基于奇异值差分谱与改进包络分析的轴承故障特征提取

Bearing Fault Feature Extraction Based on Difference Spectrum of Singular Value and Improved Envelope Analysis

针对滚动轴承振动信号故障特征难以提取的问题,提出了一种基于奇异值差分谱与改进包络分析的轴承故障特征提取方法。首先,通过奇异值分解将原始轴承振动信号分解为一系列能够线性叠加的分量信号,利用故障特征分量和噪声分量在奇异值上的差异,根据奇异值差分谱的性质筛选出有效奇异值,选择包含故障特征的分量重构信号。针对奇异值分解去噪后仍存在残余噪声,采用改进包络分析,在频域中进一步去除重构信号中的残余噪声。最后对实测轴承信号进行分析,准确地提取到故障特征明显、故障频率突出的轴承故障信号,完成故障诊断。

To overcome the difficulty of extracting fault feature from rolling bearing vibration signal, a method is put forward based on difference spectrum of singular value and improved envelope analysis. Firstly, the original bearing vibration signal is decomposed into a series of component signals which is able to be added linearly by singular value decomposition （SVD）. Due to the difference of singular values between the component signals of fault feature and component signals of noise, the effective singular values are selected according to the property of difference spectrum of singular value and then the component reconstruction signals containing fault feature are selected. Aiming at residual noise after denoised by singular value decomposition, the improved envelope analysis is adopted to further remove residual noise in reconstructed signal in frequency domain. Finally, actual inspected bearing signal is analyzed, the bearing fault signal with obvious fault feature and prominent fault frequency is extracted accurately and then fault diagnosis is completed.