Wavelet neural network and genetic algorithm neural network for faults diagnosis of railway rolling bearing

In this paper, we developed wavelet neural network (WNN) and genetic algorithm neural network for the diagnosis of railway rolling bearings. First, a discrete wavelet packet was adopted to decompose the signals of rolling bearings into sub-series with approximation and details frequencies, These sub-series were constructed into the wavelet packet energy eigenvectors, Then those wavelet packet energy eigenvectors were taken as fault samples to train BP neural network, WNN and genetic algorithm neural network. The experimental results show that the train BP neural network, WNN and genetic algorithm neural network can diagnose this kind of faults in rolling bearings, compared the results achieved by the BP neural network, the WNN and the genetic algorithm neural network, and found that the WNN and genetic algorithm neural network are superior to the BP neural network. Of the two, the WNN is better than the genetic algorithm neural network. This paper provides the theoretical foundation for fault diagnosis in rotary machines. © Sila Science. All rights reserved.