Abstract
In this paper, a new method to improve the performance of particle swarm optimization is proposed. The proposed method applies hypothesis testing to determine whether the particles trap into the local minimum or not. When the difference of means of two samples is not significant using hypothesis testing, the particles can be regarded as trapped into the local minima, the particles will be re-initialized and the global best experience is reserved. Several famous benchmarks are used to show the efficiency of the proposed technique. Moreover, optimisation results for three engineering optimisation problems with linear and nonlinear constraints demonstrate that the proposed method can effectively enhance the searching quality and stability.