Optimizing Flow Shop Sequencing Through Simulation Optimization Using Evolutionary Methods

This paper describes an approach to use an adapted Evolution Strategies (ES) algorithm to generate improved sequences for producing unique parts in a flow shop. The algorithm uses principles from both genetic algorithms and Evolution Strategies. While several alternative algorithms were considered, the focus of this paper is on the one that performed the best for this problem domain. The best algorithm is an ES that implements a new mapping technique (Genotype-Phenotype) to convert its real-valued gene representation into a valid job sequence. The approach also uses production heuristics to generate the initial set of sequences, thus providing a better starting point and accelerating the optimization process. The fitness of each sequence generated by the algorithm is evaluated by a discrete-event simulation model of the flow shop. The algorithm and simulation model are a part of a decisionsupport system that was developed to optimize ship panel construction at Northrop Grumman Ship Systems. A design-of-experiments approach is used to configure the efficiency of the algorithm for different problem sizes. This analysis helps select the optimal set of parameters. This paper includes details of routines and provides results from various optimization runs.