Application of Particle Swarm Optimization to the mixed discrete non-linear problems

Abstract

Particle Swarm Optimization is applied to the mixed discrete non-linear problems (MDNLP). PSO is mainly a method to find a global or quasiminimum for a non-convex optimization problem of continuous design variables. To handle the discrete design variables, penalty function is introduced. By using penalty function, it is possible to treat all design variables as the continuous design variables. Through typical structural optimization problem, the validity of proposed approach for MDNLP is examined. © 2005 by International Federation for Information Processing.