基于排列熵和神经网络的滚动轴承异常检测与诊断

Abnormality Detection and Diagnosis of Rolling Bearing Based on Permutation Entropy and Neural Network

针对轴承不同状态下的复杂性特征,提出基于排列熵和神经网络的异常检测与诊断方法。介绍异常检测与诊断的原理,采用排列熵算法检测信号异常,提取能够敏感反映轴承不同异常模式(滚动体异常、内圈异常和外圈异常)的排列熵、嵌入维数及关联维数等复杂度参数形成特征向量,通过神经网络对异常模式进行分类识别。结果表明:排列熵算法可以定位异常发生的时刻,以时间序列的复杂性特征参数为输入的神经网络诊断方法能够有效识别轴承的不同异常模式。

Aiming at the complexity characteristics in different working conditions of rolling bearings, the method of abnormality detection and diagnosis based on permutation entropy and neural network was put forward. Principle of the abnormality detection and diagnosis was introduced, and the permutation entropy algorithm was used to detect signal abnormality. Then the eigenvector that was formed by complexity parameters of different abnormality models （boll, inner and outer abnormality of rolling bearing） was determined. Finally, abnormality diagnosis was carried out by neural network. Result shows that the permutation entropy algorithm can determine when the abnormality happens, and the diagnosis method based on neural network with the characteristics complexity parameters of time series as the input can effectively identify the different abnormality models.