LAUSR

Abstract Due to less parameters and simple operations in PSO, PSO has attracted the attention of many researchers. However, it amy fall into local optimum and the search precision is not high. Therefore, this paper introduces an improved PSO (CNPSO). There are two new formulas: (1) when the individual is not the best particle, the gbest is replaced by a particle, which is selected from a set based on the probability calculation. Such mechanism can help the algorithm to escape local position. (2) when the individual is the best particle, it is combined with a randomly selected particle to generate a convex combination, after that opposite learning is adopted to get a reverse solution. This operation can maintain the diversity of population. Finally, CNPSO is compared with several algorithms in three experiments, and used to optimize the spring design problem. The results indicate CNPSO has good performance and high search precision.