改进小波阈值函数和ACEWT方法的滚动轴承故障声发射信号特征提取

Improved wavelet threshold function and ACEWT method for feature extraction of acoustic emission signals from rolling bearing faults

针对滚动轴承早期故障声发射信号受复杂传递路径和噪声的干扰,声发射信号信噪比较低,导致轴承故障特征难以提取的问题,提出了改进小波阈值函数-ACEWT的轴承故障特征提取方法。由于声发射信号呈冲击性与快速衰减的特点,构建一种衰减正弦型与指数型的小波阈值函数对低信噪比的声发射信号进行降噪。研究自相关运算与经验小波变换结合的方法(autocorrelation and empirical wavelet transform,ACEWT),用于滚动轴承故障声发射信号特征提取,解决了在低信噪比下经验小波变换对轴承故障特征提取的不足;引入经验小波能量比-熵指标,选取最优经验小波系数。通过与经验小波变换、改进小波阈值函数-EWT和MCKD-EWT方法进行对比研究,并试验验证。仿真和试验结果表明,所提方法明显优于经验小波变换、改进小波阈值函数-EWT和MCKD-EWT方法,可准确提取轴承故障声发射信号的频率特征。

Here,aiming at the problem of early fault acoustic emission signals of rolling bearing being interfered by complex transmission paths and noise,and having lower signal-to-noise ratio to cause difficulty in extracting bearing fault features,a bearing fault feature extraction method based on improved wavelet threshold function-autocorrelation and empirical wavelet transform(ACEWT)was proposed.Due to characteristics of impulsiveness and rapid attenuation of acoustic emission signals,a wavelet threshold function with attenuation sine type and exponential type was constructed to denoise low signal-to-noise ratio acoustic emission signals.ACEWT method for feature extraction of acoustic emission signals of rolling bearing faults was used to solve the problem of insufficient extraction of bearing fault features with empirical wavelet transform(EWT)under low signal-to-noise ratio.The empirical wavelet energy ratio-entropy index was introduced to select the optimal empirical wavelet coefficients.The proposed improved wavelet threshold function-ACEWT method with EWT,improved wavelet threshold function-EWT,and maximum correlated kurtosis deconvolution(MCKD)-EWT methods were studied contrastively,and verification tests were conducted.The simulation and test results showed that the proposed method is obviously superior to EWT,improved wavelet threshold function-EWT,and MCKD-EWT methods;it can accurately extract frequency features of bearing fault acoustic emission signals.