基于LMD的多尺度形态学在齿轮故障诊断中的应用

Application of multi-scale morphology based on LMD in gear fault diagnosis

数学形态学是一种非线性、非平稳分析方法,具有很强的抑制脉冲干扰的能力,但滤除白噪声的能力却不行。针对这一不足,提出一种基于LMD的多尺度形态学解调方法。该方法是将多分量调频调幅故障信号分解为一系列单分量调频调幅信号(PF),实现对故障信号的降噪处理,同时还可以获得原始信号的全部调制信息。选取能量高的PF分量求和重构,再用多尺度形态学差值滤波器提取出故障信号的频率特征。通过仿真和齿轮故障模拟实验证实了该方法的有效性。

Mathematical morphology is a kind of nonlinear and non-stationary analysis method,it has a strong ability to restrain pulse interference,but has no ability to filter the white noise.Aiming at this shortage,a multi-scale morphology method based on local mean decomposition （LMD ） was proposed.A fault signal with multi-component frequency modulation and amplitude modulation was decomposed into a series of production functions （PFs）to realize fault signal de-noising processing and to obtain the full modulation information of the original signal.The PFs with higher energy were selected and summed,then they were reconstructed.Finally,the multi-scale morphological difference filter was used to extract the fault signal characteristic frequency.The validity of this method was verified through simulation and gear fault simulation.