Fault diagnosis for locomotive bearings based on IPSO-BP neural network

This paper presents a BP network model based on improved PSO for bearing fault diagnosis. Combining PSO algorithm for global optimization ability with BP neural network advantages of local search, the model effectively prevents the network from a local minimum, and at the same time guarantees the accuracy of diagnosis. Simulation results show that the locomotive bearings have been effectively diagnosed. Compared with the conventional BP neural network model, this method not only improves the convergence speed, but also improves the fault diagnosis accuracy.