handbook, Hemisphere, 1983.

TEM A,
’Standards o f Tubular Exchangers Manufacturers Association (6th edn.)
N e w York, 1978.
Chapter 4 M inim um A rea o f H E N 's (New Algorithm ) P a ge 80

CHAPTER 4

M IN IM U M AREA

OF

HEAT EXCHANGER NETWORKS

( NEW ALGORITHM )
Chapter 4 M in im um A rea o f H E N 's (New A lg orith m ) Page 81

Contents

4.1 Introduction

4.2 Current Area Algorithm

4.2.1 Vertical model

4.2.2 Criss-crossing and its effects

4.3 N ew Area Algorithm

4.3.1 Equations required for new algorithm

4.3.2 The new area algorithm

4.3.3 Adding more flexibility

4.4 Summary

Nomenclature

References
C hapter 4 Minim um Area o f U R N 's ( New A lgorithm ) Page 82

4.1 Introdu ction

In current practice assumptions are made, prior to targeting and network design,

regarding stream heat transfer coefficients. These assumed coefficients are first used

in order to predict the minimum area requirement o f the heat exchanger network

(targeting), and then applied during network synthesis.

In chapter 1, w e demonstrated that assumed coefficients could lead to serious

design mistakes in both grass-root and retrofit designs. It was, hence, concluded that

stream pressure drops should be considered during targeting stage (through a new

area algorithm which base network area estimation on stream pressure drops and not

assumed heat transfer coefficients) as w ell as network synthesis.

In chapter 2, w e presented the relationships between exchanger area, stream film

heat transfer coefficient and pressure drop. Given these relationships, w e have the

possibility o f basing network area prediction on stream pressure drop rather than

assumed heat transfer coefficient.

This approach has a number o f advantages. First, it rem oves subjectivity in the

selection o f the coefficients used in targeting. Second, it makes the targeting

procedure consistent with heat exchanger design procedures. Third, and perhaps more

important, it allows us to cater for the fa ll-o ff in coefficients that w e expect during

retrofits which are subjected to pressure drop restrictions.

C hapter •/ M inim um A rea o f H E N 's (N ew A lgorithm ) Page 83

4.2 Current A r e a A lgorith m

4.2.1 V ertical m odel

The way in which knowledge o f process stream temperatures and enthalpies can

be used for the construction o f ’composite curves’ is now w ell known [Linnhoff et

all, 1979]. These curves are a key development in Pinch Technology, for they can be

used to undertake tw o important tasks. First, by superimposing the hot composite

curve on the cold one, such that they are separated by an assumed minimum

temperature approach (A T min), the minimum energy requirement o f a process can be

determined (the ’overshoot’ in Figure 4.1). Second, they can be applied to estimate

the minimum requirement o f the overall heat exchange area o f a network (A*,*).

Townsend and Linnhoff [1984] have shown that, in a network in which the heat

transfer coefficients are uniform, the overall heat exchange area is minimised if the

stream matches are "vertically a lign ed " between the composite curves (Figure 4.1).

In this arrangement the heat is transferred with maximum possible driving forces.

This model is analogous to pure counter-current flo w in a heat exchanger.

A match away from vertical (to the left thereby givin g a larger A T ) w ill gain a

local advantage. H ow ever, because the problem is in heat balance, it w ill later require

a match in the opposite direction (to the right giving a smaller A T , Figure 4.2). The

net effect o f such "criss-crossing " is an increase in area requirement. The greater the

amount o f criss-crossing, the higher the surface area requirement o f the network.
C hapter 4 Minim um Area o f H E N 's (New A lg orith m ) Page 84

Qa

Figure 4.1 : Vertical heat transfer between the composite curves.

Qa

Figure 4.2 : Criss-crossing heat transfer between the composite curves.

Chapter 4 Minim um A rea o f U R N 's (New A lg orith m ) Rage 85

The minimum heat exchange area (A „lin) for the network can be estimated from

’ enthalpy intervals’ on the composite curves, using the equation presented by

Townsend and L in n h o ff [1984]:

(4.1 )

where

AT, M.j : the logarithm ic mean temperature difference o f enthalpy interval i o f the

com posite curves

q,j : the heat change o f stream j in enthalpy interval i

hj : the heat transfer coefficient o f stream j (including film , wall and fouling

resistance).

For a system in which all matches have the same heat transfer coefficients, the

assumption o f vertical stream matching leads to rigorous minimum surface area.

However, with netw orks in which each stream has a different coefficient, the vertical

alignment concept can still be applied in order to give a minimum area prediction

which approximates (generally to within less than 10%) the true minimum [Morton,

1986].
Chapter 4 M in im um A rea o f H E N ‘s (New A lgorith m ) Pa ge 86

4.2.2 Criss-crossing and its effects

It is understood that for a network featuring vertical heat transfer, its matches

must be arranged so that their individual temperature driving forces conform exactly

to the driving forces between the com posite curves. This often leads to a network

requiring an excessive number o f heat exchange units. Figure 4.3 shows a network

for problem containing one hot and three cold streams with only tw o enthalpy

intervals (this network is termed a "spaghetti" network).

Figure 4.3 : A "spaghetti" network.

C h a p te r 4 Minim um Area o f H EN 's (New A lg o rith m ) P a ge 8 7

In terms o f the T -H diagram all the heat transfer within the interval is vertical.

H ow ever, this "spaghetti" network is not likely to be acceptable in practice. A

practical number o f units is significantly less than that present in this arrangement

Figu re 4.4 shows a typical network for the same problem featuring fewer units. In

this network, criss-crossing has been introduced in order to reduce the number o f

units.

Since criss-crossing increases the network area requirement, good practical

networks usually feature a moderate degree o f criss-crossing [Ahmad et al, 1986].

Chapter 4 M in im u m Area o f U F.N 's (New A lgorithm ) P a g e 88

4.3 New A rea Algorithm

4.3.1 Equations required fo r new algorithm

Here, w e present new equations through which it becomes possible to incorporate

stream pressure drop into area algorithm. W e start with existing equation for n etw ork

area prediction (Equation 4 .1 ). This equation can be written in a revised form :

(4.2)

where

(4.3)

Aj is termed the "area contribution o f stream j " . W e also have a set o f pressure

drop equations (o f the form bel ow ) that relates the heat exchanger area to stream

pressure drops and heat transfer coefficients.

A P j - k j A cj ( h cj ) m (4.4)

where

APj : specified pressure drop o f stream *j*

lq : constant value relating to stream ’j*

A cj : exchanger area installed on stream ’j ’ (contact area)

h^ : clean heat transfer coefficient o f stream j ’

C ha p ter 4 Minim um Area o f H E N 's (N e w A lgorith m ) Page 89

The obvious approach would be to solve these equations simultaneously. But, here

w e encounter a problem. The network area equation is written in terms o f stream area

contribution, which is independent o f stream match and only takes account o f part

o f the thermal resistance present in any individual exchanger. This is acceptable when

summing them in order to find the total network area. However, when considering

pressure drop, it is the total surface area o f the exchanger that must be considered.

Because, the area term appearing in the pressure drop relationship is true contact (or

wetted) area and here both stream thermal resistances have a role. The difference

between the two areas is illustrated in Figure 4.5.

Figure 4.5 : Relationship between area contribution and contact area.

C hapter 4 M in im um Area o f H F.N's (New Algorithm ) Page 90

Therefore, in order to make use o f the pressure drop relationships in network area

prediction, we need to calculate "stream contact area" as w ell as area contribution.

This can be achieved by adding to the 'thermal resistance’ an allow ance based on the

’thermal resistance’ o f the opposite streams it is likely to be brought into contact

with.

A number o f ways o f obtaining such a prediction can be envisaged. However, the

simplest and the most suitable is the one which is based on the "spaghetti m odel" that

underlies the Townsend Equation.

Having used the spaghetti model the stream contact area is giv e n by:

where

A^ : contact area o f stream ’j ’

: the opposing resistance o f stream ’j ’ within a given interval

K : number o f opposing streams to stream ’j ’ within that interval

CPk : specific heat o f stream ’ k ’

hk : heat transfer coefficient o f stream ’k ’

Chapter 4 M inim um Area o f I lE N 's (N ew Algorithm ) Page 91

The network area is now expressed in terms o f stream contact areas as follow s

(4.7)

Another problem is the difference between heat transfer coefficient used in

pressure drop relationship (clean h) and in network area equation (fouled h). This is

easily solved by applying the follow ing equation:

where is the specified fouling resistance o f the stream j\

W e now have the equations, necessary fo r the development o f a new area

algorithm. B y solving the network area equation (4.7) together with ’J’ (total number

o f streams) contact area equations (4.5) and ’J* pressure drop relationships (Equation

4.4), predictions for the network minimum area, ’J’ stream contact areas and \J’

stream heat transfer coefficients are obtained (i.e: 2J+1 equations and 2J+1

unknowns). The obtained coefficients are now consistent with specified stream

pressure drops.

4.3.2 The new area algorithm

An outline o f the new area algorithm is illustrated in Figures 4.6a and 4.6b.
C hapter 4 M inim um Area o f H E N 's (N ew A lgorith m ) Page 92

Figure 4.6a : N ew algorithm based on stream pressure drops.

C hapter -/ M inim um A rea o f H E N ’s (N ew A lgorith m ) P a ge 93

Figure 4.6b : Area algorithm for predicting the stream contact areas.
Chapter 4 M inim um Area o f ¡¡F .N ’s (New Algorithm ) Page 94

4.3.3 Adding more flexibility

So far, we have changed the basis o f the network area prediction from assumed

heat transfer coefficient to stream pressure drop. This would resolve the three major

drawbacks that exist in current procedures (described in chapter 1), and make it

possible to specify pressure drops right at the beginning o f the targeting stage. But

we need more flexibility. Although, some o f the process streams are dependent on

pressure drop, some others are not. For instance, for streams involving condensation

or boiling, film heat transfer coefficient is the crucial factor in prediction o f the heat

exchange area, whilst with the high pressure drop and/or fouling liquids, velocity is

the parameter that should be controlled during detailed design. Therefore, it is

desirable to have different specification options when defining streams o f different

types (Table 4.1).

T yp e Specification When used

A Pre-specified Boiling
Coefficient Condensation

B Pre-specified Gravity Controlled

Pressure Drop Flows

C Pre-optimised Compressed Gases

Pressure Drop Pumping o f Clean Liquids

D Selected High A llow able Pressure Drop

V elocity Fouling Liquids

Table 4.1 : Classification o f streams.

Chapter */ Minim um Area o f H E N ’s (N ew A lgorith m ) Pa ge 95

Fortunately, the new area algorithm allows this flexibility. Because, as long as the

pressure drop equation forms the core o f the algorithm, pressure drop, heat transfer

coefficient and velocity are related to each other. Hence, it would be possible (in

Figure 4.6a) to update each stream coefficient according to its specified parameter.

The result is a fully flexible area algorithm that allows individual streams to be

specified as they should be.

Chapter -1 M inim um A rea o f H E N 's (N ew A lgorithm ) Pa ge 96

4.4 S u m m ary

Current algorithms for prediction o f the network minimum area are based on

assumed heat transfer coefficients, and do not consider stream pressure drops. This

could lead to serious problems.

T o avoid these problems, a new Area Algorithm has been developed, using

pressure d ro p relationship (presented in chapter 2). With this algorithm, it would be

possible to use allowable pressure drops rather than heat transfer coefficients.

F lex ib ility has also been added to the algorithm, in such a way that allows the

process stream s to be specified in different types, such as pre-specified coefficient,

pre-specified pressure drop, pre-optimised pressure drop, or even selected velocity.

Chapter 4 M in im um A r e a o f H E N ’s (N ew A lgorith m )

Nomenclature

A heat exchanger surface area

contact area o f stream ’j ’

Aj area contribution o f stream *j*

Ajnù, minimum area requirement o f a heat exchanger network

CPk specific heat o f stream ’ k ’

h^ clean heat transfer co efficien t of stream *j’

hj heat transfer coefficient o f stream *j’

hk heat transfer coefficient o f stream ’k ’

J total number o f streams

K number o f opposing streams to stream *j* in a given interval

kj constant value in pressure drop relationship (for stream ’j ’ )

heat change o f stream ’j ’ within interval ’ i ’

fouling resistance o f stream *j*

R qppj opposing resistance to stream ’j ’ in a given interval

APj pressure drop o f stream *j*

AT temperature driving force

^ T LMi logarithmic mean temperature difference in interval V

A T min minimum temperature approach on composite curves

Chapter */ M in im um Area o f H E N 's (N ew A lg o r ith m ) Page 98

References

Ahmad S., Linnhoff B., Morton RJ., T joe T.N ., and Tow nsend D.W.,
’ Process Integration Research Report’ ,
U M 1ST, 1986.

Linnhoff B., Mason D.R. and Wardle I.,

Comp. & Chem. Eng., 3, pp. 283-291, 1979.

Morton R.J.,
’ Practical Heat Exchanger Network: Capital Cost, Exchanger Type and Utilities’ ,
Ph.D. Thesis, U M IS T , 1986.

Townsend D .W . and Linnhoff B.,

’ Surface Area Targets for Heat Exchanger Networks’ ,
I.Chem.E. Annual Research Meeting, Bath, U K, A p ril 1984.
Chapter 5 Grassroots Design o f U F.N's (New P rocedure) P a g e 99

CHAPTERS

GRASS-ROOTS DESIGN

OF

HEAT EXCHANGER NETWORKS

(NEW PROCEDURE)
Chapter 5 G rassroots Design o f H E N 's (N ew P rocedure) Page 100

Contents

5.1 Introduction

5.2 Supertargeting

5.2.1 Targeting fo r minimum total annualized cost

5.2.2 Minimum number o f units

5.2.3 C oefficients to be used in Network Synthesis

5.2.4 Selection o f stream pressure drops

5.3 Network Synthesis

5.3.1 Pinch Design Method

5.3.2 Network refinement

5.4 Case Study

5.4.1 The problem (Aromatics Plant)

5.4.1.1 Process flowsheet

5.4.1.2 Process description

5.4.1.3 Process data

5.4.2 Supertargeting (Aromatics Plant)

5.4.2.1 Trading o ff energy against network capital cost

5.4.2.2 Coefficients to be used in network synthesis

5.4.3 Network synthesis (Aromatics Plant)

5.4.3.1 Preliminary design

5.4.3.2 Final design

Chapter 5 Grassroots Design o f H E N ’s (N ew Procedure) Page 102

5.1 Introduction

As explained in Chapter 1. the current procedures for the design o f new heat

exchanger networks suffer from two major problems, caused by the difference

between the assumed values for film coefficients (during the targeting stage) and the

true values achieved after detailed exchanger design. These problems are

- incorrect capital-energy trade-off and network optimisation

- inconsistency between network synthesis and detailed design, in terms o f

exchanger surface area.

It was also concluded that estimating the network area on the basis o f stream

pressure drops, instead o f stream coefficients, could help to overcome these problems.

For this reason, a new area algorithm was developed in Chapter 4.

Now. it is about time to apply this new area algorithm, in a systematic grass-root

design procedure, and see how accurate the result produced. In this chapter, we first

discuss the different stages in grass-root design (targeting, network synthesis and

network optimisation). Then, the procedure is tested using a case study.

C ha p ter 5 Grassroots Design o f H F .S's (N ew Proced u re) Page 103

5.2 Supertargeting

5.2.1 T a rgetin g fo r m inim um total annualized cost

For a given set o f stream data the minimum energy requirement (energy target)

can be estimated, ahead o f design, either from com posite curves (conceptually) or

from the Problem Table Algorithm presented by Linnhoff et al [1982]. Also, it is

possible to find the minimum area requirement (area target) for the same set of

stream data, using the new area algorithm presented in Chapter 4.

W e see that both energy and area targets are functions o f the selected minimum

temperature approach (A T min). Thus, by varying the temperature approach we can

develop a relationship between the minimum energy and area needs. A typical result

is presented in the ’ area-energy plot’ shown in Figure 5.1. Straight away we see some

bound on what we can achieve through a network. Plants cannot be operated below

this ’targeting curve’ .

For a given A T min, each target can be converted to an annualized cost target to

produce the total annualized cost. Therefore, by calculating the total annualized cost

o ver a range o f A T min, the near optimal energy-network capital trade-off can be

established.

A n outline o f the procedure is schematically shown in Figure 5.2.

C hapter 5 Grass m ats Design o f U T S 's (S ew Procedure) Pa ge 104

Figure 5.2 : Supertargeting procedure using new area algorithm.

C hapter 5 Grassroots Design o f U T S 's (N ew Procedu re) Page 105

5.2.2 Minimum number of units

The capital cost o f chemical processes tends to be dominated by the number of

items on the flowsheet. This is certainly true o f heat exchanger networks and there

is a strong incentive to reduce the number o f matches between hot and cold streams.

Hohmann [1971] has shown that for a network with N number o f streams

(including utilities), minimum number o f units (including heaters and coolers) can be

estimated by

IW n * N - 1 <5 ' ”

This minimum number o f units can be achieved by follow ing the principle o f

maximising loads (i.e. "ticking o ff" stream or utility loads or residuals).

The above equation was then modified by Linnhoff et al [1979]:

* N + L - s (5 .2 )

where

u^ : number o f units (including heaters and coolers)

N : number o f streams (including utilities)

L : number o f loops

s : number o f independent stream subsets

In sub-section 5.3.2 w e explain more about loops, and paths and show how they

can be exploited in order to reduce the network capital cost.

C hapter 5 Grassroots Design o f H E fit’s (New Procedu re) P a g e 106

5.2.3 Coefficients to be used in Network Synthesis

The new area algorithm gives a sharper relationship between network area and

energy consumption than that predicted on the basis o f assumed coefficients (Figure

5.3).

Figure 5.3 : Area-energy plot for both methods.

As the energy consumption o f the network is reduced more area is needed within

the network. With the stream pressure drops being fixed, the increased flow resistance

requires a reduction in the individual stream coefficients. The new model takes

account o f this necessary variation in film coefficients throughout the optimisation

between energy consumption and network area.

C hapter 5 G rassroots Design o f U E N 's (New Proced u re) Page 106

5-2-3__ Coefficients to be used in Network Synthesis

The new area algorithm gives a sharper relationship between network area and

energy consumption than that predicted on the basis o f assumed coefficients (Figure

5.3).

Figure 5.3 : Area-energy plot fo r both methods.

As the energy consumption o f the network is reduced more area is needed within

the network. With the stream pressure drops being fixed, the increased flo w resistance

requires a reduction in the individual stream coefficients. The new model takes

account o f this necessary variation in film coefficients throughout the optimisation

between energy consumption and network area.

C ha pter 5 Grassroots Design o f H F.N's (N ew P roced u re) P a ge 107

It can be seen. From Figure 4.7a, that the targeting procedure not only yields the

near optimum energy recovery level and network area requirement but also the heat

transfer coefficients to be used in design. G iven this information network synthesis

can again proceed using established procedures.

5.2.4 Selection o f Stream Pressure Drop

T h e procedure presented here is based on stream pressure drops rather than film

heat transfer coefficients and this makes it consistent with the way in which heat

exchanger designers work. But isn’t the approach sim ply changing one set of

objective decisions (that o f specifying stream film coefficients) for another (the

specification o f stream pressure drops)? T o answer this w e must differentiate between

two different flo w situations: forced flo w and flow by gravity. In the case o f forced

flo w the answer is yes, but industrial companies often have set standards for

allowable pressure drops (they do not have standards for film heat transfer

coefficients!). In the case o f gravity flows the answer is no. The engineer has a good

appreciation o f the allowable head with these streams.

T h e pressure drop relationships on which the procedure is based can be used in

the estimation o f fluid flo w power consumption. A procedure for the optimisation o f

forced flo w pressure drops in heat exchanger networks has been developed using

these relationships [Jegede. 1990]. Consequently, the subjectivity in the selection o f

stream pressure drops can be removed.

Chapter 5 Grassmnts Design o f H F N ’s (New Procedure) Page 108

5.3 Network Synthesis

When Pinch Technology is used in practical network synthesis three stages are

involved. T h e first stage is the 'targeting stage’ , which was explained in Section 5.2.

The objective here is to get close to the optimum energy consumption fo r the process.

The second stage involves a preliminary synthesis o f the heat recovery network. Here

the Pinch Design Method [Linnhoff et al. 1982] is used to produce a network which

operates at the energy consumption identified at the ’ targeting’ stage. The third stage

is network refinement. The next two sub-sections explain the second and the third

stages in more detail.

5.3.1 Pinch Design Method (design for Maximum Energy Recovery)

A b rief review o f the Pinch Design Method is included here. This is done, because

the method is the only one at present to combine sufficient simplicity fo r hand use

with near certainty to meet the targets, even for large industrial problems. It

guarantees the energy target being satisfied and. most often, the (U min) MER target

(minimum number o f units for Maximum Energy Recovery) met.

The method begins by recognising the Pinch location on the composite curves,

and subsequently divides the problem into tw o thermodynamically distinct regions

(above and below the Pinch). Then, to achieve MER, designs are developed in such

a w ay that:
Chapter 5 G rassroots Design o f lIF .N 's (N ew Procedure) Page 109

- N o heat is transferred across the Pinch

- N o cold utilities is used above the Pinch

- N o hot utility is used below the Pinch.

In doing so. the two regions are designed separately by starting at the Pinch

(where the problem is most constrained) and movin£ awav. This means, above the

Pinch all hot streams crossing the Pinch must be brought to the Pinch temperature

only by interchange with cold streams above the Pinch. Similarly, below the Pinch

all cold streams crossing the Pinch should reach the Pinch temperature only by

interchange against hot streams below the Pinch.

Also, minimum number o f units is calculated (using equation 5.2) fo r each region

separately. Therefore, the (unim)MKR is determined by

( ^ m ln ) MER ~ ( LZmin ^ ABOVE + ^ Um ln ^ BELOW *5*

This means the (umh,)NtFR is always greater than or equal to the minimum number

o f units.

During the design, the method directs the designers towards achieving the

minimum number o f units for M ER. This can often be done by em ploying a ’ tick-off’

heuristic to identify the heat exchanger loads.

