Main Converter Fault Diagnosis for Power Locomotive Based on PSO-BP Neural Networks

Abstract

To aim at conventional BP learning algorithm of its flaws, say, low convergence speed and easy falling into local extremum, and etc, during main converter fault diagnosis system for power locomotive, this paper proposed a novel learning algorithm called PSO-BP neural networks based on particle swarm optimization (PSO) and BP neural networks. The algorithm generated the two phases: one is that PSO was applied to optimize the weight values of neural networks based on training samples, the other is that BP algorithm was applied to farther optimize based on verifying samples till the best weight values are achieved. Eventually, a practical example indicates that the proposed algorithm has quick convergence speed and high accuracy, and is ideal patter classifier.