A modified comprehensive learning particle swarm optimizer and its application in cylindricity error evaluation problem

Abstract

Particle swarm optimizer was proposed in 1995, and since then, it has become an extremely popular swarm intelligent algorithm with widespread applications. Many modified versions of it have been developed, in which, comprehensive learning particle swarm optimizer is a very powerful one. In order to enhance its performance further, a local search based on Latin hypercube sampling is combined with it in this work. Due to that a hypercube should become smaller and smaller for better local search ability during the search process, a control method is designed to set the size of the hypercube. Via numerical experiments, it can be observed that the comprehensive learning particle swarm optimizer with the local search based on Latin hypercube sampling has a strong ability on both global and local search. The hybrid algorithm is applied in cylindricity error evaluation problem and it outperforms several other algorithms.