Chapter 5 Grassroots Design o f ! IC S 's (S ew P roced u re)

5.3.2 Network refinement

Network refinement involves the energy/capital cost optimisation o f the specific

network structure developed in stage 2. It involves the exploitation o f the ’ loops and

paths’ [L in n h o ff et al, 1982] present in the structure. In the Pinch Design Method

individual sub-networks are generated for each side o f the Pinch. When these are

combined the initial result is a network having more units than necessary. Generally,

the capital cost o f the network can be reduced through using few er exchangers. This

is achieved b y identifying and breaking the ’ loops’ present in the network. During

the breaking o f a loop, the energy consumption o f the network is held constant. This

generally results in a network in which there is a violation o f the Pinch A T min. The

designer then has the option o f either accepting this violation or restoring the A T min

(fully or partially) using a ’ path’ . The restoration o f the A T niin in this manner

generally results in an energy penalty. However, this penalty can be compensated by

reduction in area requirement, and hence the overall result can be beneficial.

The procedure is fully described in I.Chem.E. U ser Guide [1982].

When the cost o f individual exchangers follow s a power law relationship, the

capital cost o f a network is dependent on the size distribution o f the exchangers. This

distribution can be manipulated by redistributing exchanger heat loadj within network

loops.
Chapter 5 Grassroots Design o f HF.N's (New Procedure) Page H I

Finally, at the initial synthesis stage we produce designs which, because they use

relatively few units, have some degree o f non-vertical stream matching. (This is why

it is common to find that the final network area requirement is a ft w percent higher

than the targeted value). During network refinement we can often im prove the

’ vertical ity’ o f the matching by redistributing heat loads within a loop.

The results o f network refinement are sensitive to initial network structure. The

Pinch Design Method can yield a number o f M E R solutions for a given problem. The

designer often only has time to examine one or two o f the available options. Thus,

the method cannot guarantee ’ optim ality’ in purist terms. It does yield practical cost

effective designs.

Acceptance o f designs is m ade on the basis o f the results o f network synthesis not

targeting. Since, targeting and synthesis are conducted by the same engineer as part

o f the same exercise, iteration between the stages (should it prove necessary) is easy.

As discussed in Chapter 1. this is not the case with the interface between network

synthesis and detailed exchanger design. What is essential fo r a practical

methodology is that the results o f network synthesis (not necessarily network

targeting) are consistent with those o f detailed exchanger design (See Chapter 6).
Chapter 5 Grassroots Design o f N F S ’s (Sew P roced u re) Page 112

5.4 Case Study

5.4.1 The problem (Aromatics Plant)

Perhaps the best industrial sized problem to test the new procedure is the

’ Aromatics Plant’ first presented in the I.Chem.E. User Guide [1982], and

subsequently simplified and used by Linnhoff and Ahm ad [1989] to demonstrate how

Pinch Technology could be used in the synthesis o f near optimal networks.

5.4.1.1 Process flowsheet

The process flowsheet is given in Figure 5.4.

Figure 5.4 : Flowsheet o f Aromatics Plant

C hapter 5 Grassroots Design o f H F N 's ( New Procedure) Page 113

5.4.1.2 Process description

Process description given here is based on the original process presented in the

I.Chem.E. U ser Guide. However, as a case study, the version that has b een simplified

by Linnhoff and Ahmad [1989] is used throughout this thesis.

" A heart cut o f naphtha from which both light and heavy ends have been stripped

is the principal plant feedstock. The feed is vaporised (H I ) and passed through a

desulphurisation reactor (R l). Heat is recovered from the reactor efflu en t in two

interchangers (A .B ) prior to condensation (C l) and gas separation ( F I ) . The liquid

from the separation stage is re-heated by reactor effluent (B ) and fe d to a stripping

column ( D l ) in which the light ends and sulphur-containing compounds are removed.

The desulphurised naphtha stream from the column is mixed with re c y c le gas. The

tw o phase mixture is preheated in a series o f process interchangers (D,C). The

mixture is fin ally raised to the reaction temperature o f 500*C by a radiant furnace

(H2) fired by a mixture o f gas and fuel oil. The reaction takes place in a series o f

reformers (R 2.R 3) with intermediate fired heating (H3). The reform er effluent, at

490*C. is cooled in a series o f interchangers. some preheating the fe e d (C,D,E) and

some providing the heat source for other process requirements (F ,X ).

Final coolin g and gas separation takes place in C2 and F2. T h e gas recycle is

compressed ( P I ) and preheated (E) prior to mixing with the liquid reform er feed. The

liquid from the flash drum is passed to a column for stabilisation (D 2) and a
C hapter 5 G rassroots Design o f H E N ’s (N ew Procedure)

conventional feed/tails interchanger (G ) is installed to reduce the reboil requirement

by adding feed preheat. T h e reformate stream is finally cooled in C3 prior to

storage."

5.4.1.3 Process data

T h e basic data for the problem are presented in Table 5.1. It should be noted that

the values fo r heat transfer coefficients w ill be used only for initialisation in new area

algorithm. T h e values to be used in network synthesis stage will be known after the

targeting is done.

Stream T, CC) T, CC) mCp (kW r C ) h (kW/m2 ‘ O

HI 327 40 100 0.5

H2 220 160 160 0.5

H3 220 60 60 0.5

H4 160 45 400 0.5

Cl 100 300 100 0.5

C2 35 164 70 0.5

C3 85 138 350 0.5

C4 60 170 60 0.5

C5 140 300 200 0.5

O IL 330 230 1.0

C.W . 10 30 2.5

Table 5.1 : Stream and U tility data for Aromatics Plant.

Chapter 5 Grass m ats Design o f H F N 's (N ew P rocedu re) Pa ge 115

N ow , we need information on fluid physical properties, stream fouling resistances

and stream allowable pressure drops. These are not given in the original problem, we

obtained physical properties from a data bank [Tables ,1987], fouling resistances from

the standards o f T E M A [1978], and stream pressure drops from the data on a typical

Aromatics flowsheet. These data are presented in Table 5.2.

Stream Q cp P k R, AP

kg/m3 J/kg ‘ C c Ps W/m *C m2 -C/W k Pa

HI 500 2000 0.25 0.11 0.00018 120

H2 55 2192 0.01 0.026 0.00014 80

H3 676 1877 0.28 0.11 0.00018 90

H4 697 5480 0.31 0.11 0.00018 60

Cl 464 2000 0.16 0.11 0.00018 20

C2 570 1590 0.30 0.11 0.00018 20

C3 1.0 10000 0.01 0.17 0.00014 30

C4 685 1580 0.27 0.11 0.00018 15

C5 667 2740 0.21 0.11 0.00018 80

Table 5.2 : Physical properties, fouling resistances and stream pressure drops for

Aromatics Plant.

In addition to above information we also need cost data, in order to be able to

trade-off between energy and network capital costs. This data is given in Table 5.3.
Chapter 5 Grassroots Design o f /IK S 's (New Procedu re) Page 116

Cost o f Hot Utility : 70 f/kw.yr Plant L ife T im e (N ) : 5 yrs

Cost o f Cold Utility : 7 £/kw.yr Rate o f Interest (i) : 15 %

Exchanger Capital Cost : 700 (Area)0"

Annualization Factor i ( i + l ) " / (< i+ l)N - 1}

Table 5.3 : Cost data fo r the Aromatics Plant.

Note : In order to obtain more realistic results, the technique for handling multiple

tube pass exchangers introduced by Ahmad and Smith [1989] has been used

throughout this thesis.

C hapter 5 Grassroots Design o f H F N 's (N ew Procedu re) Page ¡1 7

5.4.2 Sn n ertargetin g (A rom atics Plant)

5.4,2.1 T ra d in g o ff energy cost against n etw ork capital cost

The problem was solved in the stages described above. First, the energy/network

capital trade-off was examined. Table 5.4 presents the results o f supertargeting over

a wide range o f A T min.

As can be seen, the total cost in this table has a non-monotonic trend and a

relatively flat optimum. The near optimal energy recovery level occurs at tw o values

o f A T min (i.e. 15*C and 2 0 *0 , giving the total annualized costs o f 2348367 and

2385724 f/yr which are quite close to each other (1.5% difference). This is because

the Pinch location, and hence stream population on each side o f the Pinch, differs for

the values o f A T min below and above 20*C. This difference causes the slope o f the

’ Energy vs A T min* line to change at 20*C and. since the energy cost is a dominant

factor, the total cost to have a non-monotonic behaviour in this region.

However, the question arises when it com es to selection o f A T min fo r use in the

next stages, i;e. network synthesis and refinement. Each A T min, in this situation, can

result in an initial design having a different structure and energy recovery level. This

in turn affects the result o f network refinement and hence the final total annualized

cost. So, which one shall we select?

C ha p ter 5 Grassroots Design o f H F N ’s (New Procedure) Pa ge 118

STREAMS - AROM.STM U T IL IT IE S - AROM.UTY

GEOMETRY- AROM.GEO PHYS.PROP. - AROM.PRP

New D esign Range T a rg e t (F ix e d AP)

DTmin Hot U t. C old U t. 1-2 Area Nu Ns U t. C ost C a p .C o st T o t .C o s t

CC] [kW] (kW] [M2] [E / y r] [E / y r ] [£ / y r]

5.0 15130.00 22850.00 25125.73 15 94 1219050 2029104 3248154

6.0 15560.00 23280.00 21405.48 15 84 1252160 1742758 2994918
7 .0 15990.00 23710.00 18560.23 15 76 1285270 1522077 2807347
8.0 16420.00 24140.00 16704.32 15 70 1318380 1375272 2693652
9 .0 16850.00 24570.00 15666.45 15 65 1351490 1287641 2639131
1 0. 0 17280.00 25000.00 14179.46 15 60 1384600 11 69331 2553931
11.0 17710.00 25430.00 13144.43 15 57 1417710 1088500 2506210
1 2 .0 18140.00 25860.00 11869.83 15 53 1450820 987847 2438667
1 3 .0 18570.00 26290.00 10893.80 15 51 1483930 913945 2397875
1 4 .0 19000.00 26720.00 10106.65 15 48 1517040 849980 2367020
1 5 .0 19430.00 27150.00 9451.82 IS 46 1550150 798216 2348367
1 6 .0 19860.00 27580.00 9579.03 15 44 1583260 801046 2384307
1 7 .0 20290.00 28010.00 9598.71 15 42 1616370 796092 2412462
1 8 .0 20720.00 28440.00 8958.06 15 40 1649480 745524 2395005
1 9 .0 21150.00 28870.00 8433.31 15 39 1682590 706047 2388637
2 0 .0 21680.00 29400.00 7892.81 15 37 1723400 662324 2385724
21 .0 22240.00 29960.00 7648.06 15 35 1766520 639165 2405685
2 2 .0 22800.00 30520.00 7144.90 15 34 1809640 601094 2410734
2 3 .0 23360.00 31080.00 6733.40 15 33 1852760 569319 2422079
2 4 .0 23920.00 31640.00 6387.68 15 31 1895880 539188 2435068
2 5 .0 24480.00 32200.00 6091.60 15 30 1939000 515480 2454480
2 6 .0 25040.00 32760.00 5834.29 15 29 1982120 494483 2476603
2 7 .0 25600.00 33320.00 5608.08 15 28 2025240 475671 2500911
2 8 .0 26160.00 33880.00 5757.96 15 27 2068360 483202 2551562
2 9 .0 26720.00 34440.00 5483.87 15 27 2111480 464032 2575512

Table 5.4 : Supertargeting results for Aromatics Plant (Fixed Pressure Drop).

T o answer this question, w e should also take the other factors, which can affect

the final total cost, into account. They are the wideness o f the com posite curves and

the problem constraints which have significance in network synthesis and refinement.
Chapter 5 Grass m ots Design o f H F N ’s (New P rocedu re) Page 119

Having considered these factors, it is anticipated that the final cost w ill be lower

if the design is initialised using A T min o f 20*C. Although this anticipation is different

from that targeting suggests, it is, however, justifiable. The reason is that the

composite curves are wider at 20*C, and hence the initial network is less constrained.

In such a situation the loops and paths can be exploited m ore efficiently and with a

smaller penalty, when trying to reduce the number o f units. In other words, there is

more potential available fo r redu cing the number o f units.

Therefore we choose the A T mjn = 20*C as the near optimum point and w ill be

using it to demonstrate the next stages o f the procedure.

5.4,2.2 Heat tran sfer coefficients to b e used in N etw o rk Synthesis

As described above, the new targeting procedure not only yields the area and

energy targets for optimum A T min. but also the heat transfer coefficients to be used

in network design.

Table 5.5 shows the stream heat transfer coefficients associated with optimum

A T min. Included are the stream contact areas.

Chapter 5 Grassroots Design o f H F.N's (N ew P rocedu re) Page 120

Stream Exchanger Targeted h Contact Area

Name Side (W/m2 *C) (m2)

HI Shell 1098.2 2016.9

H2 Shell 1051.2 774.2

H3 Shell 1178.9 675.6

H4 Shell 1280.6 2745.2

Hot Utility Shell 1000.0 1680.8

Cl Tube 840.6 1840.4

C2 Tube 738.3 835.1

C3 Tube 1465.0 1559.6

C4 Tube 767,2 665.0

C5 Tube 1246.1 2259.2

Cold Utility Tube 2500.0 733.6

Table 5.5 : Stream film coefficients and contact areas at A T min = 20*C.

(Aromatics Plant)

5.4.3 N etw ork synthesis (Arom atics Plant)

The purpose o f this sub-section is to show the application o f the Pinch Design

Method (P D M ). using the results of the new targeting procedure.

5.4.3.1 P relim in a ry design

H aving obtained the A T„lin = 20*C, as a result o f Energy-Capital trade-off, a

heating requirement o f 21.68 M W and a cooling requirement o f 29.4 M W set the

Chapter 5 Grassroots Design o f IlF .N 's (N ew Procedure) Page 121

preliminary design objectives. W e now commence the design using P D M , keeping

in mind that w e want to achieve M ER (Maximum Energy Recovery) whilst trying to

minimise number o f units. This would result in an initial network having 14 units and

an installed area o f 8388 m2 (Figure 5.5).

It should be noted that the exchanger areas are calculated based on multi-pass

shell-and-tube exchangers, using the heat transfer coefficients obtained in targeting

stage. The total annualized cost o f this network is 2413413 £/yr which is 1.2% higher

than the targeted value.

r— i 007 /Cv ....... / = * ............. " .............. .............. ™ “V

------------------- s ' s J / ......... ................ ^
v S 0400
1— 1*®
l= J -----------------
so
-f i
M Ç
* >
__ l i u
r ü i i ü _________ ........ f
lü J f ......^
04700

r
i' ‘ “ iRr lIi
00000
100
V ,
V ------
" Ve.r * j \ _ J _________________ i— i
uo uno Ò100 4Mi !
- 1“ ____ f \ _______________
10000
•°F n
L——1

y u a ... ^
■ * = < £ ) ----------0= ---------------------------------------------------------- L _ J
s i too »00 «“ O 000

T« m p. ( C ) D u ty (kW ) m C p (kW /C)

Figure 5.5 : Initial design for Aromatics Plant At A T mil = 20*C (Fixed Pressure drop).
C hapter 5 Grassroots Design o f N F S ’s (New Procedu re) Page 122

5.4.3.2 Fin al design

A s it can be seen from Figure 5.5, this initial network features a fe w loops and

paths. Som e o f these loops are simple involving only four units (loops C1-E7-E8-C3

and H I-E 6 -E 7 -H 2 ), and one is more com plex involving six units (C2-E2-E7-E6-E1-

C3). In either case they are to be exploited, using the established techniques described

in sub-section 5.2.3, in order to a ) reduce the number o f units and b ) optimise the

heat load distribution.

mCp

r = n“ r /=\ / \ 10«
--------- v 2/ w
____—n | • , 4H i #o^
1M
l = J------------------------ y ; ; "

M
i i ® ►
1 187»
______ !____ 400
■i

'
¡0
j
- <

- <
-

*
^ •°° ; 100
w — ; | « i nu
14U ____!_________________ A “~ A ____ !_________________
r S TO
]
- <
:
* kJ
1
G

v - 14" 800
1 1MM
JW s' M
! “ fTTl

10T.4 S '- V itu

^ w w i— i
m u MM MM

T r a p . (C ) D u t , (k V ) mCp (kV /C)

Figure 5.6 : Final design for Aromatics Plant A t A T min = 20*C (Fixed Pressure drop).
Chapter 5 G rassroots Design o f M E N 'S (New Procedure) P a ge 123

Having done these refinements, a final network (Figure 5.6) can be produced

having 12 units and an installed area o f 8638 m2. The energy consumption o f this

network is 20.515 M W which is lower than the targeted value (because o f the energy

relaxation during loop breaking), and also it has a total annualized cost o f 2351566

f/yr that is 1.4% lower than the value obtained in targeting stage.

Table 5.6 shows a comparison between the results o f the three stages involved in

this case study.

Stage: targeting Preliminary Final

Results Design Network

Energy (M W ) 21680 21680 20515

Area (m2) 7893 8388 8638

N o o f Units 15 14 12

N o o f Shells 37 41 40

Energy Cost (£/yr) 1723400 1723400 1633722

Capital Cost (£/yr) 662324 690013 717845

Total Cost (£/yr) 2385724 2413413 2351566

Table 5.6 : Results o f the different stages for the new design (Aromatics Plant).
C hapter 5 Grassroots Design o f U E N 's (N e w Procedure) Page 12-4

N ow . before w e finish this section, it is w orthw hile to compare the final results
h
o f the two A T mm*s (15 & 20 C ) and see w eth er our initial anticipation has been

currect.

T o compare the two near optimum points, the whole design procedure was

repeated for A T min o f 15*C. This means an initial network was first designed and then

refined using the established techniques. The results o f the tw o stages together with

the Targeting results are compared for both cases in Table 5.7.

Targeting 19430 9451.8 2348367 21680 7892.8 2385724

19430 10443. 2402563 21680 8388.2 2413413

Refinement 20614 9184.7 2371*28 20515 2351566

Table 5.7 : Comparison between the results fo r A T mjn = 15*C and 20*C.

A s it was anticipated, the network which has been initialised using A T min o f 20*C

features three units less than the network initialised with 15’ C and it is cheaper too.
C hapter 5 Grassroots Design o f U P S 's (S ew Procedu re) Pa ge 125

5.5 Sum m ary

1. A new targeting procedure has been developed, using the new area algorithm

presented in Chapter Four.

2. This new procedure gives a sharper relationship between network area and

energy consumption than that predicted on the basis o f assumed coefficients. This is

because it takes account o f variation in film heat transfer coefficien ts throughout the

trade-off between energy consumption and network area.

3. A s a result o f 2, therefore, the new targeting procedure not on ly yields the near

optimum energy recovery level and network area requirement but also the heat

transfer coefficients to be used in network design.

4. The procedure has been tested using a case study and it has been found that the

results o f targeting and network synthesis are in a good agreement.

C hapter 5 Grassroots Design o f HF.N's ( New P rocedu re) Page 126

Nom enclature

A heat transfer surface based on outside tube area

Cp specific heat

E energy consumption

h heat transfer coefficient

k thermal conductivity

L number o f loops

m mass flowrate

N number o f streams (including utilities)

Nu number o f units

Rf fouling resistance

s number o f independent stream subsets

T, stream supply temperature

T, stream target temperature

U «* minimum number o f units (including heaters and coolers)

(Ujnin^a b o v e minimum number o f units above the Pinch

(UminW.iow minimum number o f units below the Pinch

( UminV\IER minimum number o f units for Maximum Energy Recovery

AP stream pressure drop

A T Y ,* logarithmic mean temperature difference

Q stream density

P stream viscosity
Chapter 5 Grassroots Design o f H F N ’s (N ew Procedu re) Page 127

References

Ahmad S. & Smith R.,

’Targets and Design for Minimum Number o f Shells in Heat Exchanger Networks,
Trans.I.Chem.E., 67, pp. 481-494, 1989.

Hohmann. E.C.,
’ Optimum Network for Heat Exchanger’ ,
Ph.D. Thesis, University o f S.Califomia. 1979.

Jegede F.O.,
’ Power. Capital and Energy Cost Trade-offs in Heat Exchanger Networks’ , Ph.D.
Thesis. U .M .I.S.T., 1990.

Linnhoff B., Mason D.R. and Wardle I.,

’ Understanding Heat Exchanger Networks’ ,
Comp, and Chem.Eng., 3.295. 1979.

Linnhoff B. et al.
’ User Guide on Process Integration for the Efficient Use o f Energy’ ,
IChemE. Rugby. 1982.

Linnhoff B. «Sc Ahmad S.,

’ Optimal Synthesis o f Energy Management Systems’ ,
T A S M E (J.Eng.Res.Tech.), 111(3), pp. 121-130, 1989.

TEM A.
’Standards o f Tubular Exchangers Manufacturers Association (6th edn.)’ ,
N ew Y o rk . 1978.

TABLES.
’Computer Program Library fo r Process Engineering Calculations’
Imperial Chemical Industries. 1987.
C hapter ti Interface n f Network Synthesis & Detailed H .E.D . Page 128

CHAPTER 6

INTERFACE O F

NETWORK SYNTHESIS

AND

DETAILED HEAT EXCHANGER DESIGN

C hapter 6 Interface o f Network Synthesis & D etailed 11.F..D. Page 129

Contents

6.1 Introduction

6.2 Case Study (Aromatics Plant)

6.2.1 Assumed heat transfer coefficient (h=500 W/m2 *C)

6.2.2 Fixed pressure drop

6.2.3 Assumed heat transfer coefficient (h=1000 W/m2 *C)

6.3 N ew Procedure and Optimal Network Synthesis

6.3.1 Vertical alignment and tick-off heuristics

6.3.2 Case study (Four stream problems)

6.4 Summary

References
Chapter 6 Interface o f Network Synthesis & Detailed H.F..D. Page 130

6.1 Introduction

Current procedures for targeting and design o f heat exchanger networks involve

the use o f assumed heat transfer coefficients, however, during detailed heat exchanger

design allowable pressure drops are often the most critical factors. This inconsistency

can result in a big difference between the exchanger sizes and costs anticipated by

the network designer and those realised by the exchanger designer [Polley, Panjeh

Shahi. 1991]. T h e hazards o f this inconsistency were mentioned in Section 1.2.

In Chapter 5. we presented new heat exchanger targeting and design procedures

which were based on stream pressure drops rather than stream coefficients. This

means, having used these procedures, stream pressure drops can be treated objectively

from the very first stage (network targeting) to the last (detailed exchanger design),

in a consistent manner.

