Evolutionary algorithms for production planning problems with setup decisions

Production planning problems with setup decisions, which were formulated as mixed integer program (MIP), are solved in this study. The integer component of the MIP solution is determined by three evolution algorithms used in this study. Firstly, a traditional genetic algorithm (GA) uses conventional crossover and mutation operators for generating new chromosomes (solutions). Secondly, a modified GA uses not only the conventional operators but also a sibling operator, which stochastically produces new chromosomes from old ones using the sensitivity information of an associated linear program. Thirdly, a sibling evolution algorithm uses only the sibling operator to reproduce. Based on the experiments done in this study, the sibling evolution algorithm performs the best among all the algorithms used in this study.