中文摘要：

针对滚动轴承故障诊断中特征向量难以提取与支持向量机结构参数选取依据经验的问题,提出了基于小波包与奇异值分解的GA-SVM滚动轴承故障诊断方法.首先,采用小波包对采集的滚动轴承各状态下的信号进行分解,获取表征信号局部特征的各节点系数,在此基础上构建各节点系数矩阵并进行奇异值分解,来获取特征向量进而将其作为故障诊断模型的输入;其次,利用遗传算法（GA）优化支持向量机（SVM）的惩罚系数和高斯核系数两个结构参数;最后,将上述特征向量作为输入,建立GA优化SVM的故障诊断模型,实现滚动轴承的状态辨识.实验结果表明,与BP、SVM、PSO-SVM相比,基于小波包与奇异值分解的GA优化SVM滚动轴承故障诊断方法具有更高的分类精度,能够提高滚动轴承状态辨识的效果.

英文摘要：

In order to solve the problem that the feature vectors is difficult to extract and the parameters affecting the fault classification accuracy is based on experience,it proposes a method of GA-SVM rolling bearing fault diagnosis based onwavelet packet and singular value decomposition.First,in order to get the coefficient of each node,this method decomposes the signal using the wav elet packet.Then,it decomposes the matrix that is consisted of each node＇s coefficient using the singular value theory.At last,it inputs the matrix of singular value into genetic algorithm （GA） optimized support vector machine （SVM） for classification and recognition.The test results show that this method has good classification accuracy for rolling bearing status and can well applied in the fault diagnosis of rolling bearing.