基于排列熵的振动信号小波包阈值去噪研究

工程实践中的振动信号往往存在噪声干扰而导致信号特征信息无法显露,传统小波包软、硬阈值函数去噪形式固定,无法依据信号小波包分解系数的噪声干扰情况进行调整.据此,提出一种新的介于软、硬阈值函数之间的改进小波包阈值函数,并将排列熵作为信号含噪情况表征参数引入阈值函数中.对信号小波包系数进行排列熵计算,并依据该值对阈值函数进行自适应调整,使得新的阈值函数能够对含噪较多的小波包系数进行大尺度收缩而对含实际信号特征较多的小波包系数尽可能地保留,从而达到最佳的去噪效果.对滚动轴承振动实验信号的去噪分析,并与其他方法进行对比,验证了该方法的有效性与优越性.

Research on Wavelet Packet Threshold Denoising of Vibration Signal Based on Permutation Entropy

The features of the engineering vibration signal are often obscured by the interference of the noise signal.Traditional wavelet packet soft and hard threshold functions are failed to adjust according to the noise contribution to the signal wavelet packet decomposition coefficients due to the lack of elastic denoising form.Therefore,a new improved wavelet packet threshold function between soft and hard threshold functions is proposed,and the permutation entropy is introduced into the new threshold function as the characterization parameter of noise signal.In order to shrink noisy wavelet packet coefficients in a large scale and reserve actual signal wavelet packet coefficients as much as possible,and achieve the best denoising effect,the permutation entropy of signal wavelet packet coefficients is calculated and the threshold function is adjusted adaptively based on the entropy value.The effectiveness and superiority of this method are testified by the denoising analysis of rolling bearing experiment vibration signal and the comparison with other methods.