In this chapter, the Aromatics case study is revisited using both fixed coefficient

(conventional method) and fixed pressure drop (new method) procedures and the

results o f the two methods are compared. Also, detailed exchanger design is

undertaken to show how consistency between expectation and realisation can only be

achieved if the new procedure is applied.

C hapter 6 Interface o f Network Synthesis & detailed H.F..D.

6.2 C ase Study (Arom atics Plant)

T h e case study which is discussed here is the Aromatics plant used in the previous

chapter to demonstrate how new procedure could be used in the design o f near

optimal networks.

Lin n h off and Ahmad assumed that all o f the streams involved exhibited a service

film heat transfer coefficient o f 500 W/m2 *C. These values are low er than it would

be expected to be seen in practice. H owever, they w ill be re-em ployed here with the

exception o f the coefficients o f the utility streams. These w ill be set at 1000 W/m2*C

for the hot oil stream and 2500 W/m2 *C for cooling water.

T h e stream data, physical properties, fouling resistances, stream pressure drops

and cost data for this problem are those presented in Tables 5.1, 5.2 and 5.3.

N o w . before applying the new procedure to the problem w e w ill look at a solution

resulting from the use o f assumed heat transfer coefficients.

C hapter 6 Interface o f Network Synthesis & Detailed //.F . D. P a ge 132

6.2.1 Assum ed heat tran sfer coefficients o f 500 W/m2 *C

A . T a rg e tin g

T h e first stage o f designing a heat exchanger network is the trade-off between

energy and network capital casts. Table 6.1 presents the result o f this trade-off over

a w ide range o f A T min.

STREAMS - AROM.STM U T IL IT IE S - AROM.UTY !

New D esign Range T arget (As sumed C o e f f i c i e n t )

DTmi n Hot U t. C old U t. 1-2 Area Nu Ns U t. C ost Cap.Cost T o t .C o s t

re j [kW] IkW) (M2) (C / y r] (E / yr) te / y r )

5 .0 15130.00 22850.00 41735.27 15 141 1219050 3312522 4531572

6 .0 15560.00 23280.00 36702.62 15 124 1252160 2913092 4165252
7 .0 15990.00 23710.00 32743.75 15 111 1285270 2600372 3885642
8 .0 16420.00 24140.00 30071.45 15 102 1318380 2388384 3706764
9 .0 16850.00 24570.00 28635.82 15 97 1351490 2273841 3625331
10.0 17280.00 25000.00 26449.52 15 90 1384600 2101851 3486451
1 1 .0 17710.00 25430.00 24843.82 15 84 1417710 1972117 3389827
12.0 18140.00 25860.00 22847.15 15 78 1450820 1816605 3267426
13.0 18570.00 26290.00 21281.57 15 72 1483930 1689521 3173451
14.0 19000.00 26720.00 19993.75 15 68 1517040 1588703 3105743
15.0 19430.00 27150.00 18903.70 15 65 1550150 1504884 3055035
16.0 19860.00 27580.00 19167.31 15 65 1583260 1522282 3105542 |
17.0 20290.00 28010.00 19092.12 15 65 1616370 1517324 3133694
18.0 20720.00 28440.00 18009.77 15 62 1649480 1434010 3083490
19.0 21150.00 28870.00 17109.30 15 59 1682590 1362704 3045294
20.0 21680.00 29400.00 16167.99 15 55 1723400 1284754 3008154
21 .0 22240.00 29960.00 15697.11 15 54 1766520 1249716 3016236
22.0 22800.00 30520.00 14809.20 15 51 1809640 1179240 2988880
2 3.0 23360.00 31080.00 14069.42 15 48 1852760 1118547 2971307
2 4.0 23920.00 31640.00 13437.79 15 46 1895880 1068944 2964824
25.0 24480.00 32200.00 12888.97 15 44 1939000 1024807 2963807
2 6.0 25040.00 32760.00 12405.71 15 43 1982120 988939 2971060
2 7 .0 25600.00 33320.00 11975.68 15 41 2025240 952657 2977897
2 8.0 26160.00 33880.00 12299.62 15 42 2068360 977994 3046355
2 9.0 26720.00 34440.00 11765.83 15 41 2111480 938780 3050260

Table 6.1 : Supertargeting results for Aromatics Plant, (Assumed-h, 500 W/m2 *C).
C hapter 6 Interface o f Network SynthesLi & Detailed H .E.D. P a ge 133

These results suggest that the near optimal energy recovery level will be achieved

(givin g a minimum total annualized cost o f 2963807 £/yr) if A T min is set at 25*C. The

energy consumption and area requirement corresponding to this temperature

difference are 24480 k W and 12889 m2, respectively.

B. N etw o rk Synthesis

Having set the energy recovery level, an initial network was produced using the

design procedure introduced in Chapter 5. The result was a network having 13 units.

The loops and paths present in this structure were then exploited to yield the final

network shown in Figure 6.1.

The final network features 11 units (two units less than the initial network) having

a total surface area o f 14569 m2. The energy consumption for this network is 22400

k W and the total annualized cost is 2950601 £/yr.

In the first look the design seems to be quite good and comparable with the

targeting results. However, at this stage we are just half way home. W e have not yet

accomplished the detailed exchanger design.

C h a p ter 6 Interface o f Network Synthesis & Detailed II.E .D . Page 13d

mCp
______ C v i w 100
Y s y ^
* * °fc 190
l_ l
y \
IU
r\ /S S “ fc. 90
V i 1499
: s «» 400
f i n 1* 0 ! ^
£11 î
99119
1

( -s ; :
1 100r[C«J
^n 100
90000
141.9 109.9 /- ! *9 r - ^
70
IMS
V *" ò “ LfZJ
îô t i 980
w

^ ____ ___________________________________________________________ îü L r ^ i 90
LrzJ

900
^ w i___ i
99400 9000
Temp. (C) Duty (kV ) mCp ÇkW/C)

Figure 6.1 : Final network for Aromatics Plant. (Assumed Coefficient, 500 W/m2 *C).

C. D eta iled Design

N o w . I^et’s m ove on to the detailed design o f the individual heat exchangers. The

designs were undertaken using the full Bell-Delaware method and the n ew procedure

presented in Chapter 3. Details o f the individual exchanger designs are given in

T ables 6.2a and 6.2b.

Chapter 6 In te rfa c e r tf Network Synthesis & Detailed H .E.D . P a ge 135

Exchanger El E2 E3 E4 E5 E6

Expected Sizeim 2) 118.3 1294.5 1536.0 3223.4 316.5 813.0

Actual Size (n r) 63.3 699.5 765.0 1869.3 164.1 459.2

No. o f Shells 1 5 3 8 1 3

Shell Diam. (mm) 828.0 809.7 1215.0 1034.7 1160.2 835.2

No. o f Tubes 685 653 1498 1105 1354 698

Tube Passes 2 2 4 2 4 2

Tube O.D. (mm) 19.1 19.1 19.1 19.1 19.1 19.1

Tube I.D. (mm) 15.4 15.4 15.4 15.4 15.4 15.4

Tube Pitch (mm) 25.4 25.4 25.4 25.4 25.4 25.4

Tube Length (mm) 1546 3590 2847 3537 2026 3669

Baffle Cut (% ) 22.5 22.6 36.7 27.0 32.9 38.9

No. o f Baffles 5 13 3 8 2 6

h * (W/m2 " O 780.0 759.7 1318 839.2 785.7 770.1

h„ (W/m2 'C l 1166 1186 898.4 1092 1331 1016

U (W/m2 -C) 467.4 463.1 534.2 474.5 494.1 438.1

AP, (k Pa) 1.90 15.00 35.31 20.00 5.07 13.03

AP. (k Pa) 7.18 78.62 80.00 90.19 3.57 22.75

Table 6.2a : Results o f the Detailed Exchanger Design for the Aromatics plant,
(Assumed C oefficien t, 500 W/m2 *C).
Chapter 6 Interface n f Network Synthesis & Detailed H.F..D. P a ge 136

Exchanger E7 11 Cl C2 C3

Expected Size(m 2) 3872.1 1943.9 1130.3 69.0 252.5

Actual Size (m2) 1378.4 1200.4 531.9 42.8 144.5

No. o f Shells 1 5 2 1 1

Shell Diam. (mm) 1506.6 893.1 1205.6 564.1 842.3

No. o f Tubes 2487 809 1536 266 650

Tube Passes i 2 2 6 6

Tube O.D. (mm) 19.1 19.1 19.1 19.1 19.1

Tube T.D. (mm) 15.4 15.4 15.4 15.4 15.4

Tube Pitch (mm) 25.4 25.4 25.4 25.4 25.4

Tube len gth (mm) 9270 4969 2897 2690 3720

Baffle Cut (% ) 17.4 19.8 29.9 43.6 38.2

No. o f Baffles 27 20 4 5 6

(W/m2 "C> 1402.3 1261.6 2505.2 2506.1 2506.8

h„ (W/m2 ‘ Cl 1388.7 1001.0 1408.3 919.4 1015.8

U (W/m2 *C) 697.8 558.1 901.5 672.6 722.9

AP, (k Pa) 30.00 44.69 38.10 53.20 73.6

AP, (k Pa) 43.68 152.5 12.75 4.19 7.07

Table 6.2b : Results o f the Detailed Exchanger Design for the Aromatics Plant,
(Assumed Coefficient, 500 W /m 2 *C).
Chapter ft Interface o f Network Synthesis & Detailed H.F..D. Page 137

As can be observed, the detailed exchanger design results in actual exchanger

sizes that are very different from those expected during the n etw ork design. The

overall heat transfer coefficients are by no means uniform, and th e film coefficients

assumed during the network synthesis were too conservative.

In Table 6.3 we show the energy recovery levels and network a rea needs together

with the calculated costs at each stage o f the procedure.

Stages Energy Network Energy A rea Total

o f the need Area Cost C o st Cost

solution (kW ) (m2) (£/yr) (f/ y r ) (£/yr)

Targeting 24480 12889 1939000 1024807 2963807

Network Design 22400 14569 1778840 1171761 2950601

Detailed Design 22400 7318 1778840 585121 2363961

Table 6.3 : Results o f the three stages for the Aromatics Plant,

(Assumed Coefficient, 500 W/m2 *C).

In the context o f this chapter it is the network area requirement that is o f

particular importance. With the assumed coefficients the targeting procedure predicts

a need o f 12889 m2. After network synthesis the expected area n eed is 14569 m2.

Follow in g detailed exchanger design the true network area is fou n d to be just 7318

m2. W e have a difference between the network synthesis area and the final design

value o f around 50%.

Chapter 6 Interface o f Network Synthesis & Detailed H .E.D . Page 138

N o w . let's exam ine the new procedure. In next sub-section, we apply the new

procedure to the problem and see whether this inconsistency is repeated.

6.2.2 Fixed Pressu re Drop

T h e first two stages o f the solution (i.e. targeting and network synthesis) using the

new procedure have been demonstrated in previous chapter.

A. T a rg e tin g

In targeting stage the energy/network capital trade-off was examined and the near

optimal energy recovery level established at A T min o f 20*C (Table 5.5). W e know that

the targeting stage, in new procedure, not only gives us a A T niin to initialise the design

with, but also returns a set o f stream coefficients to be used in network synthesis.

These coefficients which were different from the assumed values have been given in

Table 5.6. Having compared the two procedures, it is appreciated that this is one

significant difference in the information carried forward to the design stage.

Another important difference is the A T min itself. Here, The A Tmin to be used for

initial design is 20*C rather than 25*C. This particular change in A Tmin does not affect

the Pinch location and stream population around the Pinch (Figure 6.2) and w e can

expect the two solutions to the problem to have similar structures. H owever, in other

temperature ranges the change in A T miB can result in significant changes in final

network structure (S ee next sub-section).

C hapter 6 Interface o f Network Synthesis & Detailed II.E .D . Pa ge 139

B. N etw o rk Synthesis

In network design stage, an initial network was first produced and then refined to

give the near optimal structure. The resulting network has been illustrated in Figure

5.6.

C. Detailed Design

W e now undertake the detailed design o f the individual heat exchangers. The

results o f these designs are given in Tables 6.4a and 6.4b.

C ha pter 6 Interface <>f Network Synthesis & D etailed H .F.D. Page 140

Exchanger El E2 E3 E4 E5 E6

Expected Sizeim 2) 59.6 623.7 1159 1861 244.9 240.1

Actual Size (m 2) 59.2 607.1 1370 1873 238.3 231.7

N o. o f Shells 1 5 5 8 1 1

Shell Diam. (mm) 849.1 791.1 1348 1035 1305 1190

N o. o f Tubes 724 620 1856 1105 1732 1432

Tube Passes 2 2 4 2 4 4

Tube O.D. (mm) 19.1 19.1 19.1 19.1 19.1 19.1

Tu be I.D. (mm) 15.4 15.4 15.4 15.4 15.4 15.4

Tube Pitch (m m ) 25.4 25.4 25.4 25.4 25.4 25.4

Tube Length (mm) 1365 3272 2462 3535 2295 2697

B a ffle Cut (% ) 21.0 22.6 36.6 27.1 27.7 31.6

N o. o f Baffles 4 12 2 7 3 3

h „ (W/m2 -C) 749.9 784.0 1161 837.2 1209 755.2

h„ (W/m2 -C ) 1176 1209 812.6 1091 1391 1349

U (W/m2 " O 457.8 475.6 478.0 473.7 646.8 484.1

AP, (k Pa) 1.51 15.00 34.58 20.00 7.31 6.11

A P , (k Pa) 7.43 77.80 80.00 89.90 4.91 4.81

Table 6.4a : Results o f the Detailed Exchanger Design for the Arom atics Plant,
(Fixed AP).
C h a p ter 6 Interface o f Network Synthesis & D etailed H.F..D. P a g e 141

Exchanger E7 E8 H Cl C2 C3

Expected Size(m 2) 486.9 1970 1278 539.6 38.2 136.4

Actual Size (m2) 493.6 1902 1296 527.4 42.2 146.9

N o. o f Shells 3 1 5 2 1 1

Shell Diam. (mm) 851.8 1591 927.8 1161 589.6 835.5

N o. o f Tubes 729 2787 879 1415 288 638

Tube Passes 2 1 2 2 6 6

Tube O.D. (mm) 19.1 19.1 19.1 19.1 19.1 19.1

Tu be I.D. (mm) 15.4 15.4 15.4 15.4 15.4 15.4

Tu be Pitch (mm) 25.4 25.4 25.4 25.4 25.4 25.4

Tu be Length (mm) 3767 11380 4924 3109 2449 3841

B a ffle Cut <%) 38.1 18.0 16.4 33.0 39.2 39.5

N o. o f Baffles 6 30 24 4 5 6

h * (W/m2 "Cl 746.3 1404 1199 2500 2500 2500

h„ (W/m2 'C l 1025 1314 1000 1334 1022 978.7

U (W/m2 'C l 431.8 678.6 545.2 870.0 725.6 703.4

AP, (k Pa) 12.38 30.00 38.12 41.00 48.40 76.00

A P , (k Pa) 23.51 39.47 158.0 10.81 4.76 6.59

T a b le 6.4b : Results o f the Detailed Exchanger Design fo r the Aromatics Plant,

(Fixed API.
C hapter 6 Interface o f Network Synthesis & Detailed H.F..D. Page 142

Let us look at the predictions o f network energy consumption and area

requirement made in each o f the stages o f this solution. Th ey are summarised

together with the related costs in Table 6.5.

Stages Energy Network Energy Area Total

o f the need Area Cost Cost Cost

solution (kW ) (m2) <£/yr) (£/yr) (£/yr)

Targeting 21680 7893 1723400 662335 2385735

Network Design 20515 8638 1633722 717845 2351566

Detailed Design 20515 8787 1633722 684893 2318615

Table 6.5 : Results o f the three stages for the Aromatics Plant, (Fixed AP).

A t the targeting stage w e predict a network area need o f 7893 m2. Following

network synthesis w e expect the area requirement to be 8638 m2, and the final

installed value is found to be 8787 m2. With the new procedure the difference in area

predicted by synthesis and that actually found in detailed design is just 1.7%.

The above comparisons indicate that the heat transfer coefficients originally

specified by Linnhoff and Ahmad were poor given typical pressure drops and

physical properties. The question that now arises is ’ how do the tw o methods

compare if the coefficients had been better guessed?’ . This question is addressed in

next sub-section.
Chapter 6 Interface o f Network Synthesis & Detailed //.ED. Page 143

6.2.3 Assumed heat transfer coefficients o f 1000 W /nr *C

W e see from tables 6.4a and 6.4b that the film heat transfer coefficients can be

expected to range from 750 to 1250 W/m2 *C. A good guess would appear to be 1000

W/m2 *C for each stream. Therefore we repeat the case study using assumed

coefficients o f 1000 W/m2 *C.

A . T a rgetin g

Table 6.6 shows the results o f the Supertargeting stage. For this case, the near

optimal energy recovery level happens at A T niin o f 15*C giving a total annualized cost

o f 2384334 £/yr. The energy consumption and area requirement related to this

temperature difference are 19430 kW and 9967 m2, respectively.

B. N etw ork Synthesis

In network design stage, an initial network was first produced fo r maximum

energy recovery determined by supertargeting. T h e result was a network having 15

units. This network was then refined to obtain the cheapest structure. T h e resulting

network is illustrated in Figure 6.3. It is immediately observed that this final network

has 3 units more than the network achieved by fixed pressure drop solution. This is

because the A Tniin used for initial design was 15*C rather than 20*C. A s it was

anticipated above, within this temperature range, this change in A T min did affect the
Chapter 6 Interface o f Network Synthesis & D etailed H .F .D . Page 144

stream population around the pinch and the Pinch location itself (Figure 6.4).

Therefore w e coulc expect the two final structures to be indicatively different.

STREAMS - AROMI . STM U T IL IT IE S - AROM.UTY

New D esign Range T arget (Assumed C o e f f i c i e n t )

DTmin Hot Ut. C old Ut. 1-2 Area Nu Ns U t. C ost C ap.C ost T o t .C o s t
rei tkW) [VW] [M2] Ce/yr] [E / yr] [E /yr]

5 .0 1 5130.00 22850.00 21423.25 15 94 1219050 1777627 2996677 i

6 .0 15560.00 23280.00 18951.57 15 84 1252160 1575241 2827401
7 .0 15990.00 23710.00 16952.19 15 76 1285270 1411789 2697059
8 .0 16420.00 24140.00 15600.71 15 70 1318380 1299423 2617803
9 .0 16850.00 24570.00 14870.81 15 65 1351490 1233125 2584615
10.0 17280.00 25000.00 13767.96 15 60 1384600 1141094 2525694
11 .0 17710.00 25430.00 12957.23 15 57 1417710 1075617 2493328
12.0 18140.00 25860.00 11952.43 15 53 1450820 993550 2444370
13.0 18570.00 26290.00 11164.31 15 51 1483930 932742 2416672
14.0 19000.00 26720.00 10515.98 15 48 1517040 878456 2395496
15.0 19430.00 27150.00 9967.29 15 46 1550150 834184 2384334
16.0 19860.00 27580.00 10096.04 15 44 1583260 836770 2420031
17.0 20290.00 28010.00 10120.54 15 42 1616370 831851 2448221
18.0 20720.00 28440.00 9572.57 15 40 1649480 787731 2437211
19.0 21150.00 28870.00 9116.62 15 39 1682590 753211 2435801
2 0.0 21680.00 29400.00 8640.10 15 37 1723400 713968 2437368
21 .0 22240.00 29960.00 8399.63 15 35 1766520 690878 2457398
22.0 22800.00 30520.00 7951.60 15 34 1809640 656905 2466545
23.0 23360.00 31080.00 7578.41 15 33 1852760 628017 2480777
24.0 23920.00 31640.00 7259.94 15 31 1895880 599625 2495505
25.0 24480.00 32200.00 6983.41 15 30 1939000 577378 2516378
2 6.0 25040.00 32760.00 6740.10 15 29 1982120 557410 2539530
2 7.0 25600.00 33320.00 6523.80 15 28 2025240 539295 2564535
2 8.0 26160.00 33880.00 6684.80 15 27 2068360 546927 2615287
2 9.0 26720.00 34440.00 6417.22 15 27 2111480 528693 2640173

Table 6.6 : Supertargeting results for Aromatics Plant, (Assumed-h, 1000 W/m2 *C).
 Interface o f N etw ork Synthesis & Detailed H.F.D.

0= -ty - -<=>— ^ — (= )-

H = -------- ------------------" V
•14
----- v:-y -^ r ¡ N fa “ *►
v2/
r.

1 • 10#f^n
< ---(«y-----
JM l«4/J>»07^,S_1
__________*°J7n
/■N
IW 7
O “ ra
.W, L-J
»70
< ------------ -------------------------------
k a -
10M7 06*6-2 078.7 B600 11M.1
=a
Temp. (C) Duty (kW) mCp (kW/C)

Figure 6.3 : Final network for Aromatics Plant, (Assumed C oeff., 1000 W/m2 *C).

100 mCp
40^
m 9X7 10O
r_, 220 >» 100
00 __ 00
r=-i “ °
1 3 1
160 40 _ 400

300 100 ro-| 100

4 1* 4 “ m 70

----------------------------? i { T ] 300
170 00
•°n n
300

b
146
Tempo. (C) mCp (kW/C)

Figure 6.4 : Design Grid for A T min = 15*C (Aromatics Plant).

Chapter 6 Interface o f Network Synthesis & Detailed H .F .D . P a ge 146

C. D etailed Design

Finally, the detailed design o f the individual heat exchangers was carried out. The

detail o f these designs are demonstrated in tables 6.7a, 6.7b and 6.7c.

Again, we look at the predictions o f network energy consumption, area

requirement, and the costs made in each o f the stages o f this solution. They are

summarised in Table 6.8.

W e find that after targeting an area network o f 9967 m2 is expected. Following

synthesis this has fallen to 9728 m2. The final network area is found to be 8895 m2.

It is seen that even with a good £uess the difference between the synthesis and

detailed design is 9%.

