中文摘要：

针对轴承振动信号具有的非平稳和故障诊断样本数据难以按需获取的问题,设计了一种基于小波包分解和EMD SVM的故障诊断方法;首先,采用Mallat塔式算法对信号进行降噪,实现信号的小波分解,获得重构后的故障诊断子频带信号;然后,在经典的EMD算法的基础上定义了改进的EMD算法,采用改进的EMD算法对经过小波包降噪的故障诊断子频带信号进行特征提取,从而获得故障诊断特征向量;最后,采用适合小样本分类的SVM进行故障诊断,将经过小波包降噪和EMD特征提取的样本数据用于训练SVM,得到用于故障诊断的多个二分类SVM故障诊断模型,通过投票机制来确定样本数据最终对应的故障诊断类别：在Matlab环境下对轴承故障诊断进行实验,实验结果证明了文中基于小波包和EMD-SVM的方法一种适用于小样本的故障诊断方法,且与其它方法相比,具有诊断效率高和精度高的优点.

英文摘要：

Aiming at the bearing vibration signal having the problems such as non--stability and sample data is hard to obtain, a fault di agnosis method based on wavelet packet decomposition and EMD--SVM is proposed. Firstly, the Mallat pyramid algorithm is used to reduce the noise to realize the wavelet decomposition. Then the improved EMD algorithm is proposed based on the classic EMD algorithm, the improved EMD algorithm is proposed to realize the signal extraction to get the fault diagnosis feature vector. Finally, SVM is used to diagnose the fault and the sample data is used to train the SVM classifier based on wavelet packet decomposition, then the multi SVM diagnosis model for fault diagnosis, the voting mechanism is used to assure the final fault. The bearing experiment is implemented in the Matlab environment, the experiment result proves the method based on wavelet packet decomposition and EMD--SVM is suitable for the small sample set, and compered with the other methods, it has the higher diagnosis efficiency and accuracy.