Sciencedirect: A Lean Assessment Tool Based On Systems Dynamics
Sciencedirect: A Lean Assessment Tool Based On Systems Dynamics
Sciencedirect: A Lean Assessment Tool Based On Systems Dynamics
com
ScienceDirect
Procedia CIRP 50 (2016) 106 – 111
Abstract
Lean manufacturing is synonymous with a set of practices used in the identification and elimination of waste related with the manufacturing
system, and focusing on what creates value for the customer. Lean assessment tools enable an overall audit of the performance of lean
practices, and so are able to identify lean improvements. The interactions between lean practices and their improvements are often latent and
need to be investigated: a systems approach can be used to disclose these hidden interactions. In this article, system dynamics is used as a lean
assessment tool to assess and improve lean performance for a print packaging manufacturing system.
©2016
© 2016 TheThe Authors.
Authors. Published
Published byB.V.
by Elsevier Elsevier
This is B.V.
an open access article under the CC BY-NC-ND license
Selection and peer-review under responsibility of Professor Lihui Wang.
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organizing committee of the 26th CIRP Design Conference
Keywords: Lean manufacturing; lean assessment tool; system dynamics
1.Introduction LATs also exist such as Value Stream Mapping (VSM) and
many others [3].
Lean manufacturing (LM) as a set of practices, tools and Lean practices affect and are affected by other practices.
techniques is centered around five core principles namely: The case of routine maintenance is an example. Routine
precisely determining the value of each specific product in the maintenance is an arm of TPM and can be described as the
eyes of the customer; identifying the value flow of each type of maintenance undertaken daily by operators on their
product; making the value flow continuously; letting the machines. Such maintenance include cleaning, lubricating and
customer pull value from the manufacturer and seeking inspecting of machines. Increasing the number of routine
perfection [1]. The intended end-result is improved maintenance should naturally have a positive effect by
organizational performance through customer value enhancement reducing the frequency of machine breakdowns. However,
as a result of waste (non-value adding activities) elimination. carrying out routine maintenance results in the machine being
Representative practices relating to LM include just in time non-operational during the checks, thereby increasing machine
(JIT) management, total productive maintenance (TPM), downtime and delaying manufacturing cycle time. This and
employee involvement, continuous improvement, set-up many other interactions occur between improvement lean
reduction, customer engagement and many others. practices.
For an organization to enjoy the multiple benefits of LM, The non-consideration of the interrelations between lean
the practices associated with it need to be implemented practices has made LATs to be less than optimal [4]. Few
holistically [2]. A lean assessment tool (LAT) is typically used researchers have considered this relationship [4,5,6] in their
to audit, in a simultaneous fashion, the performance of all the LAT. However, they have done so using subjective-based
lean practices that are relevant to the type of organization. approaches such as Analytical Hierarchy Process (AHP) [4],
Many LATs have been developed and validated. Analytical Network Process (ANP) [5] and BSC [6].
Qualitative questionnaire-type models such as the Lean Subjective based approaches have the flaw of bias, and a
Enterprise Self-assessment tool (LESAT), Balanced strategic and long-term view based on small incremental lean
Scorecard, European Foundation for Quality Management improvements cannot be achieved. An objective-based
(EFQM), Malcolm Baldridge Model, and the Shingo Model approach has previously been attempted using discrete event
[3] have been used as the basis for a LAT. Quantitative based simulation (DES) [7], but focused mainly on tangible aspects
of LM such as reduction in setup, defect rates and lead-time.
2212-8271 © 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organizing committee of the 26th CIRP Design Conference
doi:10.1016/j.procir.2016.04.169
Oleghe Omogbai and Konstantinos Salonitis / Procedia CIRP 50 (2016) 106 – 111 107
In the current paper, a System Dynamics (SD) based LAT primary objective of the proposed lean improvements. SD is
incorporating both tangible and behavioral aspects of LM is used to articulate the problem in a dynamic way so that the
proposed. It is used to objectively investigate the dynamic organization can validate proposed lean improvements as well
interactions between lean practices, their performance as study the interactions between them.
