Based on the advantages of particle swarm optimization(PSO) that was easily to be realized, high precision and fast convergence, combining PSO and blind source separation(BSS), a new method of mechanical failure blind source separation based on PSO was proposed(PSO-BSS), and the method was applied to rolling bearing compound fault diagnosis. In the PSO-BSS method, the sum of the absolute value of the kurtosis was taken as the target function,the maximum of the target function was sought by PSO,then the optimal separation matrix was determined.The simulation results show that the PSO-BSS method can be used to separate the fault features of multipile compound fault signal and is significantly superior to the traditional mechanical failure blind source separation based on genetic algorithm (GA-BSS) in separation performance, algorithm convergence and operation speed. Finally, the PSO-BSS method was successfully applied to the actual rolling bearing inner and outer mixing fault blind source separation, a good result was obtained which validated its engineering practicality.