基于小波包变换的电机定子故障特征提取方法

A feature extraction method for motor stator fault based on wavelet packets transform

针对感应电机定子故障的特征频率处在低频段,小波分解系数易受电机负荷波动影响的问题,提出一种采用希尔伯特变换对信号进行预处理,利用小波包分解来实现定子故障特征的提取方法。通过小波包分解,使相应分解子频段能始终覆盖随电机转差率以及供电电源频率变化的故障特征频率。增加小波时域波形的波峰数,减少了子频段间的频域混叠及频谱泄漏现象。对原始信号进行希尔伯特变换的预处理,降低了电机负荷波动对分解系数的影响;采用子频段节点重构系数的均方根值变化率作为故障特征指标。通过对实测故障数据的应用,利用上述方法可以有效地识别出电机的定子故障。

Wavelet coefficient is disturbed by motor load fluctuation, because eigenfrequencies of motor stator fault lie in lower frequency bands. A feature extraction method for stator fault is introduced based on Hilbert transform and wavelet packets transform. Selecting the fitting decomposition, the decomposition frequency bands always cover the motor stator fault eigenfrequencies varied along with the slide and power frequency. By increasing crests of wavelet, both wavelet overlap and spectrum leak between the adjacent frequency bands are decreased. Aiming at reducing the side effects of the various decompose coefficient caused by load fluctuation, the process method for current signal based on Hilbert transform is proved to be effective. The motor fault eigenvalues are obtained by the mean-squared root method based on reconstructed node coefficients. Under the laboratory condition, an artificial motor fault experiment is conducted, and the data are recorded. Through applying the proposed method to the recorded data, the motor stator faults can be effectively identified.