(Consideration on knowledge acquisition from optimization result of genetic programming)
Shinji KYOTO, Tatsuya KYOTO, Kazuhiro Kyoto, Takayuki KYOTO · The Proceedings of Mechanical Engineering Congress Japan · 2019
By devising the production process of industrial products, it will greatly affect the shortening of production time and cost reduction. There is a job shop shceduling which is one of production method. In JSSP(Job Shop Scheduling Problems), machine assignment and job processing order greatly affect production efficiency. In this study, we used dispatching rules. We used GP(genetic programming) to express the priority function of DRs(Dispatching Rules). Moreover, when considering a production system, it is necessary to optimize for multiple purposes. In this study, multi-objective optimization was performed using NSGA-II(Non dominated Sorting Genetic Algorithm) calculate Pareto optimal solutions. Furthermore, clustering was performed using k-means in order to consider Pareto optimal solutions. The distance in k-means is defined in the objective function domain.