LMD滤噪算法及在旋转机械转子故障诊断中的应用

An Improved LMD-based Denoising Algorithm for Fault Diagnosis of Rotating Machinery

利用噪声统计特性及局部均值分解算法(local mean decomposition,LMD)在信号分解过程中的自适应性,提出了一种新的基于LMD的自适应滤噪算法。该算法完全由数据驱动,可对信号自适应降噪,并将降噪后的信号分解为若干个瞬时频率具有物理意义的PF(Product function)分量。重构的信号可有效提高功率谱故障诊断的性能。通过对2种非平稳信号的仿真实验及在实际运行状态下采集的旋转机械转子振动信号降噪的应用,结果表明提出的算法降噪性能优于中值降噪、均值降噪、小波降噪、EMD软阀值降噪等典型滤噪算法。该算法也可在频域有效地用于旋转机械转子故障的诊断。

A new adaptive signal denoising algorithm based on local mean decomposition（LMD） for machine fault diagnosis was proposed. The method was fully data driven. The denoised signal could be decomposed adaptively into a set of single AM-FM components called product functions（PFs）, and the reconstructed signals could effectively improve the power spectrum performance of fault diagnosis. Through the simulation experiment on two different unstable signals and the application to denoising of real vibration signals acquired from the faulted rotors of rotating machines, this method was proved to have better performance than the averaging, median, wavelet and empirical mode decomposition（EMD） soft threshold approaches. The experimental results show that this method can detect rotor fault features efficiently and can be applied to the fault diagnosis of the rotating machines.