Integrating Evolution Strategies into Genetic Algorithms with Fuzzy Inference Evaluation to Solve a Steelmaking and Continuous Casting Scheduling Problem

Abstract

This contribution presents a metaheuristic approach that integrates evolution strategies into genetic algorithms using a fuzzy rule based inference system to evaluate schedules in a generalized steelmaking and continuous casting production system. The genetic algorithm controls the job sequences assigned to the machines while the setting of jobs initial processing dates at the converter are optimize by means of evolution strategies. The fuzzy inference system gives an overall evaluation of the schedule quality by controlling discontinuities and transit times with different degrees of acceptance throughout the evolution process. This approach integrates an embedded search procedure to overcome one of the weaknesses of metaheuristic scheduling methods of setting initial dates for task processing and is especially suited for highly nonlinear objective functions as in this case. A general structure of the steelmaking and continuous casting production system is consider with an arbitrary number of machines at each stage, with production of several steel grades and types (e.g. slabs and billets). Technological constraints such as continuous casting between jobs (batches) and in process time of liquid steel are included. For illustration purposes, a real sized problem is solve.