摘要： 提出一种结合小波包分析(WPA)理论和支持向量机(SVM)分类器的机械故障诊断方法。该方法具有重复训练样本少，简单、直观的优点，具有很高的分类性能。利用获得的机械故障数据建立故障分类器，对不同测试集条件下的3种SVM核函数、SVM方法与神经网络方法的比较结果证明，基于小波包和支持向量机的故障诊断方法是机械故障诊断的发展方向。并对实验的最佳训练样本集进行讨论。

关键词: 小波包分析, 支持向量机, 故障诊断

Abstract: A novel method for machinery fault diagnosis combining wavelet packet analysis and multiple support vector machine classifier is put forward. The method has little duplicating training samples and is simple, and its classification accuracy rate is very high. Experimental results show that the method proposed above can successfully be applied to diagnosis of machinery faults, and the best training set is discussed in this paper.

Key words: Wavelet Packet Analysis(WPA), Support Vector Machine(SVM), fault diagnosis

中图分类号: 

• TP18