Chapter 6 Interface o f Network Synthesis & Detailed H.F..D. Pa ge 1-47

Exchanger El E2 E3 E4 E5

Expected Sizeim 2) 377.3 74.1 48.3 707.1 1062.8

Actual Size (m 2) 317.9 62.2 56.1 782.7 1044.8

No. o f Shells 2 1 1 6 4

Shell Diam. (m m ) 905.5 905.1 883.2 824.1 1315.4

No. o f Tubes 833 832 789 679 1761

Tube Passes 2 2 2 2 4

Tube O.D. (m m ) 19.1 19.1 19.1 19.1 19.1

Tube I.D. (mm) 15.4 15.4 15.4 15.4 15.4

Tube Pitch (m m ) 25.4 25.4 25.4 25.4 25.4

Tube Length (mm) 3184.1 1246.5 1186.0 3212.3 2474.1

Baffle Cut (% ) 31.2 19.1 22.1 23.2 35.3

No. o f Baffles 6 4 3 11 2

hK) (W/m2 ' O 1236.6 1237.7 708.0 738.0 1197.6

hG (W/m2 ‘ C> 1126.6 1156.6 1109.5 1150.9 870.8

U (W/m2 'C l 589.5 597.9 432.2 449.7 504.2

AP, (k Pa) 10.851 2.131 1.128 15.000 30.564

AP, (k Pa) 20.761 7.699 5.018 73.469 80.000

Table 6.7a : Results o f the Detailed Exchanger Design fo r the Aromatics Plant,
(Assumed C oefficient, 1000 W/m2 *C).
C hapter 6 Interface o f Network Synthesis & Detailed II.E .D . Pa ge 148

Exchanger E6 E7 E8 E9 E10

Expected Si7.e(m2) 1155.3 146.9 295.2 513.2 2540.1

Actual Size (m2) 1230.1 116.2 337.6 606.1 1832.7

No. o f Shells 6 1 2 3 4

Shell Diam. (mm) 1059.7 1096.1 1077.5 885.6 1580.6

No. o f Tubes 1164 1222 1177 794 2751

Tube Passes 2 3 3 2 1

Tube O.D. (mm) 19.1 19.1 19.1 19.1 19.1

Tube T.D. (mm) 15.4 15.4 15.4 15.4 15.4

Tube Pitch (mm) 25.4 25.4 25.4 25.4 25.4

Tube Length (mm) 2939.6 1586.9 2392.6 4247.7 2778.9

B affle Cut (% ) 25.6 36.7 40.9 36.0 18.9

No. o f Baffles 6 1 2 7 6

hK) (W/m2 *C) 809.3 1252.8 710.6 705.1 1415.2

h„ (W/m2 ' O 1104.5 1255.1 1119.6 1057.4 1299.8

U (W/m2 X ) 467.1 627.0 434.7 423.0 677.5

AP, (k Pa) 11.362 4.225 6.892 11.980 30.000

AP. (k Pa) 63.571 2.436 4.895 28.239 42.116

Table 6.7b : Results o f the Detailed Exchanger Design for the Aromatics Plant,
(Assumed Coefficient, 1000 W/m2 *C).
C h a p te r 6 Interface o f Network Synthesis <ft Detailed H.F.D. Page 149

Exchanger HI H2 Cl C2 C3

Expected Size(m 2) 878.4 1120.6 135.0 36.7 636.5

Actual Size (m 2) 795.1 1009.8 135.0 39.2 530.0

N o . o f Shells 8 4 1 1 3

Shell Diam. (m m) 736.9 915.5 791.5 562.8 1165.8

N o . o f Tubes 549 854 564 265 1429

T u b e Passes 2 6 6 2

T u b e O.D. (mm) 19.1 19.1 19.1 19.1 19.1

T u b e I.D. (mm) 15.4 15.4 15.4 15.4 15.4

T u b e Pitch (mm) 25.4 25.4 25.4 25.4 25.4

T u b e Length (m m) 3021.4 4937.4 3991.9 2470.4 2062.7

B a ffle Cut (% ) 15.8 19.4 40.7 44.1 33.6

N o . o f Baffles 19 19 6 4 2

h „ (W/m2 -C) 840.5 1219.6 2500.0 2401.6 2500.0

hQ (W/m2 ’ C) 992.8 1000.0 988.8 930.6 1308.2

U (W/m2 'C l 455.2 549.5 708.6 670.7 858.8

A P , (k Pa) 8.638 32.229 79.000 40.000 40.800

A P , (k Pa) 80.000 46.000 7.429 3.814 10.553

T a b le 6.7c : Results o f the Detailed Exchanger Design for the A rom atics Plant,
(Assumed Coefficient, 1000 W/m2 *C).
C hapter ft in terfa ce o f Network Synthesis & Detailed l/.F.D. Page 150

Stages Energy Network Energy Area Total

o f the need Area Cost Cost Cost

solution (k W ) (m 2) (£/yr> (£/yr) (£/yr)

Targeting 19430 9967 1550150 834184 2384334

Network Design 20060 9728 1598625 815641 2414266

Detailed Design 20060 8895 1598625 750951 2349576

Table 6.8 : Results o f the three stages for the Aromatics Plant,

(Assum ed Coefficient, 1000 W/m2 *C).

6.3 New Procedure and Optimal Network Synthesis

6.3.1 Vertical alignm ent and tick-off heuristics

Network design using Pinch Technology proceeds in stages. The philosophy is to

fix the energy consumption, thereby achieving the energy target, and then to approach

the capital cost target b y minimising the number of units (using the so called ’tick-off

heuristic’ [Lin n h off et al, 1982]) whilst choosing matches which exhibit good vertical

alignment between the composite curves.

In some cases the use o f minimum number o f units can result in designs having

significant non-vertical alignment o f matches. Poor verticality results in the true needs

being higher than the predicted needs. This is easily understood. Poor alignment
C hapter 6 Interface o f Network Synthesis & Detailed H .F.D. Page 151

between the composite curves results in reduced efficiency in the use o f temperature

driving force and an area penalty. H ow ever, when such designs are evaluated on the

basis o f assumed coefficients, this penalty need not necessarily appear large. But this

is to ignore pressure drop effects.

In chapter 2, it has been shown that stream pressure drop, exchanger area and

actual film coefficient are related by an expression o f the form:

AP = k A h " (6 .1 )

where k and m are constant values.

A consideration o f this equation shows that any area penalty resulting from poor

alignment actually, given fixed pressure drop, results in reduced heat transfer

coefficients and further area penalty. So. p o o r alignment has a ’double penalty*. Can

this have an effect on the accuracy o f our synthesis predictions?

6.3.2 Case study (Four stream problem )

The four stream problem presented by Sagli et al [1990] is helpful in

demonstrating the possible consequences o f the 'double penalty' caused by the poor

alignment. Again, the problem is solved in three stages using the information given

in Tables 1.1 and 1.2.

C ha pter 6 Interface a f Network Synthesis & Detailed H .E .D . Page 152

A . T a rg e tin g

Having applied the new targeting procedure to this problem , a A T min o f 13*C is

obtained that reflects the minimum total annualized cost o f 105663 f/yr. The energy

consumption, surface area requirement, and the m inim um number o f units

corresponding to this temperature difference are 675 k W , 483.9 m2 and 4 units,

respectively.

B. Network Synthesis

A t this A T mjn the problem becomes a ’ threshold problem ’ . In this situation, if we

aim at achieving the minimum number o f units, a four unit design that exhibits

significant non-vertical alignment is produced (Figure 6.5).

Figure 6.5 : Network for Four stream problem, (Designed fo r min. number o f units).
C hapter ft Interface o f Network Synthesis & Detailed II. E D . Pape 153

however, if during network design w e concentrate on obtaining good vertical

alignment, rather than on using the ’ tick-off heuristic’ to obtain the least number o f

matches, a design having seven units and exhibiting quite good vertical alignment can

be derived (Figure 6.6).

C. Detailed Design

In this stage we undertake the detailed design of the individual heat exchangers

o f the two networks. Comparisons o f the predicted areas during network synthesis

with the actual areas known after detailed design are given in Tables 6.9 and 6.10.
C hapter 6 Interface o f Network Synthesis & Detailed I/.E.D. Page 154

Exchanger El E2 E3 H

Expected Size(m2) 394.4 124.6 85.1 12.5

Actual Size (m2) 465.2 114.0 92.8 19.5

No. o f Shells 8 2 1 1

Shell Diam. (mm) 551.9 597.3 775.0

No. o f Tubes 286 297 556

Tube Passes 2 6 4

Tube O.D. (mm) 19.1 19.1 19.1

Tube I.D. (mm) 15.4 15.4 15.4

Tube Pitch (mm) 25.4 25.4 25.4

Tube Length (mm) 3407 3211 2793

B affle Cut (% ) 24.8 32.9 22.2

N o. o f Baffles 16 9 10

h * (W/m2 ‘ Cl 492.7 1122.1 502.6

h„ (W/m2 ‘ O 757.0 1203.3 1234.1

U (W/m2 ‘ O 298.5 580.6 357.2

AP, (k Pa) 8.22 54.53 1.78

A P . (k Pa) 20.00 17.82 12.18

Table 6.9 : Results o f the Detailed Exchanger Design for Four stream problem.
(4-unit network).
Chapter 6 Interface o f Network Synthesis & Detailed H.F..D. P a ge 155

Exchanger El E2 E3 E4 E5 E6 H

Expected Size(m 2) 56.3 62.9 37.3 33.0 140.0 131.0 23.6

Actual Size (n r) 49.6 60.4 41.3 35.8 139.8 129.9 23.4

No. o f Shells 2 2 2 2 2 2 1

Shell Diam. (mm) 346.7 382.4 259.6 290.8 443.3 488.4

No. o f Tubes 103 120 46 51 171 195

Tube Passes 1 2 2 4 2 4

Tube O.D. (mm) 19.1 19.1 19.1 19.1 19.1 19.1

Tube I D. (m m) 15.4 15.4 15.4 15.4 15.4 15.4

Tube Pitch (m m) 25.4 25.4 25.4 25.4 25.4 25.4

Tube Length (mm) 3998 4171 7616 5838 6791 5549

B affle Cut (% ) 21.0 20.2 39.3 36.1 31.9 29.1

No. o f B affles 37 36 43 31 29 23

h * (W/m2 -C ) 619.6 994.7 579.7 968.0 602.2 987.5

h„ (W/m2 *C> 943.7 939.2 706.5 770.1 1124 1191

U (W/m2 'C * 374.1 483.1 318.4 428.9 392.2 539.9

AP, (k Pa) 2.19 16.63 6.89 43.37 6.89 43.37

AP, (k Pa) 12.03 13.12 7.97 6.88 30.00 30.00

Table 6.10 : Results o f the Detailed Exchanger Design fo r Four stream problem,
(7-unit network).
Chapter 6 Interface o f Network Synthesis & D eta iled 11.F..D. Page 156

N ow . let us look at the predictions o f network area requirement made in each o f

the stages o f the tw o solutions. They are summarised together with the other

associated information in Tables 6.11 and 6.12.

Stages Energy Network Energy Area Total

o f the need Area Cost Cost Cost
solution (kW ) (m2) (£/yr) <£/yr) (£/yr)

Targeting 675 483.9 47250 58413 105663

Network Design 675 616.6 47250 72852 120102
Detailed Design 675 691.5 47250 79265 126515

Table 6.11 : Results o f the three stages for the Four stream problem,
(4-unit network).

Stages Energy Network Energy Area Total

o f the need Area Cost Cost Cost
solution (kW ) (m2) (£/yr) (£/yr) (£/yr)

Targeting 675 483.9 47250 58413 105663

Network Design 675 484.1 47250 69170 116420
Detailed Design 675 480.2 47250 68862 116112

Table 6.12 : Results o f the three stages for the Four stream problem,
(7-unit network).
Chapter 6 Interface o f N etw ork Synthesis & D etailed H .E.D. Pa ge 157

W e observe that when we have good verticality (the seven unit design) w e get

good agreement between the tw o areas (predicted 484.1 m2. actually need 480.2 m2,

0.8% difference). W e further observe that, as expected, the poor vertical alignment

o f the four unit design results in a much higher prediction fo r the area requirement

o f the network (616.6 m2 as opposed to the 484.1 m2 o f the seven unit design).

H ow ever, w e also see that it results in the true area needs o f the design being higher

than the predicted needs (691.5 m2 as opposed to 616.6 m2, a 12.1% difference).

Finally, when the tw o designs are compared in terms o f total annualized cost, it is

found that the seven unit design is infact about 8.2% cheaper than the four unit

design!

The Pinch Design Method TLinnhoff et al. 1982] introduced the ’tick-off heuristic’

in order to drive the network synthesis towards a design having a minimum number

o f units. H owever, this case study shows that when handling problems that have

significant scope fo r poor vertical ity for stream matches, and when exchanger capital

costs (table 1.1) are realistically given by cost equations o f the type used b y^ a gli et

al [1990]. this approach must now be questioned. A much safer procedure would be

to initially derive networks which exhibit good verticality, and then reduce the

number o f units through an optimisation that exploits the loops and paths present in

the network.
C h a p ter 6 Interface o f Network Synthesis & Detailed H.F..D. P a ge 158

5.6 Sum m ary

1. U s e o f assumed film heat transfer coefficients in targeting can lead to wrong

design initialisations and subsequently to non-optimal network designs.

2. Consistency between synthesis and equipment design can only be achieved if

both operations are conducted on the same basis. This means that synthesis must be

based on stream allowable pressure drops and not on assumed film heat transfer

coefficients.

3. U sing the new area algorithm which is based on stream pressure drops, it is

possible to achieve consistency between network synthesis and detailed exchanger

design conducted using the Bell-Delaware method.

4. A n y area penalty resulting from poor alignment actually, given fixed pressure

drop, results in reduced heat transfer coefficients and further area penalty. So, poor

alignm ent has a 'double penalty*.

5. T o avoid 'double penalty’ , tick -off heuristics introduced by Pinch Design

M eth od should be used with caution when handling problems that force poor

vertical ity o f stream matches.

Chapter 6 In te rfa c e o f Network Synthesis & Detailed H .E.D . P a ge 159

References

Linnhoff B. et al,
’ User Guide on Process Integration for the Efficient Use o f Energy’ ,
IChemE, Rugby. 1982.

Linnhoff B. & Ahm ad S.,

’Optimal Synthesis o f Energy Management Systems’ ,
T A S M E (J.E ng.R es.Tech). 111(3), pp. 121-130. 1989.

Polley G .T. & Panjeh Shahi M.H.,

’ Interfacing Heat Exchanger Network Synthesis and Detailed Heat Exchanger
Design’ ,
Trans. I.Chem.E. 69 A , pp.445-457, 1991.

Sagli B., Gundersen T . and Y ee T.,

’Top ology Traps in Evolutionary Strategies for Heat Exchanger Network Synthesis’ ,
Computer Application in Chemical Engineering, ed. H.Bussemaker & P.Iedema,
Elsevier. Amsterdam, 1990.
 PARTS

RETROFIT DESIGN
C hapter 7 R ctm fit n f H E N ’s (Targeting Procedure) Page 160

CHAPTER 7

RETROFIT

OF

HEAT EXCHANGER NETWORKS

(N E W TARGETING P R O C E D U R E )
 Retm/it o f H F N ' s (Targeting Procedu re) Pa ge 161

Contents

7.1 Introduction

7.2 Retrofit Targeting Procedure to Date

7.2.1 Targeting philosophy

7.2.2 Targeting using constant-a approach

7.2.3 Targeting using incremental-a approach

7.2.4 Targeting using area matrix approach

7.3 Problems w ith the Existing Targeting Procedures

7.4 N ew R etrofit Targeting Procedure

7.4.1 Minimum area requirement in retrofit

7.4.2 Capital cost estimation

7.4.3 The procedure

7.5 Summary

Nomenclature

References
C hapter 7 Retrofit o f H E N 's (T argeting Procedure) Page 162

7.1 Introduction

Current heat exchanger network retrofit procedures [T jo e & Linnhoff, 1986] [Ciric

& Floudas, 1988] [Shokoya & Kotjabasakis, 1990] make use o f information on

stream flowrates, heat capacities, and process temperatures. They also usually make

assumptions regarding stream heat transfer coefficients. Generally little, if any,

systematic consideration is made o f the rest o f process. In particular, no consideration

is taken o f heat exchanger pressure drops, available gravitational head, or o f pump

and compressor characteristics.

As mentioned in Chapter 1. the failure to consider the flo w system at an early

stage in the retrofit exercise can have a number of detrimental results. The network

design may not be compatible with the existing flow system and the designer may

opt fo r the unnecessary replacement o f pumps and/or compressors. Alternatively,

extra area may be added in an ad hoc manner in an attempt to reduce velocities such

that pressure drop limitations are met (this prejudices the accuracy o f any targeting

procedure used). Design features which make poor use o f available pressure drop

may be retained in the retrofit (this results in higher capital expenditure than

necessary). It is therefore important that the interactions between the heat exchanger

network and its flo w system are properly appreciated and full consideration o f

pressure drop is made during network retrofit.

C hapter 7 R etrofit o f H F N 's (Targeting Procedu re) Page ¡6 3

In this chapter w e demonstrate how the new area algorithm (presented in Chapter

4) can he applied to develop a new retrofit targeting procedure which is based on

allowable stream pressure drops rather than stream coefficients.

7.2 Retrofit Targeting Procedure to Date

7.2.1 Targeting philosophy

’ Targeting* [Linnhoff 8c Flower. 1978] is the concept o f predicting, prior to

detailed design, how a process can best be operated in order to achieve a desired

result. In grassroots design the ’ target’ may be a minimum total annual operating cost

and components o f this target could be a related process energy consumption, the

heat exchanger network area and even pump and compressor power consumption

[Jegede, 1990].

T h e need for targets in retrofit have been clearly demonstrated by T joe and

Linnhoff [1986]. First, they have shown that the amount o f energy being saved in a

retrofit ( ’ the scope o f the project’ ) relates directly with project payback and that

project payback increases as project scope increases. This may seem obvious and not

at first sight a problem. However, they go on to show that a network structure that

is efficien t for one energy consumption is often not efficient fo r operation at another

level. It is therefore important that the correct project scope is set ahead o f network

design.
Chapter 7 Retrofit o f H E N ’s (Targeting Procedu re) P a g e 164

T jo e and Lin n h off present a further important reason for proper retrofit targeting.

In retrofit projects it is usual that more than one economic criterion must be satisfied.

These criteria are often set subjectively without any attempt to relate them. Provided

a design fits, or is within, all o f the set criteria, it is accepted as being viable. T h is

can result in the acceptance o f designs which fall a long w ay short o f what is truly

achievable.

Energy saving plant retrofits are usually subject to at least two econom ic

considerations. A lim it is generally placed on the level o f investment and a minimum

payback is required on any investment made. In the context o f heat exchanger

networks the investment is mainly in the form o f additional heat transfer surface and

the savings (from which the payback is estimated) com e in the form o f reduced

energy consumption. Thus, if we can relate energy saving with additional area

requirement at the targeting stage w e can predict the scope o f the project that best

conforms with both o f our economic constraints ahead o f any detailed design

calculations.
C ha pter 7 R etrofit o f H E N 's (Targeting Procedure) Pa ge Ï6S

7.2.2 Targeting using c o n sta n t-« a p p ro a ch

The area-energy plot provides a means o f evaluating retrofit projects. In retrofits,

the plot only considers heat recovery area, i.e. process-to-nrocess exchangers fTioe,

19861. The first step is to compare the performance o f the existing plant with what

is possible by plotting the plant’ s current performance (Point X in Figure 7.1).

In the example shown in Figure 7.1, the comparison shows that for the current

energy consumption the process requirements could have been achieved with

significantly less capital investment (Point C). Alternatively, with the level of

investment made the process could have operated at a significantly low er energy

consumption (Point A ).
Chapter 7 R etrofit o f H E N 's (Targeting Procedure) Page 166

T joe and Linnhoff then point out that the better retrofits are those which make

good use o f the investment that has already been made. T h e ’ ideal’ retrofit would

move from the existing inefficient point (X ) to the most efficient that could have

been achieved with the existing area (Point A ). H ow ever, given that any

modifications are going to require investment w e can expect to follow a path similar

to that shown in Figure 7.2.

Figure 7.2 : A typical retrofit path.

C hapter 7 R etrofit o f H E N ’s (Targeting Procedu re)

The optimum retrofit path cannot be determined simply. Th ey suggested a way

in which a bound could be placed on its location. They argued that a retrofit which

used area less efficien tly than the existing plant was unlikely to be cost effective.

Consequently, they gave a definition fo r area efficiency. They then assumed that all

efficient retrofits w ill at least maintain area efficien cy and consequently suggested

a constant efficien cy line on the plot. The definition o f area efficiency used by Tjoe

and Linnhoff is:

Π_ ^mxlstlng{polntX) (7.1)
A idoal (pointa

and the targeting curve is shown in Figure 7.3. N ote that the constant efficiency curve

and the ideal curve are divergent.

I
Ay

A ty
Ax

Atx

Ey E * Energy

Figure 7.3 : Retrofit targeting curve (constant a ).

C h a p ter 7 R etrofit o f H E N 's (Targeting Procedu re) Page 168

7.2.3 Targeting using incremental-q approach

T h e constant efficien cy curve does provide a reasonable bound on the location of

good retrofits. H ow ever, for the targeting o f project scope the curve can be too

conservative. Silangwa [1986] recommends that if the area efficiency o f the existing

network is 0.9, o r less, a curve parallel with the ideal should be used. There is a

strong case for this. W e know how area should be used in networks. W e can,

therefore, expect the area efficiency o f the network to improve as new area is added.

Silangw a’ s method assumes that all o f the new area is added in an ’ ideal’ manner

(Figure 7.4). Here, the incremental efficien cy curve and the ideal curve are parallel.

Figure 7.4 : Retrofit targeting curve (incremental a ).

C ha p ter 7 R etrofit o f H E N ’s (Targeting Procedu re) P a ge 169

From any o f the above targeting curves we can relate incremental investment in

the form o f area (A A in Figure 7.3) with incremental savings in the form o f energy

(A E in Figure 7.3). B y turning each o f these into monetary terms, it is possible to

compare the scope o f the project with the econom ic criteria o f payback and

investment limit and obtain the target scope (Figure 7.5).

The target scope can be related with a specific minimum temperature approach

between the composite curves. This provides the starting point fo r the actual design

work.

