Engine Fault Diagnosis Based on Wavelets Packet and Neural Networks

Abstract

A fault diagnosis system is presented for connecting rod bearings in engine based on wavelet packet energy feature and BP neural network. Four-layer wavelet decomposition is conducted on the vibration signals of connecting rod bearing, and the energy of wavelet packet is extracted as the feature parameter of vibration signal of connecting rod bearing. Then these feature parameters are used to train BP neural network for fault pattern recognition. Test results show that applying wavelet packet energy and BP neural network to fault diagnosis of connecting rod bearing is feasible and effective.