基于相关性小波奇异熵的滚动轴承故障特征提取

Rolling Bearing Fault Feature Extraction Based on Correlation of Wavelet Singular Entropy

针对轴承故障信号受噪声影响严重,导致故障特征提取稳定性较差的问题,将小波变换、相关性、奇异值分解和信息熵理论相结合,提出一种基于相关性小波奇异熵的轴承故障特征提取方法。该方法首先将轴承信号进行小波分解,利用小波分解系数和噪声的相关性特点不同,引入相关计算以去除噪声的影响;然后对相关处理后的规范化系数进行奇异值分解,轴承的不同故障信息就体现在奇异值中;再利用信息熵的统计特性对奇异值进行不确定度计算;最后,以相关性小波奇异熵作为特征向量,通过概率神经网络对滚动轴承故障进行识别。实验表明:该方法能够有效地提取轴承故障特征,具有良好的容噪能力和稳定性。

Aiming at the poor stability in extracting bearing’s fault signals seriously affected by noises,and through combining the information entropy theory with wavelet transform and correlation as well as the singular value decomposition,the correlation wavelet singular entropy-based new method of extracting the bearing’s fault signal feature was proposed,in which,having the bearing signals processed through wavelet decomposi-tion;and considering the difference in correlation characteristics of the wavelet decomposition coefficients and the noises,having the related calculation introduced to remove the noise influence;having the normalized cor-relation coefficients handled with singular value decomposition can find the bearings’fault information in the singular value;according to the statistical features of the information entropy,having the singular value’s un-certainty calculated;taking the correlation wavelet singular entropy as the characteristic vector and making use of the probability neural network,the rolling bearing’s faults can be identified.Experimental results show that this method can effectively extract the bearing’s fault characteristics and it has better stability and tolerance to the noises.