中文摘要：

针对频谱分析不能确定轴承故障程度的缺点,提出PCA（principle component analysis）与优化的SVM（support vector machine）相结合的方法,研究轴承不同故障类型、同种故障类型不同故障等级的实验振动数据,同时对轴承振动信号的13个特征属性参数进行了主成分分析,确定了最优特征属性参数,并利用优化的SVM对轴承故障进行诊断。实验结果表明：该方法确定了最优属性参数,减少了冗余信息,提高了诊断准确率,减少了时间消耗,不仅有效地诊断出了轴承的故障类别,而且实现了轴承的故障等级诊断,使诊断更加精细化,为工程实际中轴承的健康管理提供了有益参考。

英文摘要：

As the conventional frequency spectrum analysis method fails to determine the fault levels of bearing, this paper proposes a method combining PCA （principle component analysis） with the improved SVM （support vector machine） to diagnose fault levels of bearing. According to the bearing vibration experiment data of different levels of fault types, 13 characteristics parameters of bearing vi- bration signals are extracted as the training and diagnosis samples to determine the bearing states. This method can determine the extracted features of the properties parameters, thereby reducing the redundant information and improving the diagnostic accuracy in the process of diagnosis. Experimental research shows that the improved method can not only diagnose the bearing fault category effectively but also identify the fault levels of bearing, which provides a useful guide for precision diagnosis of bearing malfunction.