EMD-ICA与SVM在滚动轴承故障诊断中的应用

Application of EMD-ICA and SVM in Fault Diagnosis of Rolling Bearings

针对滚动轴承非线性的早期故障信号,应用独立分量(ICA)将滚动轴承产生的故障信号从多通道混合信号中分离出来,然后采用EMD(Empirical Mode Decomposition)进行再次降噪并建立AR模型,最后提取模型的自回归参数和残差方差作为故障特征向量,并以此作为支持向量机(SVM)分类器的输入参数来区分滚动轴承的工作状态和故障类型。实验结果表明,该方法是有效的。

Aiming at the early non-linear fault signals of rolling bearings, the ICA is employed to separate the fault signals of the rolling bearing from the mixed signals collected by the multi- channel. Then, the EMD method is used to reduce the noise and establish the AR model. Finally, the self-regressive parameters and the residual square difference of the model are extracted and regarded as the fault characteristic vectors. They are used as the input parameters of the SVM classifier to distinguish the working condition and the type of faults of the rolling bearing. Experimental results show that this approach is effective.