子波神经网络在故障诊断中的应用研究

根据旋转机械常见的的故障类型和故障信号时域采样数据 ,以子波空间作为模式识别的特征空间 ,采用信息熵为代价函数的神经网络学习算法 ,由子波神经网络对故障进行学习和诊断 .实验结果表明 ,子波神经网络的故障诊断方法在不了解故障信号频率结构的情况下 ,即可对平稳和非平稳故障信号进行诊断 ,适于设备在线监测及设备的巡检

Application Research of Wavelet Neural Network to Fault Diagnosis

This paper proposes a wavelet neural network for fault diagnosis. According to the faults of rotary machinery, based on wavelet space, the neural networks for fault diagnosis are trained with fault signals in time domain. The learning algorithm is based on entropy cost function. The result indicates that wavelet neural networks can diagnose stationary and non-stationary fault signals without analyzing the frequency structure of fault signals, and this method can be applied to going on the tour of monitoring for machinery on the spot.