Construction of Fault Diagnosis Model of Metro Wheel Speed Box System Based on Convolution Neural Network

This research focuses on a fault detection method apply on the subway wheelsets. This fault diagnosis method is mainly aimed at the vibration signal to distinguish the fault and abnormal situation. At the same time, the vibration signal incorporated with the deep learning method to establish the diagnosis mechanism. Reliable online data of subway in a city of China and CNN (Convolutional Neural Network) are applied in the web training process. The degradation of vibration signal of rolling functional unit (wheelset) was summarized, and the difference between normal and fault signals of rolling functional unit (wheelset) was studied. The feedback learning mechanism makes it possible to update the neural network in real time.