基于小波包能量谱及SVM算法的轴承故障检测

Bearing fault detection method based on wavelet packet energy spectrum and SVM algorithm

针对小波分析在故障诊断时的局限性,将小波分析和支持向量机算法相结合,提出基于小波包能量谱及支持向量机算法(SVM)的故障检测方法。该方法以振动信号小波包分解后各子频带的能量作为故障检测特征,利用SVM算法对轴承故障进行检测实验。结果表明:小波包能量谱能有效地反映轴承信号特征,并对故障进行检测。该方法同基于Lipschitz指数熵、单奇异点检测,以及小波包能量谱与神经网络相结合的故障检测方法进行比较,检测率均优于其他三种常用方法。

This paper is concerned with a novel scheme designed for overcoming the limitation resul- ting from the use of wavelet analysis for fault diagnosis. This scheme is developed by combining wavelet analysis with support vector machine, builds on wavelet packet energy spectrum and support vector ma- chine algorithm, and works by using sub band energy following vibration signals wavelet packet decompo- sition as fault detection characteristics and the SVM algorithm and thereby detecting bearing faults using SVM algorithm. The results demonstrate that the wavelet packet energy spectrum is capable of an effective reflection of bearing fault characteristics and fault detection. It follows that the scheme boasts a higher de- tection rate than the other three conventional detection methods based on Lipschitz index entropy, singular point detection and the combination of wavelet packet energy spectrum and artificial neural network.