Abstract: Particle swarm optimization(PSO)is simple in theory, quick in convergence, but likely to be "premature" at the initial stage. Genetic algorithm (GA) has strong global search ability but the convergence accuracy is low. Considering both the advantages and disadvantages, the structure and the critical parameters are analyzed in this paper, genetic operators and the crossing-search methods are applied to PSO algorithm to avoid falling into locally optimal solution. In this process, inertial weight and mutation methods are improved to balance the global and the local search ability. At the same time, some swarms are mutated if the swarm population have evolved to an enough small space. And then, the novel algorithm could get higher convergence accuracy and executive capability to solve non-linear and multi-extremum in the application of the engineering field. According to the results of comparisons with other algorithms through varieties of test functions, the hybrid algorithm combining PSO and GA shows great advantages in solution accuracy, search efficiency and the ability to process different functions, and meets the engineering needs.