Figure 7.5 : Investment-saving plot

Chapter 7 R etrofit o f U R N ’s (T a rgeting Procedure)

7.2.4 Targeting using area matrix approach

Another approach for setting retrofit targets is called Area Matrix approach. A

general picture o f the approach is given here.

As explained in Chapter 4. minimum surface area o f a heat exchanger network is

achieved through vertical heat exchange between the composite curves (Equation 4.1).

This minimum area can be used as area target in grass-roots designs. H ow ever, in

retrofits the existing area has already been installed. Therefore, the retrofit objective

is to minimise the additional area that must be installed above the existing network

area.

Shokoya [1989] has presented a targeting procedure in which the minimisation o f

this additional area is aimed. The procedure calculates the area distribution between

the streams, given b y the composite curves fo r vertical heat transfer, and compares

them with the area distribution o f the existing network matches. The difference

between these values represents the total additional area that should be ideally

installed during retrofit.

C hapter 7 Retrofit o f H E N 's (Targeting Procedure) Page 171

7.3 Problems with the Existing Targeting Procedures

A ll existing targeting methods h ave tw o basic drawbacks. Both drawbacks relate

to the stream film heat transfer coefficien ts used.

The first drawback relates to the specification o f stream film heat transfer

coefficients. Currently these are based on the performance o f the existing exchangers.

This is meaningful if only one exchanger is present on a specific stream and no new

exchangers are to be added to that stream. However, i f there is m ore than one

exchanger on a stream and these exchangers have different heat transfer coefficients,

there is no systematic means o f d erivin g a single value that is properly representative

o f the stream.

The problem is compounded when it comes to the installation o f new exchangers.

Exchangers are not designed on the basis o f an assumed stream coefficient, or to

achieve a certain coefficient, but on the basis o f allowable pressure drop. The

coefficient assumed in targeting m ay not be consistent with that finally achieved in

detailed exchanger design. This inconsistency can result in poor targeting.

The second drawback relates to the assumption that the stream coefficients are

independent o f project scope. In m ost retrofits the engineer is faced with constraints

on the stream pressure drop. These arise from the often overriding need to retain

existing pumps or compressors and from physical constraints such as available

gravitational head. A s the amount o f surface in contact with a stream is increased, the
C ha p ter 7 R etrofit o f H E N ’s (Targeting P r o c e d u r e ) P a ge 172

pressure loss at a fixed velocity increases. For a fix e d pressure loss w e need to

reduce velocity if w e are going to increase the contact surface. In practical terms this

means placing new exchangers in parallel with the existin g ones, rather than in series

with a resultant decrease in clean film heat transfer coefficients.

W ith fixed pressure losses the ’ ideal* area-energy relationship based on constant

film heat transfer coefficients w ill be too optimistic. T h e true coefficients w ill fall as

the scope o f the project increases (Figure 7.6). This in turn means that the existing

targeting procedures are likely to be too optimistic.

Figure 7.6 : Comparison between ideal (minimum) area curves.

Chapter 7 R ctm fit o f H/tN’s (Targeting Procedure) P a g e 173

T o overcom e both o f these problems we need a new targeting procedure in which

the relationship between heat exchange area, stream pressure drop and heat transfer

coefficient is considered.

7.4 New Retrofit Targeting Procedure

7.4.1 Minimum area requirement in retrofit

T h e area algorithm presented in Chapter 4 can be used with either the procedure

initially developed by T joe [1986] or with the modified procedure suggested by

Silangwa [1986] to provide a method which allows proper consideration o f the

network’s flo w system. T o d o this, two important points should be taken into

consideration:

A) Multi-pass heat exchange area

In retrofit targeting using assumed heat transfer coefficients, the temperature cross­

over effects associated with multi-pass exchangers are not considered. This m eans no

regard is taken o f Fx correction factor when calculating the minimum area o f the

network (from the composite curves) and/or the area o f the existing process

exchangers. This is because the surface area efficiency (a ) should solely express

temperature criss-crossing effect in a network.

Chapter 7 Retrofit o / H E N 's (T argeting Procedure) Page 174

H ow ever, this is not necessarily true when applying fixed pressure drop area

algorithm. Here, we expect (a ) to express the overall efficiency o f the network, that

is temperature criss-crossing effect (A T efficiency) plus poor use o f pressure drop

effect (A P efficiency). In this strategy, it is the actual area o f the exchanger which

should be used when estimating the network minimum area and/or the total area of

the existing network.

W ith fixed stream pressure drops any additional area (whether due to installation

o f new exchangers or use o f multi-pass instead o f pure counter-current exchangers)

causes the true heat transfer coefficients to fall as the scope o f the project increases.

Therefore, any negligence about the difference between the multi-pass heat exchanger

area and the pure counter-current heat exchanger area results in a double error in

targeting results.

Figure 7.7 shows three area target versus A T min curves. These are based on

assumed coefficient, fixed AP, and multi-pass fixed AP, respectively. The comparison

highlights the fall-off o f heat transfer coefficients due to this effect.

C hapter 7 R etrofit o f H E N ’s (Targeting Procedure) Page 175

A rea (m 2 )

0 10 20 30 40 60 60 70 80 00
A T**,, (C)

Figure 7.7 : Area target versus A T niin for three different basis.

Area target is still based on process-to-process area. The performance o f the utility

exchangers is usually ignored. However, there are two circumstances in which the

performance o f these units is subject to separate consideration:

(i) in debottlenecking situation;

(ii) where part o f the pressure drop through the units can be freed for exploitation

in the network.
C hapter 7 Retrofit o f H E N 's (T argeting P rocedu re) Page 176

B ) specification of stream pressure drops

The first step in the procedure is the specification o f stream pressure drops. This

can in volve a number o f actions. First, the performance o f the existing exchanger

must be evaluated in order to determine how much pressure drop is currently being

used for heat transfer purposes. This should be follow ed by an analysis o f how much

o f a stream’ s head loss is due to gravitational needs and how much is due to flo w

through pipelines, valves, fittings etc. If the loss due to pipe-flow is large,

consideration should be given as to how this can be reduced such that more head is

available fo r heat transfer promotion. Finally, having determined the ’ useful’ stream

pressure drops these can be fed to the new area algorithm and the ’ ideal’ area-energy

plot can be generated.

A t this stage the designer may wish to consider the benefits o f pump replacement.

This is easily done in the targeting mode by systematically adjusting individual

stream pressure drops and generating new area-energy plot for the so revised data.

The difference in area requirement given by one o f these plots and the one based on

the existing plant provides an indication o f the likely exchanger capital cost benefits

that would result from pump replacement.

Chapter 7 R etrofit o f U E N ’s (Targeting Procedure) Pa ge 177

7.4.2 Capital cost estimation

A heat exchanger network generally involves several units and each unit may have

a number o f separate shells arranged in series or parallel. So. it is the number o f

shells (N ,) that determines the network capital cost.

In retrofits, given the additional surface area requirements (A A ) which can be

obtained from the targeting curve, the project capital or investment cost can readily

be calculated, assuming that new exchangers have equal area.

H ow ever, the number o f new heat exchanger shells (A N .) is difficult to predict

and depends on the existing units already present in the network. Often, an exchanger

requires additional shells after retrofit because o f increased surface area requirement

for that unit. Thus, a technique for estimation o f the number o f new exchangers in

retrofit is essential.

T jo e [1986] suggest that the average area o f the exchanger shells already present

in the existing network gives a reasonable representation o f the distribution o f new

area between new exchangers. This gives us:

An, AA
(7 .2 )
a v e ra g e a re a o f s h e ll
Chapter 7 R etrofit o f H E N ’s (Targeting P rocedu re) Page 178

Therefore, for a network which follow s the capital cost o f the form below

In s ta lle d C ost - a + b (A r e a )c

the investment cost for the project is given by

I n v e s t m e n t = A Na [ a +jb ( a v e r a g e a r e a o f s h e l l ) 0 ] <7 - 4>

7.4.3 The procedure

Having known the stream data, physical properties, existing energy consumption

and exchanger data, and also having finally determined what stream pressure drops

should be used, retrofit targeting can proceed in the manner illustrated in Figure 7.8.
Chapter 7 Retrofit o f H E N 's (Targeting Procedure) P a ge 179

Figure 7.8 : Summary o f the new targeting procedure.

Chapter 7 R etrofit o f H E N ’S (Ta r#clin g Procedure) P a ge 180

7.5 Sum m ary

1. A ll o f the existing targeting methods have two basic drawbacks, both o f which

relate to the assumptions made regarding stream film heat transfer coefficients. T o

overcom e these problems, a retrofit targeting procedure which is based on allowable

stream pressure drops and fully considers the existing flo w system has been

developed.

2. W ith fixed pressure drops, the true coefficients will fall as the scope o f the

project increases. This in turn means that the existing targeting procedures are likely

to be too optimistic.

3. In new procedure surface area efficiency (a ) should express the overall

efficien cy o f the network, that is temperature criss-crossing effect (A T efficien cy)

plus poor use o f pressure drop effect (A P efficiency). Therefore, actual area o f the

exchangers should be used when estimating the network minimum area and/or the

total area o f the existing network.

C h a p te r 7 Retrofit o f H E N ’s (Targeting P rocedu re) Pa ge 181

Nomenclature

a constant in capital cost law

A heat exchanger surface area

b constant in capital cost law

c constant in capital cost law

E energy consumption

Ft correction factor to the A T LM for multi-pass exchangers

h stream heat transfer coefficient

H enthalpy

N, number o f shells

R* stream fouling resistance

T temperature

a surface area efficiency o f a network

AA additional surface area

AE reduction in energy consumption

AN, number o f additional shells

AP stream pressure drop

AT temperature driving force

A T min minimum temperature approach on composite curves

Chapter 7 Retro fit o f H E N 's (Targeting Procedure) Pa ge 182

References

Ciric A.R. & Floudas C .A .,

’ Optimisation M odel fo r Retrofitting Heat Exchanger N etworks’ ,
I.Chem.E. Symposium Series 109, Manchester, 1988.

Jegede F.O.,
’Power, Capital and E n ergy Cost Trade-offs in Heat Exchanger Networks’ ,
Ph.D. Thesis. U M IS T , 1990.

Linnhoff B. & Flow er J.R.,

A.I.Ch.EJ., 24 (4): 633-654, 1978.

Shokoya C.G.,
’ Retrofit o f Heat Exchanger Networks for Energy Savings and Debottlenecking’ ,
Ph.D. Transfer Report, U M IS T , 1989.

Shokoya C.G.. & Kotjabasakis E.,

’ Retrofit o f Heat Exchanger Networks - Targeting’ ,
I.Chem.E. 17th Annual Meeting, Swansea. 5th April 1990.

Silangwa M.,
’ Evaluation o f Various Surface Area Efficiency Criteria in Heat Exchanger Network
Retrofits’ ,
M.Sc. Thesis. U M IS T , 1986.

Tjoe T.N.,
’ Retrofit o f Heat Exchanger Networks’ ,
Ph.D. Thesis. U M IS T , 1986.

Tjoe T.N . & Linnhoff B .,

’ Using Pinch T ech n o lo gy for Process Retrofit’ ,
Chem. Eng., April 28, 1986.
C hapter 8 Retrofit o f H E N 's (D esign Tools) P a ge 183

CHAPTER 8

RETROFIT

OF

HEAT EXCHANGER NETWORKS

( DESIGN TOOLS )
C ha pter 8 R etrofit o f H EN 's (D esign Tools) Page 184

Contents

8.1 Introduction

8.2 Cross-Pinch Exchanger Analysis

8.3 Driving Force Plot

8.4 Remaining Problem Analysis

8.4.1 Assumed-h R.P.A.

8.4.2 Fixed-AP R.P.A.

8.4.3 Combined use of R.P.A.

8.5 Exchanger Shifting

8.6 Ideal Pressure Drop

8.7 Summary

Nomenclature

References
Chapter 8 R etrofit o f l/ F N ’s (Design Tools) P a ge 185

8.1 Introduction

The new retrofit targeting method introduced in Chapter 7 determines the scope

o f the project that could be achieved within the constraint o f the required payback

and investment limitation. Knowing this scope, the Pinch temperatures and the

starting value o f A T min are readily identified from the composite curves. T h is gives

us the starting point fo r our retrofit design.

H owever, re-designing an existing network in retrofit is more d ifficu lt than

conducting a grass-roots design. Heat exchangers and pumps/compressors are already

installed. So, design is constrained by the structure o f the existing network and more

importantly by the existing flo w system.

The continued use o f existing heat exchangers, within the pressure drop constraint

set by the existing pumps/compressors, requires little new investment and does not

impose new space requirements. Therefore it is highly desirable to use as much of

the existing exchangers as possible. T o do so there are two approaches:

1) Producing a number o f grass-roots designs using targeting results and new

grass-roots design procedure presented in Chapter 5, and then selecting the network

which has structure similar to the existing network.

2) using existing heat exchangers as effectively as possible whilst making as few

changes to the existing network as possible.

Chapter 8 R etrofit o f H E N 's (D esign Tools) P a ge 186

Experience shows that the first approach is only suitable for cases in which the

composite curves are wide apart and many design options exist. It is difficult to apply

in "tight" problems. In this case the designer is forced to apply the second approach.

W hatever the approach, tools are needed to examine each individual exchanger

in the network and assess its contribution towards the overall performance o f the

network. In this way those exchangers which make good contribution can be

considered well placed and may be retained, and those which are poorly placed

and/or use excessive pressure drop can be corrected or used elsewhere.

In this chapter the existing design tools are briefly described. They are as follows:

- Cross-Pinch exchanger analysis,

- D riving Force Plot,

- Remaining Problem Analysis,

- Exchanger shifting.

A lso, new tools that best suit the objective o f the new retrofit procedure are

developed. These are:

- a new Remaining Problem Analysis that takes the pressure drop

effect as w ell as temperature driving force effect into account;

- estimation o f the exchanger’ s ’ ideal pressure drops’ .

Chapter 8 R etrofit o f H E N ’s (D esign Tools) Page 187

8.2 Cross-Pinch Exchanger Analysis

W e know that the Pinch divides the problem into two separate regions. In designs

meeting energy targets no heat is transferred between these tw o regions. But, in

inefficient designs, there is often a heat flo w across the Pinch which corresponds to

the excess utility requirement.

The relative position o f the composite curves, in retrofit, is determined from the

required energy saving (A E ) obtained in targeting stage. This gives us the Pinch

temperatures to be used in the design stage. Having known the Pinch position, the

energy saving can be determined as the amount o f heat thè existing network transfers

across the pinch (Figure 8.1).

Figure 8.1 : Heat transfer across the Pinch by the existing network.
C hapter 8 R etrofit o f H E N 's (Design T ools) P a ge 188

Therefore, one simple approach is to identify exchangers so positioned and to

correct or rearrange them so that this flow is eliminated [Tjoe, 1986].

This analysis appears to present a significant step forward towards designing

compatible retrofit networks. It identifies exchangers which are poorly placed in that

they transfer heat across the Pinch. Figure 8.2 shows tw o cross-Pinch exchangers

before and after the retrofit.

Figure 8.2 : Cross-pinch exchangers before and after the retrofit.

Cross-Pinch exchanger analysis can only be used to identify exchangers which are

thermally inefficient. It does not help the designer to find those which make poor use

o f pressure drop.
C hapter 8 R etro fit o f H EN 's (D esign Tools) P a ge 189

8.3 Driving Force Plot

Driving Force Plot is a tool fo r the identification and manipulation o f poor heat

transfer alignment between the composite curves. The plot consists o f aligned hot and

cold stream temperatures and is made up o f a series o f straight lines as shown in

Figure 8.3.

Te

Figure 8.3 : A typical Driving Force Plot.

Chapter 8 R etro fit o f H E N ’s (D esign T ools) P a ge 190

The slope o f each o f these lines is equal to the ratio o f the heat capacity flowrates

o f the hot and cold composites in that temperature interval. Included in the plot is the

line T h = T c. The distance between the plot and this line indicates the "tightness" o f

the problem.

A n y individual exchanger in a network is represented by a line superimposed on

this plot (Figure 8.4). Where the line perfectly coincides with the plot, the stream

matches in that exchanger are vertically aligned. The greater the separation between

the exchanger line and the plot, the poorer the alignment. Cross-Pinch exchangers can

also be identified. Their lines cross the hot and/or cold Pinch temperatures [Tjoe,

1986]. Again, the Driving Force P lot only provides information regarding the thermal

aspects o f a network. It may be usefully applied in correcting badly placed heat

exchangers.

Figure 8.4 : A heat exchanger line superimposed on a D.F.P.

Chapter 8 R etro fit o f H EN 's (D esign T ools) Pa ge 191

8.4 Remaining Problem Analysis (R.P.A.)

8.4.1 Assumed-h R.P.A.

Driving Force Plot is strictly a qualitative measure. The effect o f poor alignment

is evaluated quantitatively using Remaining Problem Analysis [Tjoe, 1986]. This

analysis can briefly be described as follows.

The elements o f the stream data associated with the exchanger being examined

(exchanger i) are removed from the data set (Figure 8.5). The A T min associated with

the targeted energy consumption given the "residual problem" (A T min,) is then

calculated. The minimum area requirement fo r the remaining data (A ^ ,) is then

determined. This is added to the area o f the exchanger (A j). Then, by dividing the

minimum area requirement fo r the original full data set (A *,*) by this sum, we

determine the maximum a (a __ ,) that can be obtained if the proposed exchanger is

accepted as part o f the final network design.

_____ A n ln ( 8 .1 )
“ max.i
A 1 + A n in . r

Exchangers having an close to unity are efficient units and can be left alone.

Those having an a mMX below the efficiency (a ) o f the existing network cannot be

accepted and their contribution to the network performance need to be improved.

Chapter 8 R etrofit o f H E N 's (Design Toots) Pa ge 192

This R.P.A. is conducted using the assumed-h area algorithm and its result is

referred to below as a ml<h.

8.4.2 Fixed-AP R.P.A.

Assumed-h R .P .A . provides a clear image about the thermal efficien cy o f the

exchangers. But. it tells us nothing about how w ell we are using pressure drop. It will

be seen later that poor use o f pressure drop results in poor use o f area (and hence

capital). H owever, given the new area algorithm (presented in Chapter 4) a R.P.A.

which looks at the combined effects o f individual exchanger pressure drop and
C h a p te r 8 R etrofit o f H E N 's (D esign Tools) Pa ge 193

exchanger placement on network performance is possible [Polley & Panjeh Shahi,

1990].

A g a in , for each exchanger we simply rem ove the stream elements (now including

the individual pressure drops) associated with the unit from the data set and re-solve

(F ig u re 8.6). The subsequent result o f this analysis is referred to below as a muiAp.

F igu re 8.6 : Remaining Problem Analysis based on fixed pressure drop.

C hapter 8 Retrofit o f H E N ‘s (Design Tools) P a ge 194

8.4.3 Combined use o f R.P.A.

Remaining Problem Analysis based on assumed film heat transfer coefficients tells

us how efficiently th e area is being used. Remaining Problem Analysis based on

fixed pressure drop tells us how w ell the exchanger is perform ing overall for it

considers the effects o f both temperature driving force and pressure drop.

B y using the t w o tools in combination, w e can clearly differentiate between

inefficiency due to thermal effects and those due to pressure drop effects. This can

be illustrated using a result from the case study described in Chapter 10.

With one exchanger we find that the a maxAP is 0.66. This tells us that the

exchanger is a poor one. However, we do not know if the problem is one o f use of

pressure drop or use o f temperature driving force. However, the value o f a maxjl which

is 0.99 tells us that th e exchanger makes good use o f temperature driving force. So,
e
through the com bined use of the tw o analys s w e can see that the exchanger is poor

because o f its use o f pressure drop. If w e can simply reduce (in this instance) the

pressure drop through the exchanger w e have an efficient exchanger and an

opportunity to im p rove the performance o f the network generally.

Chapter 8 R etrofit o f H E N "s ( Design Tools) Pa ge 195

8.5 Exchanger Shifting

One way o f improving the area e fficien cy o f an existing exchanger in retrofit is

"exchanger shifting". This type o f m odification is usually preferred to other types

such as exchanger relocation, for it is sim ple and no structural change to the network

is involved.

The position o f an exchanger on th e Driving Force Plot described above is

determined by five factors: the cold stream terminal temperatures, the hot stream

terminal temperatures, and the heat capacity ratio o f the two streams. These local

factors can be changed in order to obtain better stream alignment.

An exchanger shift can be made by adjusting the A T (Figure 8.7), changing the

heat capacity ratio (Figure 8.8), or even both in combination (Figure 8.9). however,

these adjustments have consequences that must be recognized.

Adjustments to the temperatures, at w h ich the exchanger operates, will change the

heat duty. The new heat duty is often less than the existing duty.
C hapter 8 R etrofit o f H E N 's (Design Tools) P a ge 196

Figure 8.7 : Exchanger shift involving change o f AT.

Changing the heat capacity ratio (which can be made either through the

introduction o f a stream splitting or altering the shells arrangement from series to

parallel) w ill reduce the stream heat transfer coefficients and hence the heat duty of

the exchanger. Thus, the performance o f the shifted exchanger should be rated again,

in order to find the actual heat duty o f the exchanger in the new position. It should

be noted that in current practice stream coefficients are alw ays assumed fixed, even

when the stream heat capacity flowrate is changed [T jo e & Linnhoff, 1987].
Chapter 8 R etro fit o f H E N 's (D esign Tools) P a g e 197

Figure 8.8 : Exchanger shift involving change o f CP ratio.

With both modifications, therefore, more surface area is required to fill the created

heat pockets. This can be achieved by altering the size o f either the exchanger under

study, the size o f adjacent exchangers in the network or by introducing new

exchangers into the network.

O nce an exchanger has been shifted the Remaining Problem Analysis should be

applied to check whether the shift has been successful.

C ha p ter 8 R etro fit o f H E N 's (Design Tools) Page 198

Figure 8.9 : Exchanger shift involving change o f both A T and C P ratio.

C hapter 8 Retrofit o f H E N 's ( Design Tools) Pa ge 199

8.6 Ideal Pressure Drop

The efficient use o f pressure drop is analogous in some ways to the effective use

o f temperature driving force. The overall pressure drop is fixed. Assume that there

are tw o exchangers on a stream, each having identical heat duties and temperature

driving forces. If the pressure drop through the first exchanger is high w e have a

local benefit o f high velocities, high coefficients and small area. H ow ever the high

pressure drop through the first exchanger results in only a low pressure drop being

available for use in the second. The result is low velocities, lo w coefficients and high

area. The overall result is an increased area requirement.