outcomes and other system variables. The approach is further
applied to generate an optimized setting of lean improvements 3.1.Stock and flow diagram
that are needed to minimize manufacturing lead-time, using an
optimization add-on tool of the SD simulation software. Activities and outcomes relating to three lean practices
have been chosen: Total Productive Maintenance (TPM),
2.Systems dynamics modeling approach Quality Management (QM) and Employee Morale. Although
many others can be investigated alongside these, the three
SD is “a perspective and set of conceptual tools that enable chosen are pertinent to the problem and are sufficient enough
us to understand the structure and dynamics of complex to illustrate the intended approach.
systems” [8]. It uses simulation to investigate how a system An SFD (Fig. 1) is first developed for the case. AnyLogic
will respond, dynamically, to a set of changes [9]. 7.2 SD Software was used for the study. The problem is
The use of SD has been well researched within modeled around two main stocks: job order backlog and
manufacturing systems and quite a number are lean-related. defects. Job order backlog is the outstanding work that is
SD has been used to control the cost of quality [9], to replenished with new order through job entry rate and
investigate the performance of a lean cell under uncertainty depleted through throughput. Defects cannot be depleted and
[10], to improve the performance of a foundry operation [11] are accumulated through defect rate.
and as a productivity improvement tool in a print shop [12]. It The number of job orders and the time between orders
has also been used as a dynamic LAT for takt time influences Job order entry rate. When the number of job
improvement [13]. In the current study, SD is used to examine orders increases, the variety of WIP increases because each
multiple aspects of lean as well as their interactions. job order is unique and is typical of MTO systems. If the
SD is a modeling technique, consisting of a stock and flow variety of WIP increases, there are more setups and
diagram (SFD). The SFD is a causal loop diagram, which changeovers than when few and similar jobs are in constant
maps the essential variables of the system under review (lean production. When a machine is being setup and changed over
practices in the current article) and the causal influences for a job, the machine is not processing, but stalled. The more
between them. Stocks are accumulations, expressed in there are setups and changeovers in the system, the more
quantities that characterize the system [14], for example machines are stalled and this increases the production idle
inventory and works-in-process (WIP). Flows are rates, time.
typically in quantities over a specified time, which deplete or When there is an increase in total job orders the pressure
replenish the “stock” level, such as shipment rate and to maintain customer specified lead-time is increased. Lead-
production rate respectively for a manufacturing system. time pressure affects the routine maintenance as management
An SFD on its own cannot be simulated. It needs a set of aims to minimize downtime during routine services. In
governing equations that describe the various causal addition, work hours are increased to ensure lead-time is not
relationships [15,16]. Subsequently, the SFD can be used for adversely affected, but this is to a maximum of 11 hours per
scenario analysis, optimization, and other simulation day (work hours are flexible for the plant ranging between 8
analysis/applications. Several guidelines to using SD can be and 11-hour days depending on work load). With increased
found in [8]. In the current paper, a manufacturing case study work hours, employee fatigue sets in, which reduces the
is used for a better illustration of the SD modeling approach number of routine quality checks and productivity through
for lean assessment. employee output per time. If employee output drops, then
throughput also drops. Various process errors are generated
3.Case study illustration whenever employees are fatigued for example mistakes with
mixing printing inks.
The production operations of a print packaging When the stock of defects increases, the plant over
manufacturer have been used as a case study. The produces to account for increased defects in the process. This
organization has been implementing lean practices for a few is anti-lean but inevitable since the plant is expected to meet
years and seeks to investigate the interactions between with the specified job order quantity. However, the company
proposed lean improvements. The prior knowledge about seeks ways to eliminate this as part of their lean
these interactions is needed to design an optimal set of lean transformation. The overproduction for defects affects the
improvements for the company. needed throughput to ensure job order completion is on target.
The printing industry is a make-to-order (MTO) production Other lean practices and system variables are depicted in the
system with custom products [12]. Top on the list of SFD (Fig. 1).
customer-specified values for the print industry is Usually each arrow in the SFD is denoted by a “+” or “-”
dependability [17]. On-time or before-time deliveries makeup characterizing if the “cause” and “effect” variables change in
a dependable print packaging supplier. The primary concern the same direction or not. For the purpose of de-cluttering the
of the company is to meet up with customer delivery time. SFD, these signs are sometimes omitted.
