Investigation of Evolutionary Algorithm Effectiveness in Optimal Synthesis of Certain Mechanism

Abstract

This paper presents the comparison of four evolutionary algorithms (EA) in terms of their effectiveness as applied to solve an engineering optimization task. The four tested algorithms differed in structure. The problem to be solved concerned optimization of a four-bar linkage dimensions so that its movement follows a given trajectory. Best EA parameters values were sought, namely: crossover probability, mutation probability and pool size. The main EA evaluation criterion was the number of iterations necessary to set the size of a mechanism that would realize the trajectory with a given tolerance. Substantial dispersion of results obtained from the algorithms after multiple repetitions of calculations were observed. For that reason the second criterion was introduced, a statistical measure of iterations number dispersion. The combined criteria, the number of iterations and standard deviation, enabled different EAs effectiveness comparison and indicated proper selection of their parameters values.