关键词: 小波分析, 神经网络, 故障诊断

Abstract: Structure model and algorithms of Wavelet Neural Network（WNN） are designed combining the advantages of both wavelet transform and Artificial Neural Network（ANN）.Compared with the slow convergence rate of conventional BP neural network，the algorithms using variable learning rate and momentum factor can be evidently improved.Wavelet neural network is applied to typical faults diagnosis of rolling element bearings.Experimental data obtained from 7216 tapered bearings experiment table is used to train the network，using vibration signal as its input.The simulation results indicate that the WNN can accurately classify faults.The WNN obviously surpasses the conventional BP neural network，and it efficiently accomplishes the fault diagnosis of rolling element bearings.

Key words: wavelet analysis, neural network, fault diagnosis