Abstract: Aiming at the problems of denoising in mechine vibration signal extrations, based on the wavelet packet multi-threshold criterion denoising method, an improved wavelet packet denoising method to multi threshold standards was proposed herein, referred to improved FMC denoising method. This method was applied to pre-process original vibration signals of machine tools by pseudo point elimination technique to eliminate interferences of external noise signals to vibration signal extractions. Then the method of determining decomposition level was optimized by the method of determining maximum decomposition level with minimum frequency components of useful signals, and optimal wavelet packet basis was determined by minimum cost principle. Finally, vibration signals were reconstructed by wavelet packet multi threshold denoising criterion. Simulation experiments were implemented, and the results show that the method has good noise reduction effect, especially in strong noise background with a mutation of noise signal recovery. Experimental results show that the improved FMC denoising method may effectively eliminate band noises and improve separability of signal characteristics. Experiments were carried out to analyze the vibration signals of air bag dressing machine using this method, and experimental results show that the improved FMC denoising method may effectively eliminate the band noise and improve signal feature separability.

Key words: vibration signal, denoising, wavelet analysis, probe interpolation, multi threshold, class distance criterion

摘要： 针对机械振动信号提取时面临的去噪问题，在小波包多阈值准则去噪法的基础上，提出一种改进的小波包多阈值准则综合去噪方法(改进FMC去噪法)。该方法首先采用探测插值法对机床原始振动信号进行预处理，剔除受外界干扰产生的突变噪声信号；再以小波包分析为基础，根据有用信号的最小频率确定最大分解层数，并按最小代价原理确定信号分解的最佳小波包基；最后采用小波包多阈值降噪准则对振动信号进行重构，得到去噪后的机床振动信号。针对含噪blocks信号、doppler信号及模拟的含噪振动信号进行的仿真实验结果表明，改进后的FMC去噪法去噪效果优于传统方法。将该方法应用于气囊修整机振动信号分析中，结果表明，改进FMC去噪法能够有效剔除振动信号各频段的噪声，提高信号特征的可分离性。

关键词: 振动信号, 去噪, 小波分析, 探测插值, 多阈值, 类间距判据