Jegede [1987] has shown that, even when the two exchangers have ’ opposing

thermal resistances’ (film resistance on the other stream plus fouling and wall

resistances) which differ b y orders o f magnitude, the area requirements o f the overall

system are within 2% o f the minimum when the same stream velocities (and hence,

film coefficients) are used in both exchangers. This means that the stream pressure

drop is best used uniformly through the use o f a constant velocity (and, constant

coefficient).

On the other hand, w e have a linear relationship between A P and A , provided the

exchanger makes efficient use o f AT. W e also know the overall stream pressure drop

and an estimate o f the overall stream contact area, hence, the ideal pressure drops can

be determined as shown in Figure 8.10.

Chapter 8 R etrofit o f H E N 's (D esign Tools) P a ge 200

s tre a m d a ta

s tre a m i (S )
- 0 -
AP4 . A0

s tre a m j (T ) APj . A
0

a c c e p ta b ly p la c e d

06 A M
< Ap. = APt
A d

06 A 1
= AP,
A «J

Figure 8.10 : Ideal pressure drops for a given exchanger.

C hapter 8 Retrofit o f H E N 's (D esign Tools) Page 201

8.7 Sum m ary

Design tools are needed to examine each individual exchanger <n the network and

assess its contribution towards the overall performance o f the network.

The existing design tools can only be used fo r assessment (and correction, if

necessary) o f the thermal efficiency o f the exchangers. They provide no information

regarding how well the exchangers are using pressure drop. Therefore, new tools have

also been developed. These are:

- A new Remaining Problem Analysis that takes the pressure drop effect as well

as temperature driving force effect into account. B y using this tool and the

conventional R .P.A . in combination, we can clearly differentiate between inefficiency

due to thermal effects and those due to pressure drop effects.

- ’ Ideal pressure drop’ estimation o f the exchangers. This can be used as a guide

when correcting an exchanger which inefficiently uses pressure drop.

Chapter 8 R etrofit o f H E N 's (D esign Tools) Pa ge 202

Nomenclature

A heat exchanger surface area

A ci contact area o f stream V

Ae surface area o f exchanger ’ E ’

Aj surface area o f exchanger ’ i ’

A,,,*,, minimum area requirement o f a heat exchanger network

Anji,,, minimum area requirement o f remaining heat exchanger network

CPc specific heat o f hot stream

CPh specific heat o f hot stream

H enthalpy

Qi heat duty o f the exchanger ’ i ’

T temperature

Tc cold interval temperature on the composite curves

Th hot interval temperature on the composite curves

a surface area efficiency o f a network

a mtx maximum surface area efficiency o f a network

a fflMh ctn,,, o f a network based on assumed coefficient

a maxi o f 3 network after exchanger T accepted

a mt. AP a,,,,, o f a network based on fixed pressure drop

AE energy reduction in retrofit

AP pressure drop

APj pressure drop o f stream ’ i ’

A PS shell-side pressure drop

C ha pter 8 R etrofit o f H E N 's (Design T ools) Pa ge 203

APt tube-side pressure drop

AT temperature driving force

A T mm minimum temperature approach on composite curves

A T min, A T niin for the remaining problem

Chapter 8 R etro fit o f H E N 's (Design Tools) Rage 204

References

P olley G.P. & Panjeh Shahi M .H.,

’Process Integration Retrofit Subject to Pressure Drop Constraint’ ,
Process Technology Proceedings, 9, pp. 31-36, Elsevier, 1990.

T jo e T.N .,
’ Retrofit o f Heat Exchanger Networks’
Ph.D. Thesis. U M IS T , 1986.

T jo e T .N . & Linnhoff B.,

’ A chieving the Best Energy Saving Retrofits’ ,
A IC h E Annual meeting, Houston, Texas. U SA, March 1987.
C h a p te r 9 R etrofit o f H E N 's (Design Procedu re) Page 2 05

CHAPTER 9

RETROFIT

OF

HEAT EXCHANGER NETWORKS

( NEW DESIGN PROCEDURE )

Chapter 9 R etrofit o f H E N 's (D esign Procedure) P a ge 206

Contents

9.1 Introduction

9.2 Analyzing E xisting Exchangers

9.3 Improving the Performance o f Inefficient Exchangers

9.4 Placing N e w Exchangers

9.5 Evolution

9.6 Summary

Nomenclature

References
Chapter 9 Retrofit o f H E N s (D esign Procedure) Page 207

9.1 Introduction

In retrofit design we aim to m inim ise the number o f changes we make to the

network in achieving the targeted goal. In this strategy, a four stage procedure that

has been proposed by T joe [1986] can b e used. H owever, adaptation is needed to

make it suitable for the new retrofit procedure. These four stages are as follow s:

1. Analyze a ll the exchangers in th e existing network, and keep those which

operate efficiently in terms o f b o th A T and AP utilisation.

2. Improve those elements that o p e ra te inefficiently by changing the temperature

fie ld around them (to im prove A T inefficiency) and!or applying other

techniques such as changing the shell arrangem ent (to im prove AP

inefficiency).

3. Add new exchangers in ord er to match the heat sources and the heat sinks

identified in stage 2 and to ob ta in the required energy saving.

4. Finally, evolve the resulting netw ork by exploiting the ’loops and paths ’

present in the network in o rd e r t o sim plify the design.

In this chapter, the above stages are described in more details and the design

procedure is then presented.

C ha pter 9 Retrofit o f H EN 's (Design Procedure) Page 208

9.2 Analyzing Existing Exchangers

In retrofit study, the existing exchangers can be analyzed using

Cross-Pinch exchanger analysis, to identify those exchangers that transfer heat

across the Pinch;

D riving Force Plot, as a qualitative measure, to s e e how vertical heat is

transferred by the match; and

e
T h e two Remaining Problem Analys s in com bination, in order to determine

both the A T and the A P efficiencies o f individual exchangers.

In this stage, the designer can decide if those exchangers which make good

contributions to the overall utilisation o f temperature d rivin g force and pressure drop

are to be retained. Exchangers that are deemed essential f o r safety or operability

reasons can also be considered. In making these decisions the designer has full

know ledge o f the penalties involved.

Summary o f the procedure for analyzing the existing exchangers is illustrated in

Figure 9.1.
C hapter 9 R etrofit o f H E N 's (D esign Procedure) P a g e 209

Figure 9.1 : Procedure fo r analyzing the existing exchangers.

Chapter 9 R etrofit o f H E N 's (D e s ig n Procedure) Page 210

9.3 Improving the Performance o f Inefficient Exchangers

Here, the designer examines how the p oor exchangers can be improved, with

respect to utilisation o f temperature d rivin g force and stream pressure drop. When

improving both temperature driving force and pressure drop, the temperature driving

force correction must be made first. T h is is because A P and exchanger area are

related and the use o f an ideal pressure drop requires the use o f the correct area

[Polley, Panjeh Shahi, 1990].

For the thermal correction o f an exchanger, the driving Force Plot and the

exchanger shifting techniques are used and the results are checked using assumed-h

R.P.A. [Tjoe, 1986].

H owever, when improving the A P e ffic ie n c y o f a poor exchanger during retrofit,

this can be studied in a number o f ways. T w o o f them are described here.

1. One way is by determining the ’ uniform ’ stream heat transfer coefficients

associated with the targeted A T min and then comparing the performance o f the

existing exchanger with these values.

2. Another way is b y determining the "ideal pressure drop" for both sides o f the

exchanger and using them as a modification guide when improving an existing

exchanger.
Chapter 9 R etrofit o f H E N ’s (D esign Procedure) Page 211

W hatever the approach used, the modification techniques are the same. If the

existing exchanger is operating at too high a coefficient (or pressure drop), this can

be rectified by

changing from a system with shells in series to one in parallel (with multi­

shell exchangers);

changing exchanger header arrangements;

swapping from shell-side to tube-side; or

adding additional area in parallel with the existing unit.

If the unit is operating at too low a coefficient (or pressure drop), this can be

rectified by placing shells in series rather than parallel. H o w ever, experience to date

shows that in retrofit in which area is being added to the netw ork, such exchangers

do not present a real problem and the designer w ill usually d ecid e that changes are

not justified.

A ll exchanger corrections must be follow ed by the application o f the fixed-AP

R.P.A. to check whether or not the exchanger has been corrected successfully.

Summary, o f the procedure for improving the perform ance of inefficient

exchangers is illustrated in Figure 9.2.

Chapter 9 R etrofit o f H E N 's (Design Procedure) Page 2 1 2

R em ain in g P ro b lem

Id e n tity ex c h a n g e r w h ich
re q u ire s Im p ro v e m e n t
A T m ln ,r OCmazJi 06 max. AP

Im p rove OCmaz, AP
using availa b le
tech n iqu es

1
Cheok CCmm-r, A P
using fix e d AP
R.P.À.

I--------------------1 No
I N ex t sta g e p----------

Figure 9.2 : Procedure for improving the inefficient exchangers.

C ha p ter 9 Retrofit o f f/ E N 's <Design Procedure) Page 213

9.4 Placing New Exchangers

Once all o f the existing exchangers are well placed, w e have to focus our effort

to fill the heat pockets (that remain after various exchanger shifts and/or changing the

shell arrangements) in order to com plete the design. T o do this, the driving force plot

and the new heat exchanger design algorithm (presented in Chapter 3) can be used.

A lso, the Remaining Problem Analysis is again applied to check the efficiency o f the

new exchangers.

T h e approach involves the fo llo w in g steps:

1. P la ce a new match such that i t fits the driving fo rc e plot.

2. Calculate the 1-1 area o f the match using targeted coefficients.

3. Check the A T efficiency o f the match using assumed-h R .P .A

4. I f the efficiency o f the p rop osed match is acceptable, then remove the stream

elements associated with that match fro m the problem and draw the resulting

driving fo rc e plot. Otherwise, return to step 1 and im prove the match o r select

another match.

5. Repeat steps 1 to 4 until the design is completed.

6. Distribute, linearly, the rem aining stream pressure drops between the new

matches (analogous to what we do in grass-roots design).

7. Perform the detailed design o f the exchangers using the new algorithm

presented in Chapter 3.
Chapter 9 R etrofit o f H E N 's (Design P rocedu re) P a ge 214

Summary o f the procedure for placing new exchangers is illustrated in Figure 9.3.

Figure 9.3 : Procedure for placing new exchangers.

C hapter 9 R etrofit o f H E N 's (D esign Procedure) P a g e 2 15

9.5 Evolution

A t the end o f stage 3 we have a network which fulfils the process objectives and

achieves the energy target. However, the design can usually be improved. For

example:

M aking the structure even more compatible with the existing network.

Reducing the number o f units required.

This improvement is achieved by the redistribution o f heat loads around loops

and, if necessary, along paths.

C ha pter 9 R etrofit o f H E N 's (D esign Procedure)

9.6 Summary

1. A four stage design strategy has been adopted and m odified to suit the new

retrofit procedure. In this strategy, exchangers which make good use o f both

temperature driving force and pressure drop are first retained. This is follow ed by

correcting inefficient exchangers. Once all exchangers are efficient, new matches are

added and the resulting network evolved.

2. When dealing with thermal aspects o f the exchangers (i.e. analyzing,

im proving or placing new matches), the conventional tools and procedures which are

all based on one tube-pass exchangers can be used. This includes the Driving Force

Plot, the exchanger shifting techniques and assumed-h R.P.A.

3. When the combined effects o f individual exchanger pressure drop and

exchanger placement on network performance is studied, all the calculations

(including exchanger rating, exchanger design and fixed-AP R .P .A .) must be based

on multi-pass heat exchangers.

C h a p ter 9 R etrofit o f H E N ’s (D esign Procedure)

Nomenclature

minimum area o f the network

oc surface area efficien cy o f a network

maximum surface area efficiency o f a network

o f a network based on assumed coefficient

^m »x.A P a,,,,,, o f a network based on fixed pressure drop

AP pressure drop

AT temperature driving force

A T min minimum temperature approach on composite curves

A T niin for the remaining problem

Chapter 9 Retrofit o f H E N 's (D esign Procedu re) Page 218

References

Polley G.P. & Panjeh Shahi M.H.,

’ Process Integration R etrofit Subject to Pressure Drop Constraint’ ,
Process Technology Proceedings, 9, pp. 31-36, Elsevier, 1990.

Tjoe T.N.,
’ Retrofit o f Heat Exchanger Networks’
Ph.D. Thesis. U M IS T , 1986.
Chapter IO R etro fit o f H E N 's (Case Study) Page 219

CHAPTER m

RETROFIT

OF

HEAT EXCHANGER NETWORKS

( CASE S T U D Y)
Chapter 10 R etrofit o f U E N ’s (Case Study) P a ge 220

Contents

10.1 Introduction

10.2 T h e Problem

10.2.1 Process flowsheet

10.2.2 Existing network

10.2.3 Physical properties

10.2.4 Cost data and project scope

10.3 Preliminary Calculation

10.3.1 Rating o f the existing exchangers

10.3.2 Estimation o f the stream h and A P

10.3.3 Equivalent 1-1 area o f the existing exchangers,

using calculated stream coefficients

10.4 Targeting

10.4.1 Theoretical curve

10.4.2 Targeting curve

10.4.3 Equivalent 1-1 area o f the existing exchangers,

using targeted stream coefficients

Chapter 10 R etrofit o f IIE N 's (C ase Study) P a g e 221

10.5 Design

10.5.1 Analyzing the existing exchangers

10.5.1.1 Cross-pinch exchanger analysis

10.5.1.2 Driving Force Plot

10.5.1.3 Remaining Problem Analysis

10.5.1.4 Making decisions

10.5.2 Improving the inefficient exchangers

10.5.2.1 Correcting exchanger E6 for A T

10.5.2.2 Correcting exchanger E6 for AP

10.5.2.3 Correcting other exchangers

10.5.3 Placing new exchangers

10.5.4 Evolution

10.6 Summary

Nomenclature
C hapter 10 R etrofit o f H E N ’s <Case Study) Pa ge 222

10.1 introduction

In Chapter 1 we discussed the hazards o f ignoring pressure drop during the retrofit

studies. W e also mentioned a project which was abandoned, because it had been

found that the retrofitted network would have required the replacement o f pumps

costing around twice o f the additional exchanger costs.

In chapters 7,8 and 9 a new retrofit method was developed, in order to overcome

the drawbacks associated with the current methods. In the new method the

interactions between the heat exchanger network and its flo w system were properly

appreciated and full consideration o f pressure drop was made during network retrofit.

N o w , in this chapter, we are going to apply the new method to the above project,

and see if we can manage to design a network which is run by the existing pumps,

whilst consumes less energy.

Chapter 10 R e tro fit o f H E N 's ( Case Study)

10.2 The Problem

An area in which the new targeting procedure is likely to prove particularly useful

is in the retrofitting o f crude o il pre-heat train. In these plants the crude oil pumps

are usually very costly. If a retrofit can be achieved without a need fo r pump

replacement, it would often be very welcomed.

10.2.1 Process flowsheet

The process flowsheet is shown in Figure 10.1, in a sim plified form.

As can be seen, the crude oil feed stream is pre-heated in two sections by

interchange with the hot fractions resulting from the distillation columns. The first
Chapter 10 R etrofit o f H E N ’s (Case Study) P a ge 224

section runs from storage to a desalter unit and the second from the desalter to the

crude tower. Process heating is provided by a furnace, which pre-heats the crude into

the crude tower.

10.2.2 Existing network

The heat exchanger network shown in flowsheet form in Figure 10.1 is reproduced

in grid form in Figure 10.2, with stream temperatures, match heat loads and specific

heat mass flowrates marked. The hot streams are grouped together in the top part o f

the grid and numbered from 1 to 6. The only cold stream is drawn below and

numbered 7.

Figure 10.2 : Existing network for the Crude Pre-heat Train.

C hapter 10 R etrofit o f H E N 's (C ase Study) P a ge 2 25

In order to estimate the stream pressure drops available for heat transfer, we have

to rate the performance o f the existing exchangers. The exchanger specifications are

also given in Table 10.1.

El E2 E3 E4 E5 E6

Shell side Kero. Naph. HGO LG O BPA ATB

Tube side Crude Crude Crude Crude Crude Crude
Area (m2) 280 1480 280 800 2760 1360
Ser x Par 1x 1 lx l 1x 1 2 x 1 3x2 4x1
Shell ID (mm) 940 1524 940 1143 1219 1143
B affle Spacing 255.3 1246.4 197.3 419.3 605.1 509.1

Tube Count 1075 2827 1075 1590 1810 1590

Tube Passes 2 2 2 2 2 2
Tube ID (mm) 15.4 15.4 15.4 15.4 15.4 15.4
Tube O D (mm) 19.1 19.1 19.1 19.1 19.1 19.1
Tube Pitch(mm) 25.4 25.4 25.4 25.4 25.4 25.4

Table 10.1 : Exchanger specifications for Crude O il Pre-heat Train.

10.2.3 Physical properties

The Representative values o f physical properties associated with individual

streams (such as density, specific heat, viscosity and thermal conductivity) together

with the stream flowrates are tabulated in Table 10.2.

Chapter 10 R etro fit o f H E N ’s (Case Study) Page 226

St. Stream Flow Q cp P k

No. Name (kg/s) (kg/m3) U/kg ‘ O (cP) (W/m ’ C)

1 Kerosene 23 700 2600 0.3 0.12

2 LGO 44 700 2600 0.4 0.12
3 H GO 13 750 2600 0.5 0.12
4 ATB 56 750 2600 0.5 0.12
5 Naphtha 253 630 2600 0.2 0.12
6 BPA 148 750 2600 0.4 0.12
7 Crude 200 800 2600 1.0 0.12

Table 10.2 : Flowrates and physical properties for Crude O il Pre-heat Train.

10.2.4 Cost data and project scope

In addition to above information we also need cost data and specification on the

project scope. This information is given in Table 10.3.

Cost Data
Hot Utility : 70 £/kw.yr
Cold Utility : 7 £/kw.yr
Exchanger Capital : 8600 + 6 7 0 »(A rea )“ ' 3

Proiect Scope
Investment Lim it : £ 2M
Maximum Payback : 2 years

Table 10.3 : Cost data and project scope for Crude O il Pre-heat Train.
Chapter 10 R etrofit o f H E N 's (C ase Study) Pa ge 227

10.3 Preliminary Calculation

10.3.1 Rating o f the existing exchangers

W e start the study with an analysis o f the performance o f the existing exchangers.

This provides information on the clean heat transfer coefficients, fouling resistances

and pressure drop currently available for use in heat transfer. The analysis used an

exchanger rating program based on Bell-Delaware method. T h e results are given in

Table 10.4.

El E2 E3 E4 E5 E6

h „ <W/m2 *C) 1947.2 898.0 1947.2 1422.6 736.6 1422.6

h„ (W/m2 ’ C ) 1690.2 2432.9 1144.6 1485.6 1577.8 1419.7

AP, (k Pa) 32.835 11.560 32.835 31.272 10.776 53.164

AP, (k Pa) 45.634 65.363 29.476 59.886 74.763 85.964

R ,„ (m2 'CAW .00144 .00137 .00140 .00142 .00157 .00142

R ,„ (m2 ' C M .00144 .00137 .00140 .00142 .00157 .00142

Table 10.4 : Performance o f the existing exchangers.

C ha pter 10 R etrofit o f H E N 's (C ase Study) Pa ge 228

10.3.2 Estimation o f the stream A P and H T C

K now ing the tube-side and shell-side pressure drops o f the individual exchangers

and also the route through the network a particular stream is flowing, the A P

(associated with heat transfer surface) can easily be calculated for that stream.

Know ing the stream pressure drops, representative clean coefficients for individual

streams (as opposed to individual exchangers) can be estimated using the basic

pressure drop relationship applying to that stream:

(1 0 .1 )

The resulting clean coefficients are then converted to fouled heat transfer

coefficients using stream fouling resistances.

The results o f this calculation are given in Table 10.5.

C h a p te r 10 R etrofit o f H E N 's (C ase Study) Page 229

Stream Stream Clean H TC Fouling H TC AP

N o. Name (W/m2 ’ C) (m2 X / W ) (W/m2 *C) (k Pa)

1 1 Kerosene 1690.2 0.00144 492.2 45.634

2 LG O 1485.6 0.00142 477.8 59.886
3 HGO 1144.6 0.00140 439.8 29.476
4 ATB 1419.7 0.00142 470.7 85.964
5 Naphtha 2432.9 0.00137 561.5 65.363
6 B PA 1348.4 0.00157 432.6 74.763
7 Crude 1135.2 0.00147 343.0 172.442

T ab le 10.5 : HTC, Fouling Resistances and A P ’s for the streams.

T h e rest o f stream data for the plant is presented in Table 10.6.

U £
Stream Supply Temp. Target Temp. HTC
o. V*
u
No. CC) C C) (W/m2 *C)

1 180.0 30.0 59.8 492.2

2 270.0 40.0 114.4 477.8

3 350.0 30.0 33.8 439.8

4 380.0 50.0 145.6 470.7

5 150.0 100.0 657.8 561.5

6 290.0 190.0 384.8 432.6
7 20.0 390.0 520.0 343.0

T a b le 10.6 : Stream data.

Chapter 10 Retrofit o f H E N 's (C ase Study)

10.3.3 Equivalent 1-1 a r e a of the existing exchangers, using calculated stream

coefficients

Before starting the targeting stage, there is one more figu re which should be

calculated if the conventional method (based on fixed coefficien t) is to be applied.

This is equivalent 1-1 area o f the existing network. W e calculate this value only

because it w ill later be used for comparison o f the two methods.

Exch. Duty hs hT U A T ,.M A,.,

Name (M W ) (W /m 2 ’ C) (W/m2 -C) (W/m2 *C) re t (m2)

El 6.00 492.2 343.0 202.1 97.4 304.7

E2 23.00 561.5 343.0 212.9 78.8 1371.4
E3 7.50 43 9 .8 343.0 192.7 129.5 300.6
E4 15.00 477.8 343.0 199.6 90.4 831.3
E5 38.48 432.6 343.0 191.3 83.3 2415.2
E6 80.42 47 0 .7 343.0 198.4 78.0 1421.6

Table 10.7 : A,., o f the existin g exchangers using calculated coefficients.

