Article ID: | iaor20123677 |
Volume: | 9 |
Issue: | 3 |
Start Page Number: | 352 |
End Page Number: | 381 |
Publication Date: | Apr 2012 |
Journal: | International Journal of Productivity and Quality Management |
Authors: | Ray Pradip Kumar, Mukherjee Indrajit |
Keywords: | heuristics: tabu search, heuristics: genetic algorithms |
A common problem in manufacturing industry is to optimise multistage multiple response processes. Stage‐wise non‐linear behaviour of process response(s), interdependency between the various stages and process constraint conditions at various stages make the optimisation problem more challenging for researchers and practitioners. In such situations, considering each stage in isolation (or treating independently) and thereby determining individual stage optimal conditions may lead to spurious sub‐optimal solution. Classical optimisation generally fails to provide exact optimal solution for a multi‐stage multi‐modal problem. Researchers generally prefer metaheuristics search and near‐optimal condition(s) for such type of problems. A popular integrated solution approach is to use empirical stage‐wise process models, composite desirability function and metaheuristic search strategy. Earlier research claims that a 'modified tabu search (MTS)' provides better solution quality than real‐valued genetic algorithm (RGA) for linear multi‐variate process models. In this paper, superior performance of MTS, as compared to RGA, is demonstrated, considering non‐linear process models (at each stage) and constraint conditions. This paper also demonstrates how stringent constraint conditions deteriorates the solution quality of RGA in a multi‐stage multiple response problem.