结合小波包与神经网络的压缩机故障诊断方法

介绍了一种针对滚动转子式压缩机故障的小波包神经网络诊断方法。利用压缩机壳体顶部或侧部获取的振动信号, 通过小波包分解与单支重构, 提取出该振动信号各频率段的能量作为特征信息, 利用BP神经网络将正常和异常压缩机区别开来。此方法具有检测效率高、可靠性高等优点, 受生产现场环境的干扰小, 可用于压缩机产品的在线诊断。小波包与神经网络诊断方法对其他机械设备的故障实时在线诊断也具有一定的工程实用价值。

Compressor fault diagnosis based on wavelet packet and neural network

A wavelet packet and neural network detecting method is applied to rolling rotor compressors. The vibrating signals from the top and flank of the compressor housing are decomposed and reconstructed by wavelet packet. The energy of different frequency ranges is put into back-propagation neural network to judge normal and malfunction compressors. The method has high efficiency and reliability and can be used to detect compressor products online.