电机转子振动信号故障特征提取方法

Fault Feature Extraction Method of Motor Rotor Vibration Signals

电机振动信号具有非平稳、非线性特性,在进行时频域特征提取时需要人工确定时间窗口和基函数。针对该问题,提出一种基于EEMD(Ensemble Empirical Mode Decomposition)分解和权重熵变换的特征提取方法。该方法应用EEMD进行自适应分解,同时提取IMF(Intrinsic Mode Function)方差作为初始特征值,并提出将权重熵作为衡量特征值重要性的标准,通过权重熵对原始特征值进行空间变换,扩大特征向量间的差异。为验证该方法的有效性和优越性,对4种状态下的电机转子进行振动信号采集,用于制作转子故障数据集,并运用EEMD进行特征提取。实验结果表明,基于EEMD分解和权重熵变换的特征提取方法能够更好地从振动信号中提取特征向量,在对电机进行故障诊断时具有更高的准确性。

Because of the non-stationary and nonlinear characteristics of motor vibration signals,it is necessary to determine the time window and base function manually when extracting the features in the time-frequency domain.In this paper,a feature extraction method based on EEMD decomposition and weight entropy transformation is presented.In this method,the EEMD is applied to adaptive decomposition,and the IMF variance is extracted as the initial eigenvalue.And the weight entropy is put forward as a criterion to measure the importance of the eigenvalues.Through the spatial transformation of the original eigenvalues by using the weight entropy,the difference between eigenvectors is enlarged.To verify the effectiveness and superiority of the method,vibration signals of motors in four states are collected to form the rotor fault dataset,and the EEMD is used to carry out the signal feature extraction.The experimental results show that the feature extraction method based on EEMD decomposition and weight entropy transformation can better extract feature vectors from vibration signals,and has higher accuracy in motor fault diagnosis.