Know ing the 1-1 area o f the exchanger units and also the number o f shells in each

unit (Table 10.1), a total network area o f 6644.7 m2 and an average shell size o f

553.7 m2 are obtained.

Chapter 10 R etro fit o f l ¡ U N ’s (C ase Study) R a ge 231

10.4 Targeting

10.4.1 Theoretical/ideal curve

Having established both stream data (Table 10.6) and individual pressure drops

(Table 10.5) the area-energy plot can be developed. In order to demonstrate the

importance o f considering pressure drop during targeting this is undertaken fo r both

a fixed pressure drop situation and a fixed heat transfer coefficient situation. The

comparison is given in Figure 10.3.

A r e a (M 2 )

Figure 10.3 : Comparison between fixed-h and fixed-AP theoretical curves.

Chapter 10 Retrofit o f llE N 's (Case Study) Pa ge 232

The comparison o f the two approaches shows the importance o f pressure drop

considerations. A t the existing energy consumption, the overall area based on pressure

drop is about 7% lower than that based on heat transfer coefficients. What does this

signify? Both minimum area predictions are based on vertically aligned temperatures

so w e can rule out thermal inefficiencies. Both areas represent possible solutions

utilising the same stream pressure drops. The difference results from the poor use of

pressure drop in existing plant, with some units using higher velocities than that

associated with a uniform coefficient. W e see that better use o f pressure drop would

have resulted in higher coefficients overall and a useful saving (potentially about 7%)

in overall area.

10.4.2 Targeting curve

L et’ s carry on the study with our new method which is based on fixed pressure

drop. When the existing plant performance is compared with the fixed pressure drop

ideal curve, it is found that the plant has an area efficien cy o f 0.818. Therefore, an

incremental-a target curve parallel to the ideal curve can be used. Also, a total

existing area o f 6960 n r and an average shell size o f 580 m2 have been used to

estimate the additional area and the investment cost.

The economic criteria to be applied to the retrofit is an investment ceiling o f £2M

with a maximum payback period o f 2 years. The project seems to be investment

(rather than payback) controlled. Thus, a project costing f 2M and resulting in 30%

energy saving (giving a payback o f 1.06 years) is identified.

C hapter 10 R etrofit o f H E N ’s (Case Study) P a ge 233

The targeting results can be presented in a tabular form (rather than

investment/saving plot) as shown in Table 10.8

STREAMS - CRUDE STM UTILITIE S - CRUDE UTY

GEOMETRY - CRUDE GEO PHYS PROP. - CRUDE PRP

R e t r o f i t Range T a rge t (F ix e d AP)

DTmin U t . Reduc. Add 1-2 A Alpha S avin g In vestm en t P ay Back

I*CJ (kW) (M2) (E / yr) (£) C yr)

2 .0 42512.00 83253.55 0.986 3273424 20158816 6 .1 58

j 6 .0 40432.00 44679.29 0.975 3113264 10809872 3 .4 7 2
10.0 38352.00 32208.39 0.967 2953104 7807837 2 .6 4 4
14.0 36272.00 24855.21 0.959 2792944 6018661 2 .1 55
16.0 34192.00 19875.20 0.952 2632784 4800823 1 .8 2 3
22.0 32112.00 15899.97 0.943 2472624 3834381 1.551
26.0 30032.00 13801.11 0.938 2312464 3345746 1 .4 4 7
30.0 27952.00 11297.95 0.929 2152304 2718763 1 .2 6 3
34.0 25872.00 9461.31 0.921 1992144 2279577 1 .1 4 4
36.9 24364.00 8286.13 0.915 1876027 1996035 1 .06 4
38.0 23792.00 7883.95 0.91 3 1831984 1920157 1 .04 8
42.0 21712.00 7402.31 0.910 1671824 1798430 1 .0 7 6
46.0 19632.00 6092.04 0.901 1511664 1489248 0 .9 8 5
50.0 17552.00 5007.59 0.892 1351504 1222881 0 .9 0 5
54.0 15472.00 4090.68 0.883 1191344 988166 0 .8 2 9
58.0 13392.00 3302.45 0.874 1031184 808424 0 .7 8 4
62.0 11312.00 2615.71 0.865 871024 647643 0 .7 4 4
66.0 9232.00 2010.77 0.856 710864 471431 0 .6 6 3
70.0 7152.00 1472.93 0.847 550704 369973 0 .6 7 2
74.0 5072.00 990.97 0.837 390544 248396 0 .6 3 6
78.0 2992.00 556.13 0.828 230384 135826 0 .5 9 0
82.0 912.00 161.49 0.818 70224 54186 0 .7 7 2
83.75 0 .0 0 0.0 0 0.818 0 0

Table 10.8 : Retrofit range targeting based on fixed pressure drop.

Underlying the use o f a parallel targeting curve is the assumption that the area

efficiency o f the network w ill increase as the scope o f the retrofit increases. This is

because all additional area is assumed to be installed ideally. W ith the quantity of

additional area proposed for the targeted project, the area efficiency o f the network

was expected to improve to a value o f 0.915 (from the existing value o f 0.818).
Chapter 10 Retrofit o f H E N ’s (Case Study) Page 23d

A s described in Chapter 4. the new area algorithm not only estimates the network

area as a function o f A T min, but also indicates the heat transfer coefficients associated

with each A T min. Table 10.9 shows the stream heat transfer coefficients and the

contact areas corresponding to targeted A T min.

Stream Stream Exchanger Targeted h Contact Area

No. Name Side (W/m2 *C) (m 2)

1 Kerosene Shell 482.7 373.7

2 LG O Shell 447.1 1874.2
¡ 3 HGO Shell 393.2 917.7
4 ATB Shell 430.5 4068.8
5 Naphtha Shell 551,7 2053.4
6 BPA Shell 437.8 4664.2
7 Crude Tube 329.5 13952.0

Table 10.9 : Stream coefficients and contact areas at A T mjn = 36.9 *C.

10.4.3 Equivalent 1-1 area of the existing exchangers, using targeted stream

coefficients

K now ing the targeted values for the stream coefficients, we can now calculate the

equivalent 1-1 area o f the existing exchangers to be used in design stage. These

values are necessary for undertaking the fixed-h Remaining Problem Analysis and

exchanger shifting (Table 10.10).

Chapter 10 R etrofit o f H E N 's <Case Study) P a ge 235

Exch. Duty hs hr U A T lm A,.,

Name (M W ) (W/m2 ’ C ) (W/m2 'C l (W/m2 ’ C) C C) (m2)

El 6.00 482.7 329.5 195.8 97.4 314.5

E2 23.00 551.7 329.5 206.3 78.8 1415.3
E3 7.50 393.2 329.5 179.3 129.5 323.1
E4 15.00 447.1 329.5 189.7 90.4 874.9
E5 38.48 437.8 329.5 188.0 83.3 2457.6
E6 80.42 430.5 329.5 186.6 78.0 1511.1

Table 10.10 : A,., o f the existing exchangers, using targeted coefficients.

10.5 Design

As it was discussed in Chapter 9, in design stage we aim to minimise the number

o f changes w e make to the network, within the pressure drop constraints, in order to

achieve the targeted goal. This procedure consists o f four main steps, o f which

analyzing the existing exchangers is the first.

10.5.1 Analyzing the existing exchangers

Here, w e determine which exchangers are well placed and make efficient use of

pressure drop and consequently can be retained unchanged in the network. In doing

this, cross-Pinch exchanger analysis. Driving Force Plot and Remaining Problem

Analysis can be used.

Chapter 10 R etrofit o f H E N ’s (Case Study) Page 236

10.5.1.1 Cross-Pinch exchanger analysis

Given the targeted A T min, and hence the Pinch temperatures, a grid highlighting

the heat exchangers crossing the Pinch is produced (Figure 10.4 ). A s it can be seen,

exchangers E3 and E6 and also heater H are identified as transferring heat across the

Pinch and hence must be corrected if energy is to be saved.

T« np . ( C ) Duty ( k¥ ) m Cp ( k ¥ / C )

Figure 10.4 : Cross-Pinch exchangers with A T min o f 36.9’ C.

T o determine the amount o f heat being transferred by each unit, the follow ing

calculation is undertaken (Table 10.11):

C hapter 10 Retrofit o f H E N ’s fCase Study) P a ge 237

Unit Calculation Heat (k W )

E3 (350 - 290) * 33.8 2028

E6 (380 - 290) * 145.6 13104

Furnace (H ) (253.1 - 235.35) * 520 9230

Total 24362

Table 10.11 : T h e total amount o f heat being transferred across the Pinch by the
existing network is equal to the targeted value o f the energy reduction.

10.5.1.2 D riving Force Plot

Now. let us see how well the existing exchangers are located on the D riving Force

Figure 10.5 : D rivin g Force Plot showing the placement o f the existing exchangers.
Chapter 10 R etrofit o f H E N 's (Case Study) P a ge 238

Again, it is observed that none o f the existing exchangers are transferring heat

vertically, and to accept them we need to apply exchanger shifting. However, before

making any decision we have to analyze them, quantitatively, using Remaining

Problem Analysis.

10.5.1.3 Remaining Problem Analysis

Application o f the two R.P.A. to the existing exchangers at the targeted energy

requirement gives the results presented in Table 10.12.

A T miBJ„ = 36.9’ C Constant C oefficient Fixed Pressure Drop

QHMt = 56056 kW „ = 13215 m2 A = 13952 m2

Exchanger Rem. Exis. Rem . Exis. Rem. ^Tn»x

Name A T m„ A,., A ,.,,» (h) A |.2 A-1-2.min (A P)

El 36.9 314.5 13046 0.989 280 14031 0.975

E2 36.9 1415.3 12025 0.983 1480 13130 0.955
E3 33.0 323.1 14314 0.903 280 15879 0.863
E4 22.6 874.9 13690 0.907 800 16428 0.810
E5 -ve 2457.6 ... ... 2760 ... ...

E6 7.8 1511.1 21924 0.564 1360 292578 0.047

(E1+E2) 30.4 1729.8 12100 0.956 1760 13582 0.909

Table 10.12 : Remaining Problem Analysis.

Chapter 10 R etrofit o f H E N ’s (Case Study) Page 239

This analysis shows:

- Whilst exchanger E l appears poorly placed on the D.F.P., this unit is actually

quite efficient giving a thermal efficiency o f 0.989. A value o f 0.975 for the

overall efficien cy (a niMiAp) indicates that the unit also makes good use o f pressure

drop. Therefore, it can be retained unchanged.

- Exchanger E2 is also not too bad and may be retained. H ow ever, if we propose

to accept more than one match, we must repeat the analysis in order to determine the

combined effect. Therefore, the R.P.A. are again applied, to check if the retention o f

both exchangers (E1.E2) is acceptable. As the results (in the bottom line o f the table)

suggest, this unit can not be kept in current operation condition and should be

modified.

- Exchangers E3 and E4 are violating the targeted A T min. T h e values o f and

(Om«Ap) f ° r these units indicates that they are operating inefficiently in terms o f both

temperature driving force and pressure drop. Hence, they both require improvement.

- Exchangers E5 and E6 are cross-Pinch exchangers. So, they show a larger

temperature driving force violation. The large difference between the thermal and the

overall efficiency o f the Exchanger E6 (0.564 & 0.047) indicates that this unit is very

poor in terms o f pressure drop utilisation.

Chapter 10 R etrofit o f H E N ’s (C ase Study) P a ge 240

10.5.1.4 M a k in g decisions

The use o f Remaining Problem Analysis shows that on ly exchanger E l is

operating efficien tly, but that all o f the others have to be improved. H ow can w e

achieve these improvements?

As it was discussed earlier, when improving both temperature driving force and

pressure drop, the temperature driving force correction must be made first. The

D riving Force Plot shows the direction in which temperature changes have to be

made.

Initial inspection o f the Driving Force Plot indicates that the placement o f these

exchangers can be improved using exchanger shifting techniques. However, having

just one cold stream in this case study suggests that stream 7 must be splitted. This

suggestion is supported when the pinched grid (shown in Figure 10.4) is inspected.

As Figure 10.4 features, there are 2 pinched hot streams above and 3 below the

Pinch. A ll these streams should be matched with pinched cold streams, if a good

degree o f heat transfer verticality is to be achieved as a result o f exchanger shifting.

Therefore, o n e possible solution would be to break the cold stream at a point near the

Pinch and split the resulting segments (above and below the Pinch) into 2 and 3

branches, respectively. In doing this, the slope o f the most constrained matches (i.e.

E6 in above and E3 and E5 in below the Pinch) on the D.F.P. can be used as a guide

when distributing the stream 7 ’s flowrate between the branches.

C h a p ter 10 R etro fit o f H E N 's (C ase Study) Pa ge 241

10.5.2 Improving the inefficient exchangers

Having retained exchanger E l, w e proceed with improving the other exchangers.

T h e exchanger shifting techniques and the Driving Force Plot fo r the remaining

problem should be first used to correct the thermal inefficiency o f the poor units.

Figure 10.6 shows the Driving Force Plot associated with the remaining problem.

Which exchanger should be corrected first? Design is generally most difficult in

the region where the temperature driving force is small (i.e.: at the Pinch). W e would

therefore correct exchangers in this region first. Exchanger E6 lies in such region and

transfers the largest amount o f heat (13104 k W ) across the Pinch. It is also violating

the A T min and, hence, can be shifted first. Moreover, when this unit is corrected, the

breaking point o f the stream 7 is determined.

Now, in order to see how correction o f a poor exchanger proceeds in different

steps, we first study the correction o f the exchanger E6 and its sole effect on the

network performance, and then check for the combined effect o f the exchangers E l

and E6.
Chapter 10 Retrofit o f H E N 's (Case Study) Pa ge 242

10.5.2.1 Correcting exchanger E6 fo r AT

The Driving F o rc e Plot (Figure 10.5) suggests that if E6 is shifted horizontally to

the right (taking the hot end temperatures o f 380*C and 282.2#C as a reference point

for the shift and m aking small adjustment to its slope), it w ill fully superimpose on

the Driving Force P lo t. Also, as a result o f this adjustment, the cold stream (above

the Pinch) is splitted as decided above. Figures 10.6 and 10.7 illustrate the corrected

placement o f the E 6 on the D.F.P. and in the network, respectively.

A fter E6 has been shifted, the fixed-h R .P.A . is used to check on the

improvement. For this, the new stream elements associated with the corrected

exchanger E6 are rem oved from the problem. The result shows an a mMxit o f 0.998 and

a A T min, o f 36.9’ C. Hence, we can consider this to be a good improvement. But, is

this sufficient?

L e t’s look at the overall efficiency o f E6. Application o f the fixed-A P R.P.A. to

the problem gives an a ma.xAP of 0.659. This means that the exchanger is not

performing effic ie n tly overall. The comparison o f the tw o analysis tells us that, since

the thermal e ffic ie n c y o f the unit is 0.998, the overall inefficiency is because o f the

poor use o f pressure drop. Therefore, the unit should be m odified with respect to AP

utilisation.
Chapter 10 Retrofit o f H E N 's (Case Study) P a ge 243

Figure 10.6 : D.F.P. showing E6 after correction for AT.

mCp

[I F -

[IP
0 = ---------------------------------------------------------------- @JU

Figure 10.7 : E6 after correction for A T .

C h a p ter 10 R etrofit o f H E N 'a (Case Study) Page 244

10.5.2.2 C orrectin g exchanger E6 fo r A P

B efore w e can improve the use o f pressure drop in E6, w e need to know whether

the unit uses too much or too little AP. T o know this, ideal pressure drop can be

estimated for both sides o f the unit. The result o f this estim ation is presented in Table

10.13. Included in the table are the existing pressure drops and the values which will

be resulted if the unit is only corrected for AT.

Status o f E6 Tube-side A P Shell-side AP

In existing condition 53.164 kPa 85.964 kPa

A fter correction for A T 41.132 kPa 85.964 kPa
Ideal pressure drops 15.380 kPa 26.291 kPa !

T a b le 10.13 : Pressure drop utilization in unit E6.

A s the comparison o f the figures indicates, the unit is consuming much higher

pressure drop than that is necessary. There are a number o f ways o f rectifying this.

A simple, but efficient, way would be switching from an arrangement o f four shells

in series to tw o banks in parallel. Having done this m odification, the is found

to im prove to 0.971.

T h e results o f the R .P.A . for different steps o f the correction are presented in

T ab le 10.14.
Chapter 10 Retrofit o f H E N ’s (Case Study) P a g e 245

A T minJ. = 36.9‘ C Constant Coefficient Fixed Pressure D ro p

Q hj—• = 56056 kW A , . , * * « = 13216 m2 A « * » = 13952 m 2

Status o f Rem. Exis. Rem. Gtm»x Exis. Rem.

E6 A T min A,., Am». (h) A ,.2 ^l-2.min (A P )

existing 7.8 1511.1 21924 0.564 1360 292578 0.047

A T corrected 36.9 1511.1 11725 0.998 1360 19820 0.659
A P corrected 36.9 1162.1 12057 1.000 1360 13014 0.971

Table 10.14 : Remaining Problem Analysis for the correction o f E6.

In this table, the comparison between a mMj, o f the second and the third steps is

interesting. It arises, because when unit is corrected for AP, the exchanger load is

decreased and hence the placement o f the exchanger line on D.F.P. is again im proved

(Figures 10.8 and 10.9).

N ow , let us look at the performance o f the network after E l has been accepted

and E6 corrected. Application o f both fixed-h and fixed-AP R.P.A. to the problem

gives an o f 0.989 and a niJtx^P o f 0.975 (Table 10.15).

Chapter 10 R etrofit o f H E N ’s (Case Study) Page 246

A T minI^ = 36.9‘ C Constant Coefficient Fixed Pressure Drop

Q Htm = 56056 kW A ,.,,*.», = 13216 m2 A,.2,nin„ = 13952 m2

Exchanger Rem. Exis. Rem. Exis. Rem. «m «

Name A T min A,., A ,.lm« (h) A,_2 Ai.2.min (AP)

El 36.9 314.5 13046 0.989 280 14031 0.975

E6 36.9 1162.1 12057 1.000 1360 13014 0.971
(E1+E6) 36.9 1476.6 11888 0.989 1640 12670 0.975

Table 10.15 : Remaining Problem Analysis for E l, E6 and (E1+E6).

In this table:

- The comparison between the three a nttxit’ s demonstrates that since the placement

o f the unit E6 has been corrected 100%, it has no longer a contribution in thermal

inefficiency o f the network. Therefore, the a m4x-h for (E1+E6) is the same as for E l.

- The comparison between the three a ^ ^ ’ s is also interesting. When E6 is

studied alone, the overall efficiency is 0.971. This is because the pressure drop

through the unit E6 is now slightly less than the ideal value and this results in a local

penalty. H ow ever, when E l and E2 are considered together, the efficiency is higher

(0.975). Th is is because E l consumes A P more than the ideal value, and hence that

penalty is now partially compensated fo r in the unit E l (the fixed-h R.P.A. does not

recognise A P interactions and benefits). Figures 10.8 and 10.9 illustrate the final

placement o f the E6 on D.F.P. and in the network, respectively.

C hapter 10 R etro fit o f H E N ’s (C ase Study) Page 2 47

Figure 10.8 : D.F.P. showing E6 after correction fo r A T and AP.

Figure 10.9 : Unit E6 after correction for A T and AP.

 R etrofit o f H E N ’S (Case Study) P a ge 248
C hapter 10

i 0.5.2.3 C o rrectin g other exchangers

I) Exchanger E5

Figure 10.10 shows the Driving Force Plot for the remaining problem. T h e next

most critical unit to be corrected is E5. It is the biggest one in size and violating the

A T„,„ such that the A T „,„, becomes negative. Simple observation suggests that the unit

can be improved by shifting as shown in the figure. H owever, the below-the-Pinch

part o f the cold stream should first be splitted into 3 branches as discussed in sub­

section 10.5.1.4. Figure 10.11 demonstrates the placement o f the exchanger E5 In the

network after it has been corrected for both A T and AP.

C hapter 10 R etrofit o f H E N ’S (C ase Study) Page 249

0 40 80 120 180 200 240 280 320

To

Figure 10.10 : D.F.P., showing E5 before and after corrections.

Figure 10.11 : Unit E5 after correction for A T and AP.

C hapter 10 R etrofit o f H E N ’s (Case Study) P a ge 250

The improvement o f this unit proceeded in 3 steps:

step 1 - shifting, for A T correction.

step 2 - switching from an arrangement o f tw o banks to three banks in parallel,

fo r A P correction.

step 3 - swapping the hot stream from shell-side to tube side, fo r further A P

improvement.

Table 10.16 shows the Remaining Problem Analysis after each step.

A T min.lar = 36.9-C Constant Coefficient Fixed Pressure Drop

Q i.,« = 56056 kW = 13216 m2 A ,.* ,,.» , = 13952 m2

Exchanger Rem. Exis. Rem. ®max Exis. Rem. ®mix

Name/step A T „„ A,., A|.|.min <h) A|-2 Al-2.min (A P)

(E1+E6) 36.9 1476.6 11888 0.989 1640 12670 0.975

(E1+E6+E5)
step 1 33.7 3662.7 9717 0.988 4400 14162 0.752
step 2 33.7 3312.5 10066 0.988 4400 10675 0.925
step 3 33.7 3573.5 9806 0.988 4400 10352 0.946

Table 10.16 : Remaining Problem Analysis fo r the correction o f E5.

Chapter 10 R etro fit o f H E N 's (C ase Study) Page 251

II) Exchangers E2 and E3

The D riving Force Plot for the remaining problem together with the exchanger

lines are illustrated in Figure 10.12. Both exchangers E2 and E3 require thermal

improvement and can be shifted independently (shifting exchanger 2 w ill not affect

the operating temperatures o f exchanger 1 and vice versa). So, the order in which

they are tackled is not critical.

Let us correct exchanger 2 first. The Driving Force Plot shows the placement o f

E2 can be improved by shifting it in a direction shown in the figure, maintaining the

inlet cold stream temperature at 31.54*C. The cold stream segment with the largest

C p is selected fo r this match, because, in this case, the slope o f the exchanger line

fits the plot.

