BP Neural Network Based Fault Diagnosis in Vehicle Braking Control System

Abstract

W232 the popularity of the high-speed railway in China, its operation safety issues has been attracting more and more attention nowadays. How to improve the high-speed railway reliability has become an emerging research focus. This paper investigated the fault diagnosis in vehicle braking control system, which is a core subsystem of EMU (electric multiple unit). To diagnosis the sensor fault in braking control system, back propagation (BP) neural network based method with two different learning approaches were utilized and compared. Test results based on raw data collected from EMU experimental platform showed that both approaches can accurately diagnose the faults, while the moment based learning approach provided a faster outcome compared with conventional gradient descent approach.