Fault Diagnosis of the Train Speed and Distance Measurement Equipment Based on Rough Set Theory

Abstract:
The Train Speed and Distance Measuring Equipment is very important during train operation and is one of the equipment with a high failure rate. However, in actual operation and maintenance, the fault detection of Speed and Distance Measurement Equipment only depends on the experience of field operators, and the efficiency of fault diagnosis relatively low. Therefore, this paper uses Rough Set Theory (RS) to optimize Back Propagation (BP) Neural Network, and intelligently diagnoses the fault of Train Speed and Distance Measuring Equipment. This paper summarizes the fault types of Train Speed and Distance Measuring Equipment based on the fault case database. First, the RS is used to reduce the fault information, delete the irrelevant condition information, and obtain the simplified fault decision table. Finally, the BP Neural Network is used for fault diagnosis. The simulation experiment is carried out with 80 sets of data from CTCS-300T on-board equipment as samples. The results show that: compared with BP Neural Network, RS_BP reduces the number of Neural Network input layers through RS attribute reduction. Algorithm iteration times, faster convergence speed and higher fault diagnosis rate.