Lead-time minimization has therefore been chosen as the
108 Oleghe Omogbai and Konstantinos Salonitis / Procedia CIRP 50 (2016) 106 – 111
The main aim of this article is to generate a SD model for In this section the model is simulated under various
lean assessment and verification/validation of SFD achieves extreme conditions to investigate the interactions between
this. After inputting the governing equations, the model is lean practices and their performances. Altering the values of
then run and verified against the real life dynamics of the case parameter variables in the model achieved this. Parameter
organization using the reference mode. The reference mode is variables are not affected by other variables and are used to
a pattern of system behavior over time [15]. In the current adjust model behavior. The authors of the present article set
analysis, the reference mode supports the hypothesis that out to alter the values of parameter variables in the model as
variations exist within the system performance parameters listed in Table 2.
when demand fluctuates between high and low values. The The values presented in Table 2 are the extreme best values
SD model is run for one year under current conditions. Fig. 3 for the case study under analysis. For example normal routine
(a-d) shows the model run results for lead-time (Customer maintenance for the current system is uniformly distributed
Delivery Performance), defect rate (TQM), employee fatigue (0.017, 0.05) i.e. routine maintenance is done once randomly
(Employee Morale) and machine efficiency (TPM) and anywhere from 20 days (0.017) to 58 days (0.05). For
verifies the variations in lean performance in accordance with Experiment 1, the test value is set at 0.02 i.e. routine
demand loads. maintenance is fixed at once every five days.
The graphs in Fig. 3 are indicative of the lean Figures 4-6 are the simulation results for the three
performances for the system for one year under current experiments. The simulation results show marginal
conditions, and assuming no changes to lean parameters. For interactions between setup time reduction, reduction in
example, Employee Morale (Fig. 3c) ranges between 0.7 and process related errors and improvements in routine
0.85 on an increasing scale of 0 to 1, where 1 represents the maintenance.
highest level of employee morale. Machine efficiency is a
typical measure for TPM and ranges between 40% and 60%
(Fig. 3d).
Lead-time
Variable Equation
job order entry rate number of job orders / time between orders time
Backlog)/dt value=25)
variety of WIP total job orders / variety coefficient
overproduction for
defect rate
defects (d) TPM performance
d(Defects)/dt defect rate (initial value=0.001)
Fig. 3. Current state lean performances for (a) lead-time, (b) TQM, (c)
quality checks Normal quality checks / employee fatigue
Employee Morale, (d) TPM
110 Oleghe Omogbai and Konstantinos Salonitis / Procedia CIRP 50 (2016) 106 – 111
5.Optimization experiments to achieve lean objective fatigue. In reality, SD models are perfected over extended
periods to capture more of the causal relationships within the
An optimization experiment was undertaken to establish system. The case study organization has used the SD model
the optimal set of lean parameters to minimize lead-time. generated in this article as a base model for their lean
Table 3 summarizes the parameter values for the optimization assessment studies. The model is also re-useable in other
experiment and represents values for current system similar instances.
parameter variables. The SD model was run with the The SD model was used to validate lean improvements.
optimization add-on function of the AnyLogic software, and For example, when the SD model is simulated with routine
the results are presented in Fig. 7 for 250 iterations. The maintenance set at 0.2 days as opposed to the current value
optimization results did not change beyond this number of that is uniformly distributed (0.017, 0.05), lead-time was
iterations. shown to improve by approximately 27%. This implies that if
the number of routine maintenance is increased, lead-time
Table 2. Parameter values for simulation experiment with the SD model reduces.
