A multi-objective heuristic-based hybrid genetic algorithm

摘要

Most attempts by researchers to improve upon multi-objective genetic algorithms (MOGAs) involve different implementations of the traditional genetic algorithm operations (i.e., fitness assignment, crossover, and mutation). However, adherence to models that remain consistent with evolutionary theory may be stifling further performance gains that could be realized through the development of hybrid algorithms. This paper presents a hybrid MOGA that combines a baseline MOGA with heuristics specifically tailored to address deficiencies often encountered in multi-objective optimization. The new hybrid technique is compared to the baseline MOGA in its application to a two-objective two-bar truss design and a three-objective packaging design of a power electronic module. The comparison is aided by four recently developed metrics that provide a balanced quantitative measurement of performance. The new technique was shown to consistently outperform the baseline MOGA for the application examples.