The improvement o f E2 is undertaken in 2 steps:

step 1 - shifting, for thermal correction.

step 2 - swapping the hot stream from shell-side to tube side, for A P improvement.

The Remaining Problem Analysis (fixed-h and fixed-A P) are subsequently used

to check on the improvements. The result shows an a mMjl o f 0.988 and an

0.945 (Table 10.17). Hence, w e accept these improvements and m ove on to the next

unit (i.e. E3).

C hapter 10 R etro fit o f H E N 's (Case Study) Page 252

Exchanger E3 is one o f the exchangers that transfer heat across the Pinch. The

observation o f the D.F.P. and the Pinch temperatures suggests that this unit can be

improved by matching stream 3 with one o f the cold stream segments (the one which

have smallest Cp), taking the Pinch temperatures as a reference point for the shift.

Having done this, the unit closely fits the plot and no longer transfers heat across the

Pinch.

The improvement o f E3 is also undertaken in 2 steps:

step 1 - shifting, for thermal correction.

step 2 - swapping the hot stream from shell-side to tube side, for A P improvement.

Applying o f the Remaining Problem Analysis (fixed-h and fixed-A P) we find an

a mMXJt o f 0.987 and an o mstxAP o f 0.943 (Table 10.18). A good improvement has been

made. The new locations and the heat loads o f these tw o units are illustrated in

Figure 10.13.

Now, w e can rem ove the stream elements (including the new A P ’s) associated

with exchangers E2 and E3 from the remaining problem, and m ove on to correct the

last exchanger (i.e. E4).

Chapter 10 R etro fit o f H E N 's (C ase Study) Pa ge 253

Th 400

300

320

280

240

200

100

120

80

40

0
0 40 00 120 100 200 240 280 320
To

Figure 10.12 : D.F.P., showing E2 and E3 before and after corrections.

Figure 10.13 : Units E2 and E3 after correction fo r A T and AP.

C hapter 10 R etrofit o f H E N 's (C ase Study) Pa ge 254

A T minIW = 36.9‘ C Constant Coefficient Fixed Pressure Drop

Q „ lir = 56056 kW A ,.1-w, = 13216 m2 A|-2.m«.uc = 13952 m2

Exchanger Rem. Exis. Rem. Exis. Rem.

Name/step A T m„ A ,., A ,.,*, (h) A|-2 ^l-2jnin <AP)

(E1+E6+E5) 33.7 3573.5 9806 0.988 4400 10352 0.946

(E1+E6+E5
+E2)
step 1 33.7 4780.8 8595 0.988 5880 9918 0.883
step 2 33.7 4948.9 8432 0.988 5880 8887 0.945

Table 10.17 : Remaining Problem Analysis for the correction o f E2.

I--------------------------------------
A T mku„ = 3 6.9’ C Constant Coefficient Fixed Pressure Drop
Q hj* = 56056 kW = 13216 m2 = 13952 m2

Exchanger Rem. Exis. Rem. «mu Exis. Rem. «mu

Name/step A T min A,., (h> A,.2 A 1-2.min (A P)

(E1+E6+E5
+E2) 33.7 4948.9 8432 0.988 5880 8887 0.945
(E l +E6+E5
+E2+E3)
step 1 32.5 5078 8304 0.987 6160 10443 0.840
step 2 32.5 5102 8281 0.987 6160 8637 0.943

Table 10.18 : Remaining Problem Analysis for the correction o f E3.

Chapter 10 R e tr o fit o f H E N 's (Case Study) P a ge 255

II I ) Exchanger E4

The Driving Force Plot for the remaining problem is illustrated in Figure 10.14.

Unlike exchangers E3, E5 and E6, this unit is away from the Pinch and, hence, it has

a more flexible reference point fo r the shift. H owever, a good shift can be started

from a point such that the exchanger line fits the plot, namely, the cold end

temperatures o f 152*C and 104*C.

Again, the improvement o f this unit is proceeded in 3 steps:

step 1 - shifting, for A T correction.

step 2 - switching from an arrangement of two shells in series to tw o shells in

parallel, for A P correction.

step 3 - swapping the hot stream from shell-side to tube side, fo r further A P

improvement.

The results o f the Remaining Problem Analysis for this unit is presented in Table

10.19. As can be seen, the values o f 0.988 and 0.935 for and a mtxAP,

respectively, indicates that the unit is now operating efficiently. T h e new locations

and the heat loads and temperatures o f the improved unit is shown in Figure 10.15.

The stream elements (including the new A P’s) associated with exchangers E4 are

removed from the remaining problem . A ll existing exchangers are now w ell placed

and make efficient use o f pressure drop. Note that parts o f the streams’ heat and

pressure drop are now available fo r placing new matches. H ow ever, in order to
C hapter 10 R etrofit o f H E N 's (Case Study)

estimate the remaining stream pressure drops, w e need information on pressure drops

being consumed by the improved units. These information, which have been obtained

through rating o f the exchangers in new circumstances, are summarised in Table

10.20.

A T miBJâr = 36.9'C Constant Coefficient Fixed Pressure Drop

Q h,iw = 56056 kW Al-I.mta.ur = 13216 m2 A ,.3.„ta.ur = 13952 m2

Exchanger Rem. Exis. Rem. a mmx Exis. Rem. «n »,

Name/step A T mJn A,-, A,.,mta <h) A ,.2 A ,.,.-. (AP)

(E1+E6+E5
+E2+E3) 32.5 5102 8281 0.987 6160 8637 0.943
(E1+E6+E5+
E2+E3+E4)
step 1 32.5 5781 7608 0.984 6960 9920 0.826
step 2 32.5 5596 7784 0.987 6960 7969 0.934
step 3 32.5 5611 7769 0.988 6960 7961 0.935

T ab le 10.19 : Remaining Problem Analysis for the correction o f E4.

Exchanger Hot stream A P (k Pa) Cold stream A P (k Pa)

El 45.634 (shell side) 32.835 (tube side)

E2 16.244 (tube side) 11.949 (shell side)
E3 0.223 (tube side) 17.632 (shell side)
E4 0.280 (tube side) 15.500 (shell side)
E5 2.682 (tube side) 21.135 (shell side)
E6 10.744 (shell side) 5.892 (tube side)

T ab le 10.20 : Pressure drops being consumed by the corrected units.

C hapter IO Retrofit o f H E N 's (Case Study) Page 257

Figure 10.14 : D.F.P., showing E4 before and after corrections.

Figure 10.15 : Unit E4 after correction fo r A T and AP.

C hapter 10 Retrofit o f H E N 's (C ase Study) Page 258

10.5.3 Placing new exchangers

The driving force plot for the remaining problem with all the new proposed

matches indicated are shown in Figure 10.16. A ll new matches fit the Driving Force

Plot except for exchangers B,D and G, where placement o f these units were

considered to be necessary for compatibility and the area penalties were small. The

thermal efficiency (a *) o f the network is 0.917 compared with the initial value o f

0.870 and the targeted value o f 0.936.

Figure 10.16 : D.F.P. showing the placement o f new matches.

The complete design grid which involves all the corrected units as w ell as new

matches is illustrated in Figure 10.17.

C hapter 10 R etrofit o f H E N ’s (Case Study) Page 259

Figure 10.17 : Complete design grid showing the n ew matches A to H.

N ow , w e can linearly distribute the remaining stream pressure drops between the

new matches and undertake the detailed design o f the exchangers using the new

algorithm presented in Chapter 3.

The results o f this stage are presented in Tables 10.21 and 10.22.
Chapter 10 R etrofit o f H E N ’s (Case Study) Page 260

Exchanger A B C D

Area (m2) 254.9 2492.4 615.3 1535.4

Shells (ser x par) 1x 1 5x1 1x 1 4x1

Shell Diameter (mm) 520.3 956.1 955.8 953.2

No. o f Tubes 263 938 934 932

Tube Passes 1 2 2 2

Tube O.D. (mm) 19.1 19.1 19.1 19.1

Tube I.D. (mm) 15.4 15.4 15.4 15.4

Tube Pitch (mm) 25.4 25.4 25.4 25.4

Tube Length (mm) 16157 8870 10957 6875

B affle Cut (% ) 38.4 25.7 34.9 25.7

No. o f Baffles 48 11 9 8

AP, (k Pa) 5.892 C 40.058 H 10.026 H 25.136 H

A P, (k Pa) 9.923 H 63.715 C 17.380 C 39.980 C

Table 10.21 : Results o f the detailed design for new units (Crude O il Pre-heat Train).
C hapter 10 R etrofit o f H E N 's (C ase Study)

Exchanger E F G H

Area (m 2) 711.7 438.1 532.4 1997.1

Shells (ser x par) 1x 1 5 x 1 2 x 1 4x1

Shell Diameter (mm) 1571.4 380.1 670.1 1112.1

No. o f Tubes 2580 128 451 1292

Tube Passes 4 1 i 2

Tube O.D. (mm) 19.1 19.1 19.1 19.1

Tube I.D. (mm) 15.4 15.4 15.4 15.4

1 Tube Pitch (mm) 25.4 25.4 25.4 25.4

Tube Length (mm) 4592 11361 9871 6451

Baffle Cut (% ) 28.2 33.2 24.6 26.1

No. o f Baffles 6 55 20 7

AP, (k Pa) 20.509 C 19.330 H 14.520 C 72.080 H

| AP, (k Pa) 49.119 H 116.08 C 59.606 H 62.740 C

Table 10.22 : Results o f the detailed design for new units (Crude O il Pre-heat Train).
Chapter 10 R etrofit o f H E N ’s (Case Study) Pa ge 262

10.5.4 E volution

The initial network can now be evolved using loop breaking to improve

compatibility with the existing network. W e see that new match E and unit E2, F and

unit E3, G and unit E4, and H and unit E5 can be merged and form four single units.

Therefore, the final network features 10 heat exchanger units as illustrated in Figure

10.18.

Figure 10.18 : Final re-designed network for Crude O il Pre-heat Train.

Chapter 10 R etrofit o f H E N 's (C ase Study) Page 263

The resulting final network saves 24.364 M W o f energy (1876027 f/yr) which is

exactly the target value. It requires the addition o f 8577 m2 o f new area which is

3.5% above the targeted value o f 8286 m2. The overall efficien cy (a^p) o f the

network has improved as a result o f retrofit (from 0.818 to 0.898) and compares with

the expected figure o f 0.915. The required investment for this project is £2 125 450

which is 6.5% higher than target (£1 996 035). The results o f targeting and design

are compared in Table 10.23.

Targeting Design Difference

Energy saving (M W ) 24.364 24.364 0.0 %

Area requirement (m2) 8286.1 8577.3 +3.5 %
Overall efficiency 0.915 0.898 -1.7 %
Saving (£/yr) 1876027 1876027 0.0 %
Investment (£) 1996035 2125450 +6.5 %
Payback (yrs) 1.064 1.133 +6.5 %

Table 10.23 : Comparison between targeting and design.

Chapter 10 Retrofit o f H E N 's (Case Study)

10.6 Sum m ary

N ew retrofit method (both the targeting and the design procedures) have been

effectively applied in a case study involving the retrofit o f a Crude Oil Pre-heat

Train.

Application o f the new procedures has made possible to install additional area (in

order to reduce energy consumption) in the network, with no requirement o f any

pump replacement.

The design tools have also been used during network design and all o f them found

to be useful. In particular, the combined use o f Remaining Problem Analysis proved

to be a very powerful means o f differentiating between inefficiency o f a given

exchanger due to thermal effects and those due to pressure drop effects.

This case study has also shown that the procedures are simple and with a thorough

understanding o f the underlying principles they can give meaningful and accurate

results in practical applications.

Chapter 10 R etro fit o f H E N ’s (Case Study) Page 265

Nom enclature

A,., surface area o f exchanger, based on one tube pass

A ,.1 -u « minimum area target o f a HEN, based on one pass exchangers

A 12 surface area o f exchanger, based on multi tube pass

“ I-2.min.wr minimum area target o f a HEN, based on multi pass exchangers

Cp specific heat

heat transfer coefficient inside the tubes

h„ heat transfer coefficient outside the tubes

hs shell-side heat transfer coefficient

hT tube-side heat transfer coefficient

k thermal conductivity

Q h .Mt hot utility target for a HEN

Rf.» fouling resistance inside the tubes

Rf.o fouling resistance outside the tubes

u overall heat transfer coefficient

«h surface area efficien cy o f a network, based on fixed-h

ctmix o f a network based on assumed coefficient

a m«xAP a max o f a network based on fixed pressure drop

C^AP surface area efficien cy o f a network, based on fixed-AP

APS shell-side pressure drop of stream

AP t tube-side pressure drop o f stream

a t lm logarithmic mean temperature difference

A Tmjn minimum temperature approach on composite curves

Chapter 10 R etrofit o f I/ E N 's (Case Study) Page 266

A T min fo r the remaining problem

A T minJM targeted A T min for the remaining problem

e density

visco sity
C hapter 11 C on clu sio n s and Future Work Page 267

CHAPTER U

C O N C L U S IO N S

AND

F U T U R E WORK
C h a p ter I I Conclusions and Future W ork Page 268

C ontents

11.1 Conclusions

11.1.1 Pressure drop and networking

11.1.2 Grass-roots design

11.1.3 Retrofit design

11.2 Future Work

Chapter 11 Conclusions and Future Work P a g e 269

11.1 Conclusions

The thesis was organized in three parts with an aim towards the incorporation o f

pressure drop in process integration. W e can now assess the contribution o f each part

in achieving the ob jective o f the thesis.

11.1.1 Pressure d r o p and netw orking

In current practice, the design o f heat exchanger networks is concluded in three

stages. These are targeting, network synthesis and detailed exchanger design. A ll of

the techniques currently used in the first two stages make use o f assumed film heat

transfer coefficients, and leave the stream pressure drops to be considered at the

detailed design stage.

In Chapter 1, w e have demonstrated that this delay in considering pressure drop

can lead to serious design mistakes such as:

1. inconsistency between network synthesis and detailed design, with respect to

surface area o f the exchangers;

2. incorrect capital-energy trade-off and network optimisation; and

3. misleading targeting results during retrofit projects.

C hapter 11 Conclusions a n d Future Work P a ge 270

It has, therefore, been made clear that the targeting and synthesis procedures

should be based on stream allowable pressure drop and not assumed heat transfer

coefficient. In order to develop such procedures, we need a simple expression that

relates the allowable pressure drop o f a stream to the surface area and the film heat

transfer coefficient o f the stream.

In Chapter 2, various integral and analytical models for the shell-side performance

have been surveyed and amongst them the Bell-Delaware method has been selected.

Then, a new pressure drop relationship has been developed for the shell side

performance using the selected model.

In deriving the above relationship, w e have also found that a simple relationship

exists between pressure drop, exchanger area and heat transfer coefficient for

exchangers that exhibit geometrical similarity. This conclusion is supported by Kern

and confirmed by Bell-Delaware m odel, and can be extended to modem and

commercial heat exchanger design methodologies.

In Chapter 3, we have developed a n ew rapid design algorithm for shell-and-tube

heat exchangers (using Bell-Delaware m odel) which ensures that full use is made o f

the allowable pressure drops o f both o f the streams. This algorithm appears to be the

first one which makes full use o f both allowable pressure drops and thereby identifies

the smallest exchanger for a given duty. W e need this design algorithm because

efficient use o f pressure drop through heat exchanger networks demands that the

individual exchangers fully utilise what is available.

C hapter 11 Conclusions and Future Work P a ge 271

Because o f application o f the Bell-Delaware model, the algorithm appreciates

different baffle leakage and bundle bypass streams, and hence ensures that the shell-

side pressure drop is not overestimated. It should be noted that, since the design is

often based on a permissible value o f pressure drop, overprediction o f pressure drop

may seriously affect the heat transfer design o f the exchanger and may result in an

oversized unit.
Chapter 11 Conclu sions and Future Work P a ge 272

11.1.2 G rass-roots design

In Chapter 4, a new algorithm has been developed for network area estimation

using pressure drop relationship presented in chapter 2. This area algorithm enables

us to use allowable pressure drops rather than heat transfer coefficients and, therefore,

to overcome the drawbacks discussed in Chapter 1.

Also, we have made the algorithm flexible by allowing the process streams to be

specified in different types, such as pre-specified coefficient, pre-specified pressure

drop, pre-optimised pressure drop, or even selected velocity.

Use of assumed film heat transfer coefficients in targeting can lead to wrong

design initialisations and subsequently to non-optimal network designs.

In Chapter 5, we have applied the above area algorithm and established a new

targeting procedure. This new procedure gives a sharper relationship between network

area and energy consumption than that predicted on the basis o f assumed coefficients.

This is because it takes account o f variation in film heat transfer coefficients

throughout the trade-off between energy consumption and network area. Therefore,

the new targeting procedure not only yields the near optimum energy recovery level

and network area requirement but also the heat transfer coefficients to be used in

network design.
Chapter I I Conclusions and F utu re Work P a ge 273

The procedure has been tested using a case study and it has been found that the

results o f targeting and network synthesis are in a good agreement.

In Chapter 6, w e have studied the interface between network synthesis and

exchanger detailed design and concluded that consistency between synthesis and

equipment design can only be achieved if both operations are conducted on the same

basis. This means that synthesis must be based on stream allowable pressure drops

and not on assumed film heat transfer coefficients.

H owever, given fixed pressure drop, any area penalty resulting from poor

alignment in network synthesis actually results in reduced heat transfer coefficients

and further area penalty. So, poor alignment has a ’double penalty’ . Therefore, to

avoid ’double penalty’ , tick-off heuristics introduced by Pinch Design Method should

be used with caution when handling problems that force poor verticality o f stream

matches.
C hapter 11 Conclusions and Future Work P a ge 274

11.1.3 R etrofit design

A ll o f the existing retrofit targeting methods have tw o basic drawbacks, both o f

which relate to the assumptions made regarding stream film heat transfer coefficients.

These are:

1. Inconsistency between the assumed coefficients and those finally achieved in

detailed design o f new exchangers to be installed.

2. Lack o f recognition o f the dependency o f the stream coefficients on project

scope, which does exist because o f constraints on the stream pressure drops.

With fixed pressure drops, the true coefficients w ill fall as the scope o f the project

increases. This in turn means that the existing targeting procedures are likely to be

too optimistic.

In Chapter 7, a new retrofit targeting procedure has been developed to resolve the

above problems. The procedure is again based on allowable stream pressure drops

and fully considers the existing flo w system.

When it comes to the design stage, tools are needed to examine each individual

exchanger in the network and assess its contribution towards the overall performance

o f the network. The existing design tools can only be used fo r assessment (and

correction, if necessary) o f the thermal efficiency o f the exchangers. They provide no

information regarding how w ell the exchangers are using pressure drop.
Chapter 11 Conclusions and Futu re Work Page 275

In Chapter 8, new design tools have also been developed. These are:

1. A new Remaining Problem Analysis that takes the pressure drop effect as well

as temperature driving force effect into account. By using this tool and the

conventional R.P.A. in combination, w e can clearly differentiate between

inefficiency due to thermal effects and those due to pressure drop effects.

2. ’ Ideal pressure drop’ estimation of the exchangers. This can be used as a guide

when correcting an exchanger which inefficiently uses pressure drop.

In Chapter 9, a four stage design strategy has been adopted and m odified to suit

the new retrofit procedure. In this strategy, exchangers which make good use o f both

temperature driving force and pressure drop are first retained. This is follow ed by

correcting inefficient exchangers. Once all exchangers are efficient, new matches are

added and the resulting network evolved.

In Chapter 10, the new retrofit targeting and design procedures have been

effectively applied in a case study involving the retrofit o f a Crude Oil Pre-heat

Train. Application o f the new procedures to this case study has made possible to

reduce energy consumption by 30% (through installation o f additional area in the

network), with no requirement o f any pump replacement.

C ha pter 11 Conclu sions and Futu re Work P a ge 276

The design tools have also been used during network design and all o f them found

to be useful. In particular, the combined use o f Remaining Problem Analysis proved

to be a very powerful means o f differentiating between inefficiency o f a given

exchanger due to thermal effects and those due to pressure drop effects.

This case study has also shown that the procedures are simple and with a thorough

understanding o f the underlying principles they can give meaningful and accurate

results in practical applications.

C hapter 11 Conclu sions and Futu re W ork Page 277

11.2 Future W o rk

In Part 3 o f the thesis (R etrofit), we have shown h ow stream pressure drops in the

existing network can be determined and used as fixed parameters in targeting stage.

One question that remains to be answered is whether it would be beneficial to replace

the existing pumps and/or compressors. In order to assess this, further w ork needs to

be done on pressure drop optimisation at the targeting stage.

Pressure drop optimisation can also be extended to grass-roots design, in order to

develop a targeting procedure in which a three-way trade-off between area, energy

and power costs is possible.

Another possible improvement in the new retrofit procedure is to predict the

increased fou lin g due to the reduced velocities, and prevent that by building into the

analysis an allowance for increased fouling as velocity reduces. This analysis can

then be used along with the economic assessment o f pump replacement in order to

properly evaluate fouling costs, and pump replacement, in the targeting methodology.
 Publications Page 278
Appendix A

APPENBMA

PUBLICATIONS
Appendix A Pub lication s Pa ge 279

Contents

The follow in g papers have been written and published during the course o f this

research. Within them, the second on e has been recognized by the Institution o f

Chemical Engineers as the best paper in 1990 and awarded the Moulton Medal.

Copies o f these papers are attached.

1. Pol ley G.P. & Panjeh Shahi M .H .,

’ Process Integration Retrofit Subject to Pressure Drop Constraint’ ,

Process Technology Proceedings, 9, pp. 31-36, Elsevier, 1990.

2. P olley G.T. , Panjeh Shahi M .H . and Jegede F.O.,

’ Pressure Drop Considerations in the Retrofit o f Heat Exchanger N etw orks’ ,

Trans. I.Chem.E. 68A , pp. 211-220, 1990.

3. Polley G.T. , Panjeh Shahi M .H . and Nunez M .P.,

’ Rapid Design Algorithms fo r Shell-and-tube and Compact Heat Exchangers’ ,

Trans. I.Chem.E. 69A, pp. 435-444, 1991.

4. P olley G.T. & Panjeh Shahi M .H .,

’ Interfacing Heat Exchanger N etw ork Synthesis and Detailed Heat Exchanger

D esign’ ,

Trans. I.Chem.E. 69A, pp. 445-457, 1991.