Reference point based evolutionary multi-objective optimization with dynamic resampling for production systems improvement
| Pages | 489-512 |
| Date | 12 November 2018 |
| Published date | 12 November 2018 |
| DOI | https://doi.org/10.1108/JSIT-10-2017-0084 |
| Author | Amos H.C. Ng,Florian Siegmund,Kalyanmoy Deb |
| Subject Matter | Information & knowledge management,Information systems,Information & communications technology |
Reference point based
evolutionary multi-objective
optimization with dynamic
resampling for production
systems improvement
Amos H.C. Ng and Florian Siegmund
University of Skövde, Skövde, Sweden, and
Kalyanmoy Deb
Michigan State University, East Lansing, Michigan, USA
Abstract
Purpose –Stochastic simulation is a popular tool among practitioners and researchers alike for
quantitative analysis of systems. Recent advancement in research on formulating production systems
improvement problems into multi-objective optimizations has provided the possibility to predict the
optimal trade-offs between improvement costs and system performance, before making the final
decision for implementation. However, the fact that stochastic simulations rely on running a large
number of replications to cope with the randomness and obtain some accurate statistical estimates of
the system outputs, has posed a serious issue for using this kind of multi-objective optimization in
practice, especially with complex models. Therefore, the purpose of this study is to investigate the
performance enhancements of a reference point based evolutionary multi-objective optimization
algorithm in practical production systems improvement problems, when combined with various
dynamic re-sampling mechanisms.
Design/methodology/approach –Many algorithms consider the preferences of decision makers
to converge to optimal trade-off solutions faster. There also exist advanced dynamic resampling
procedures to avoid wasting a multitude of simulation replications to non-optimal solutions.
However, very few attempts have been made to study the advantages of combining these two
approaches to further enhance the performance of computationally expensive optimizations for
complex production systems. Therefore, this paper proposes some combinations of preference-
based guided search with dynamic resampling mechanisms into an evolutionary multi-objective
optimization algorithm to lower both the computational cost in re-sampling and the total number of
simulation evaluations.
Findings –This paper shows the performance enhancements of the reference-point based algorithm,
R-NSGA-II, when augmented with three different dynamic resampling mechanisms with increasing
degrees of statistical sophistication, namely, time-based, distance-rank and optimal computing buffer
allocation, when applied to two real-world production system improvement studies. The results have
shown that the more stochasticity that the simulation models exert, the more the statistically advanced
dynamic resampling mechanisms could significantly enhance the performance of the optimization
process.
This work was partially financed by the Knowledge Foundation (KKS), Sweden, through the Blixt-
Sim project (2011-2014). The authors gratefully acknowledge their provision of research funding and
thank the industrial partners, Volvo Car Corporation and AB Volvo, for their collaborative supports
during the project.
Production
systems
improvement
489
Received4 October 2017
Revised17 February 2018
Accepted21 February 2018
Journalof Systems and
InformationTechnology
Vol.20 No. 4, 2018
pp. 489-512
© Emerald Publishing Limited
1328-7265
DOI 10.1108/JSIT-10-2017-0084
The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/1328-7265.htm
Originality/value –Contributions of this paper include combining decision makers’preferences and
dynamic resamplingprocedures; performance evaluationson two real-world production system improvement
studiesand illustrating statistically advanceddynamic resampling mechanism is needed fornoisy models.
Keywords Multi-criteria decision making, Multi-objective optimization, Dynamic resampling,
Production systems improvement
Paper type Research paper
1. Introduction
Real-world optimization problems very often involve multiple objectives that have to be
considered simultaneously. In terms of production systems improvement, there are almost
always at leasttwo objectives to be considered –the targetedcondition (Rother, 2009)and the
cost of the improvement. As in many real-world multi-objective problems (MOP), it is
obvious that these two objectives are in conflict with each other, so that an improvement in
one objectivecan only be obtained at the expense of degradationof the other objective. MOPs
can be readily solvedby a priori approaches which transformthe problems into some single-
objective ones by formulating the objective functions into single, weighted-sum functions.
However, this is usually not applicable if the decision maker (DM) does not explicitly know
how to weigh the variousobjectives before any optimal alternatives are known. At the same
time, it is not easy to understand the effects of the weights, in terms of correlation and
nonlinear outcomes, meaning that a small change in weights can alter the solution
dramatically. In contrast to any a priori approaches in which the DM has to explicitly
determine their preference regarding the objectives before the optimization process, a
posteriori approaches aim to find the entireset of best “trade-off”, or so-called Pare to-optimal
solutions, so that theDM can decide which solution to implement afterthe optimization has
been completed. The goal of a posteriori optimization approach is therefore to find a
converged set of solutions that also feature widediversity, to spread as much as possible in
the objectivespace and form an optimal trade-offcurve/surface, or so-calledefficient frontier
(EF). This approach is desirable, as it allows the DM to obtain a complete picture about the
problem under study, e.g. the relationship between the decision variables and theobjectives
(Bandaru et al.,2017) and provides the DM with many alternativesto choose from. However,
it becomes a critical issue if the optimization process involves computationally expensive
functionevaluations, e.g. stochasticsimulation runs on large-scale,complex models.
Regarding production systemsanalysis, stochastic simulation is not only a popular tool
for the evaluations of long, complex and real-world production systems, but probably the
only feasible choice, especially when the processing times and downtimes follow non-
exponential or non-normal distributions (Negahban and Smith, 2014). Stochasticsimulation
is the only available choice for researchers and practitioners in industry alike if more
complex flows and other types of variability (e.g. setups) are included in the study of
unbalanced production lines. As claimed by Tempelmeier (2003),“If quantitative
performance evaluation is carried out at all (in the industry), then in almost any case
simulation is the only tool used.”Wang and Chatwin (2005) summarize the three major
weaknesses of analytical/mathematical methodologies, compared to computer simulation:
analytical evaluation is impractical when it encounters stochastic elements, such as many
random and non-linear operations that exist in virtuallyany manufacturing system; due to
randomness in a dynamic system which changeswith time, the mathematical modeling of a
complex dynamic system requires many simplifications which may cause the models to
become invalid; analyticalmethods are not sufficient for optimization because mathematical
models can only be built with simplifying assumptions that may affect the accuracy of the
JSIT
20,4
490
To continue reading
Request your trial