In order to balance local and global search ability of particle swarm optimization algorithm, a particle swarm optimization algorithm with multi-adaptive strategies (MAS-PSO) has been proposed. In the process of particle evolution, the algorithm adopted adaptive velocity setting strategies which were based on the evolution degree of particles and local opening chaotic search. The MAS-PSO is applied to BP neural network training of analog circuit fault diagnosis, and it solved effectively the problems of slow network convergence rate in conventional BP algorithm and easily falling into partial minimum. The simulation results show it works with quicker convergence rate and higher forecast precision.