EEMD-ICA联合降噪的旋转机械故障信号检测方法

Fault Signal Detection Method of Rotating Machinery Based on EEMD-ICA Joint Denoising

针对旋转机械前期故障信号微弱、易被噪声淹没、故障特征难以提取的问题,提出一种聚合经验模态分解(Ensemble Empirical Mode Decomposition,EEMD)和独立成分分析(Independent Component Analysis,ICA)相结合的故障特征提取方法。首先,运用EEMD理论将振动信号分解为一系列的固有模态函数(Intrinsic Mode Function,IMF),然后根据相关系数和均方根准则选取含有原始信号多的IMF分量构造观测信号,引入虚拟噪声通道;最后,通过FastICA算法将噪声与故障特征信号进行分离,并对分离出的有用信号进行频谱分析,突显故障频率。通过仿真信号验证所提出方法的有效性,并将其应用于轴承的内外圈故障识别,与传统的EEMD-WTD降噪方法对比,结果表明:所提出的方法能提取出清晰微弱故障特征信号,对低频噪声的抑制效果明显优于EEMD-WTD方法。

Aiming at the problems that early fault signals of rotating machinery are weak, easy to be submerged by noise, and it is difficult to extract the fault features, a fault feature extraction method combining ensemble empirical mode decomposition(EEMD) and independent component analysis(ICA) is proposed. First of all, EEMD theory is used to decompose the vibration signal into a series of intrinsic modal functions(IMF). Then, the IMF component with more original signals is selected according to the correlation coefficient and root mean square criterion, and the virtual noise channel is introduced. Finally, the FastICA algorithm is used to separate the noise and the fault characteristic signal, and FFT spectrum analysis is perform on the separated useful signal to highlight the fault frequency. The effectiveness of the proposed method is verified by simulation signals, and it is applied to the fault identification of inner and outer rings of bearings. The results show that compared with the traditional EEMD-WTD noise reduction method, the proposed method can clearly extract weak fault characteristic signals. In particular, the suppression effect of low-frequency noise is significantly better than the EEMDWTD method.