Multiple-objective optimization of machine cell layout using genetic algorithms

A prototype genetic algorithm-based system has been developed to group machines into manufacturing cells, subject to objective functions such as gross part (inter- and intra-cell) movement, cell load variation and machine set-up costs. The system is an improvement over two contemporary techniques because it can handle complex machine groupings with multiple objective functions. In order to assess chromosomes from different populations, a new crossover technique is proposed. The prototype system was evaluated using four machine groupings, namely Models 1, 2, 3 and 4 and three Approaches involving three different objective (fitness) functions. Models 1 and 2 have the twin objective functions of minimizing gross part movement and cell load variation. Models 3 and 4 are Models 1 and 2, respectively, with the minimizing machine set-up cost as the third objective function. In Approach 1, the chromosomes were drawn from two distinct populations, while Approaches 2 and 3 were defined by the authors to handle 3 or more objective functions. A chromosomal performance index, Cobj, is defined to evaluate the fitness values of Approaches 2 and 3. It was found that Approach 3 was consistently able to produce sub-optimal solutions in all four models after weights are applied to different objective functions, whereas Approach 2 can perform as well as Approach 3 only if there is a sufficient number of iterations. On the other hand, Approach 2 requires a much shorter computational run time than Approach 3.