基于LMD-ICA降噪的滚动轴承故障特征提取方法研究

在滚动轴承进行故障识别中,针对局部均值分解(LMD)方法分析非平稳、非线性含噪信号时,存在端点效应,易产生虚假分量和单通道独立成分分析(ICA)盲源分离时的欠定问题,提出了基于LMD-ICA降噪的振动信号特征提取算法.首先对原始信号进行LMD,并抑制端点效应,得到n个瞬时频率具有物理意义的乘积函数(PF)之和;然后对得到的PF分量以连续的3阶PF分量为一序列组合进行ICA,可以得到n-2个重构分量;最后利用n-2个分量进行重构,得到降噪后的故障信号,并再次进行LMD或功率谱计算,提取故障特征.经验证,该方法可有效识别滚动轴承的多类故障.

Research of the Rolling Bearing Fault Signal Feature Extraction Method Based on the LMD-ICA Noise Reduction

Aiming at the problem of existing the end effect,easily producing false component.The local mean decomposition (LMD) was used in analysis of non-stationary and nonlinear signals containing noise and the underdetermined problem of the single channel independent component analysis (ICA) blind source separation,the feature extraction algorithm of the vibration signals based on the noise reduction LMD-ICA was proposed.By applying the algorithm,the fault of rolling bearing is diagnosed,firstly,the original signals are decomposed by LMD,the n instantaneous frequency product function (PF) components with physical significance are obtained and the end effect is limited.Then the obtained PF components which are arranged into one sequence combination per constant third-order PF components are analyzed by independent component.And n-2 refactoring components can be obtained by ICA.At last,the fault signals of which the noises are reduced are decomposed by LMD or calculated by power spectrum again to extract the fault.According to the experiment,the multi-types fault signals of rolling bearing can be effectively recognized by the method proposed above.