基于LCD降噪与LS-SVM的滚动轴承故障诊断方法

Fault Diagnosis Method of Rolling Bearing Based on LCD Denoising and LS-SVM

在实际工况下滚动轴承较易发生故障,为了保障机械运行可靠性,对其进行故障诊断研究显得非常重要,提出一种基于局部特征尺度分解(Local Characteristic-scale Decomposition,LCD)降噪与最小二乘支持向量机(Least Squares Support Veotor Machine,LS-SVM)的滚动轴承故障诊断方法。首先,利用LCD对轴承信号进行自适应性分解,得到一系列内禀尺度分量(Intrinsic Scale Component,ISC),然后结合峭度准则筛选出包含主要特征信息的分量,完成信号降噪预处理,并与经验模态分解(Empirical Mode Decomposition,EMD)进行对比,研究LCD算法的优越性;最后提取ISC模糊熵作为信号的敏感特征集,输入到训练好的LS-SVM分类器中进行轴承状态识别。实验研究表明,提出的基于LCD降噪与LS-SVM的轴承故障诊断方法能有效地识别出多种轴承类型,识别率高达84%,是一种行之有效的轴承诊断算法。

In actual conditions,the rolling bearing is more prone to failure. In order to guarantee the reliability of mechanical operation,it is very important to study the fault diagnosis of the machine. The fault diagnosis method of rolling bearing based on local characteristic-scale decomposition( LCD) denoising and least squares support vector machine( LS-SVM) is proposed. Firstly,the bearing signal is adaptively decomposed by LCD. A series of intrinsic scale components( ISC) are obtained. Then,combined with the kurtosis criterion used to select mainly contains the characteristic information of the components,the signal denoising preprocessing is completed. Compared with EMD,the superiority of the LCD algorithm is studied. Finally,ISC fuzzy entropy is used as a sensitive signal feature extraction and the input of trained LSSVM classifier. Experimental results showthat the proposed bearing fault diagnosis method based on LCD denoising and LS-SVM can effectively identify a variety of bearing types and the recognition rate is up to84%. The method is a kind of effective bearing diagnosis algorithm.