基于小波包和LMS自适应降噪的柴油机振动诊断

摘要

由于柴油机振动信号的特征频带和噪声频带存在重叠现象,利用小波阈值消噪时难以选取合适的小波阈值,针对该问题提出一种基于小波包的LMS(Least Median of Squares)自适应滤波降噪方法.该方法将小波包与LMS自适应滤波相结合,首先利用小波包变换对信号进行多层分解,然后以噪声干扰对应尺度上的第一层"细节"分量及最大分解尺度上的逼近分量重构信号,将重构后的信号作为LMS自适应滤波器原始输入信号,再以小波包最大分解尺度上的高频细节信号作为自适应抵消器的参考输入信号,进行LMS自适应滤波降噪处理.仿真计算和工程应用表明,该方法参数设置较少,易于控制,不涉及小波阈值降噪中阈值的选取问题.对比试验信号的分析验证方法的有效性,将该法应用在柴油机振动诊断中提高故障识别率.%Because of the overlap for frequency band both of vibration characteristic signal and noise in diesel engine,it is difficult to choose suitable wavelet threshold for wavelet threshold denoising in vibration signal processing. Aiming at this problem, a new Adaptive Filtering Denoising method based on LMS (Least Median of Squares) wavelet packet transform for diesel engine' s vibration signals processing is proposed. Firstly, the multi-layer decomposition is carried out for the vibration signal using wavelet packet transform. Then, the signal is reconstructed using the first-layer wavelet detailed components and the asymptotic components of the maximum decomposition scale which corresponds to the noise disturbance. Finally, the reconstructed signal is used as the input signal for the LMS adaptive filter, and the detailed signal with high frequency in the maximum decomposition scale of wavelet packet transform is used as reference input signal for the adaptive eliminator. Then the optimal filtering for signal-noise decomposition is realized. Both computer simulation and engineering application demonstrate that this method is excellent to realize the optimal estimate for the valuable signal and noise of the diesel engine vibration signal. It doesn't need to choose suitable wavelet threshold for wavelet threshold denoising in vibration signal processing. It has less parameter setting and is easy to control. Using this new method, the effect of the engine vibration fault recognition is raised.