Analyzing Production Process Performance through Evolutionary Multiobjective Optimization

Nature-inspired computational techniques are nowadays being developed for and employed in various application domains. Evolutionary algorithms, known as general and robust optimizers, have recently been extended to deal with multiobjective optimization where search for the best among candidate solutions is performed not according to one, but multiple, usually conflicting objectives. In this paper we show how evolutionary multiobjective optimization can be used in parameter tuning and performance analysis of a metallurgical production process where several empirical criteria are applied to ensure the process safety and product quality. The paper introduces the concept of multiobjective optimization and the applied algorithm, describes the considered task of process parameter tuning and the performed numerical experiments, and discusses the obtained results. In contrast to single-objective optimization, the multiobjective approach enables a much better insight into the process performance.