Performance Comparison of Variants Based on Swarm Intelligence Algorithm of Mathematical and Structural Optimization

In this paper, three different variants of the canonical particle swarm optimization (PSO) algorithm are compared in terms of their optimal solutions to engineering optimization problems. The PSO often face to premature convergence due to its weak capability in local exploitation, which leads to a low optimization precision or even failure. Moreover, the computational burden is also one of the obstacles for the PSO when solving complex problems. To overcome these drawbacks, the original PSO algorithm is integrated with the knowledge sharing and auxiliary mechanisms of comprehensive learning (CL) and Gaussian local search (GLS) strategies to form the so-called GLS-PSO, CLPSO, and GLS-CLPSO algorithms. To demonstrate the accuracy and robustness of the three variants, their performance is tested on a benchmark spatial truss structure with continuous variables. The obtained optimization results are then compared with those by other PSO variants in the literature.