Research on fault recognition method of on-board equipment based on BP neural network optimized by Bayesian regularized
Cai Baigen, Yang Jiaming, Wei Shangguan, Bin Chen · 2017
As the core of guaranteeing the safety of high-speed railway traffics and improving transportation efficiency, Chinese Train Control System (CTCS) is the nerve center of railway transportation. As the key component of the CTCS, the on-board equipment is of great significance to diagnose its fault location quickly and effectively. At present, the fault diagnosis method based on artificial intelligence of on-board equipment is few, and mainly depends on artificial experience. Taking the data of CTCS-300T as the research objects, this paper presented a fault classification and recognition method for the on-board equipment based on Back-Propagation (BP) neural network model, and then used Bayesian regularization algorithm to optimize neural network model. The simulation results shows that the optimized model has more stable performance and higher generalization ability, and has obvious superiority compared with the BP network. And the accuracy of the unknown samples' classification is improved significantly.