Multiobjective batch plant design: A two-stage methodology. 2. Development of a genetic algorithm and result analysis

This second part of this series of papers deals with the second stage of the methodology developed for multiobjective batch plant design, that is, the implementation of a dedicated genetic algorithm. This procedure can be viewed as the search engine for workshop configurations and uses the simulation model as a subroutine for evaluating the generated structure feasibility. First, the basic principles of genetic algorithms (GAs) are briefly recalled; second, the GA, especially developed for treating batch plant design, is largely presented. An interesting concept has been introduced, that is, the so-called gene fridge to prevent population degradation by allowing the introduction of previous genes during evolution. Besides, this procedure allows a better screening of search space and the evaluation of the problem combinatorics. Finally, this concept drives the search toward critical steps, by introducing a varying mutation probability function of the gene locus. An illustrative example, already presented in part 1 of this series, is largely analyzed and provides useful guidelines for treating similar problems. The parametric study is particularly interesting for GA parameter setting, which constitutes a key problem.