A Fault Diagnosis Method of 25Hz Phase-Sensitive Track Circuit Based on BP Neural Network Optimized by Genetic Algorithm
Xinkai Wang · Journal of Physics Conference Series · 2022
In this paper, considering the actual situation of the track circuit, aiming at the fault diagnosis of the 25Hz Phase-Sensitive track circuit, because the BP neural network (BPNN) model possess the characteristics of succinct construction and powerful nonlinear fitting ability as well as good fault tolerance performance, A fault diagnosis system of 25Hz Phase-Sensitive track circuit based on BPNN is designed. In addition, considering that the BPNN is greatly affected by the initial weight. Different results can be obtained when BPNN fault classifier is initialized with different weights, and it is easy to converge to the local minimum, so the genetic algorithm is selected to optimize the initial weight and threshold of the BPNN. The research results show that the BPNN optimized by genetic algorithm has fewer iteration steps, lower average error, higher classification accuracy, and is simple and easy to handle, which can better deal with nonlinear system problems such as track circuit fault diagnosis.