Fault Bearing Identification Based on Wavelet Packet Transform Technique and Artificial Neural Network

Bearing race faults have been detected by using wavelet packet transform (WPT) technique, combined with a feature selection of energy spectrum. Vibration signals from ball bearings having defects on inner race and outer race have been considered for analysis. In the present fault diagnosis study, the artificial neural network techniques both using radical basis function (RBF) neural network and conventional back-propagation (BP) neural network are compared in the system to evaluate the proposed feature selection technique. The experimental results pointed out the proposed system achieved fault recognition rate of over 90% for various bearing working conditions. And RBF neural network is more effective than BP neural network in this fault diagnosis system.