基于最优Morlet小波和隐马尔可夫模型的轴承故障诊断

Fault diagnosis of rolling element bearings based on optimal morlet wavelet and hidden markov model

提出一种从信号时频域提取故障特征的新方法,先将振动信号作Morlet小波变换,再将小波系数顺序划分成多个子列,各子列协方差矩阵的特征值为所需的特征参数。为了更有效地提取信号的振动特性及周期性成分,使用了最小香农熵准则和奇异值分解技术选择Morlet小波参数,并用比较实验证明了参数优化的有效性。状态辨识使用了连续型隐马尔可夫模型,在三种故障程度下分别实现轴承正常状态,滚动体故障,内圈和外圈故障的正确辨识,平均精度都大于93%。

A new approach for fault diagnosis of rolling element bearings was presented using the optimal Morlet wavelet and the statistical characteristic of the wavelet coefficients.It was shown that the optimization of the wavelet parameters benefits to extract the effective features.Thus,the criterion of the minimum Shannon entropy and the singular value decomposition technology were used to optimize the parameters of Morlet wavelet.The feature extraction firstly divided Morlet wavelet coefficients into a series of segments.Then,the infinite-norm of the covariance matrix for each segment was calculated,which was also used to construct the observation vectors of the hidden Markov model.Finally,the test results of bearing faults identification and isolation were presented and all the identification accuracies were greater than 93%.