利用最优小波尺度循环谱的滚动轴承早期故障特征提取

Rolling Element Bearing Incipient Fault Feature Extraction Based on Optimal Wavelet Scales Cyclic Spectrum

滚动轴承微弱故障信号往往包含在二阶循环平稳信号中,但容易受到噪声干扰。对循环平稳信号进行基于短时傅里叶变换的循环周期谱分析可以提高周期故障的识别能力,但其结果受到窗函数大小的影响且对于微弱故障的诊断效果不佳,提出最优小波尺度循环谱进行滚动轴承的早期故障诊断。首先利用连续小波变换对信号进行处理获得小波系数;接着采用相关峭度方法选择最优的分析尺度;然后沿着时间轴对该尺度范围内的小波系数进行循环谱分析;最后对最优尺度下的循环谱平均进行特征提取。与循环周期谱的分析结果进行对比,验证了该方法在早期故障特征提取方面的有效性。

Rolling Element Bearing fault characteristic information is within the second order cyclic stationary signal. But it is susceptible to noise interference. The method of cyclic periodogram based on the short-time Fourier transform used on the cyclic stationary signal analysis can improve the recognition ability of cyclic failure. But it is not good for weak fault characteristic identification and its results affected by the size of the window function. The method of optimal wavelet scales cyclic spectrum is proposed for detection of rolling element bearing early faults. The continuous wavelet transform is carried on vibration signal processing to obtain the wavelet coefficients firstly. Then, the optimal scale is selected by correlated kurtosis values. And then, the wavelet coefficients in this scale range are analyzed by using cyclic spectra along the time axis. At last, the average value of the cyclic spectra under the optimal scales is calculated for feature extraction. Comparison with the result of cyclic periodogram, it can be concluded that the proposed method has good performance for rolling element incipient fault feature extraction.