The optimization experiment (section 5) was used to
Exp.* Exp. Description of generate an optimal set of values for the parameter variables,
Variable Current Value
No. Value experiment values
needed to minimize lead-time. Results of the optimization
normal Time between experiment, Fig. 7, indicate that the best lead-time of 1.42days
Uniform 0.2
1 routine maintenance is once is achieved when variety coefficient= 9, average setup time=
(0.017, 0.05) days
maintenance every five days 0.004, normal machine efficiency= 0.735, normal routine
maintenance= 0.03, normal process errors= 0.233 and normal
normal 2 % of orders have
2 process
Uniform
0.02 process related quality checks= 0.174. These values are based on the ranges
(0.15, 0.25)
related errors errors specified for the optimization experiment (Table 3). Altering
the parameter ranges as well as changing the optimization
average setup Uniform 0.003 Average setup time
3
time (0.0042, 0.0083) days in day units
objective function can be used to configure other optimization
experiments.
*Experiment
A significant contribution of the current study is that
Table 3. Parameter values for lean optimization experiment “hard” and “soft” aspects of LM can be assessed and
improved in tandem in one LAT. This kind of lean assessment
Value is lacking [18, 19]. For example, measuring Employee Morale
Parameter Type is often done using questionnaire-based, subjective-type lean
Min. Max.
self-assessment tools, while setup and changeovers are
variety coefficient discrete 3 9 tracked using quantitative based LATs such as the VSM. With
average setup time continuous 0.004 0.008 the SD modeling approach both qualitative and quantitative
normal machine efficiency continuous 0.7 0.8 measures of LM can be accumulated under the same
normal routine maintenance continuous 0.017 0.05 assessment.
normal process related errors continuous 0.15 0.25
normal quality checks continuous 0.1 0.2
Machine efficiency
Machine efficiency
In the current article, a SD based lean assessment tool was
generated. The SD modeling approach enabled the validation
Time
of proposed lean improvements as well as the analysis of the
inter-relationships between lean variables. The model was
also used to establish an optimal set of values for lean
Defects rate
Time
References
[1] Womack JP, Jones DT. Lean thinking: banish waste and create wealth in
Fig. 5. Simulation results for Experiment 2 with process related errors = 0.02 your corporation. 2nd ed. New York: Free Press; 2003.
[2] Bonovia T, Marin JA. An empirical study of lean production in the
ceramic tile industry in Spain. Int J Prod Op Manage 2006;26:505-531.
[3] Oleghe O, Salonitis K. Leanness assessment frameworks: strengths and
weaknesses. Submitted for publication to Journal of Manufacturing
Machine efficiency
Manage 2013;30:751–64.
[10] Deif A. Dynamic analysis of a lean cell under uncertainty. Int J Prod Res
2012;50:1127–39.
[11] Rodrigues LLR, Hebbar S, Shetty D, Hoskote RN. Performance
improvement of foundry through lean methodology: a modeling and
Time simulation approach. IEEE Computer Society. 7th Asia Modeling
Symposium. 2013.
[12] Uribe J. Print Productivity: a system dynamics approach. A Research
Fig. 6. Simulation results for Experiment 3 with average setup time= 0.003 Monograph of the Printing Industry Center at Rochester Institure of
Technology;2008:5
[13] Ali RM, Deif AM. Dynamic lean assessment for takt time
implementation. Procedia CIRP 17 2014:577-581.
[14] Grigoryev I. AnyLogic 7 in three days: a quick course in simulation
modeling. 2015. Available at http://www.anylogic.com/free-simulation-
book-and-modeling-tutorials
[15] Vensim Version 6.3E. Introduction and Tutorial.
[16] Rodrigues LLR, Dharmaraj N. System dynamics approach for change
management in new product development. Manage Res News 2006;
29:512-523
[17] Pellow B, Sorce P, Frey F, Oslon L, Moore K, Kirpichenko S. The
advertising agency’s role in marketing communications demand creation.
Printing Industry Center at Rochester Institute of Technology. 2003:5
[18] Oleghe OA, Salonitis K. Improving the Efficacy of the Lean Index
through the Quantification of Qualitative Lean Metrics. Procedia CIRP
2015; 37:42-47
[19] Oleghe OA, Salonitis K. Variation Modeling of Lean Manufacturing
Performance Using Fuzzy Logic Based Quantitative Lean Index. Procedia
Fig. 7. Snapshot of results of optimization experiment for lead-time
CIPR 2016; 41:608-613
minimization