基于VMD-KPCA-OSELM算法的滚动轴承故障诊断

Rolling Bear Fault Diagnosis Based on VMD-KPCA-OSELM

为了提高滚动轴承故障诊断的准确性,提出了变分模态分解(VMD)与在线贯序极限学习机(OSELM)相结合的VMD-KPCA-OSELM方法对轴承故障进行诊断。首先利用VMD对所提取轴承信号进行去噪;其次,应用KPCA对去噪后的数据进行故障特征提取和降维压缩;最后,运用OSELM方法对轴承故障进行分类。实验结果表明,所提的集合型方法能通过计算每个模态的中心频率将带宽的求解转化为约束问题的寻优,有效地区分出不同的模态,对故障信号进行滤波并提取故障特征,且在诊断快速性方面优于传统单梯度下降学习方法。所提的集合型VMD-KPCA-OSELM方法比常规的单一型方法更适用于滚动轴承的故障诊断。

To improve the accuracy of the rolling bearing fault diagnosis, the variational mode decomposition(VMD)combined with online sequential extreme learning machine(OSELM), called VMD-KPCA-OSELM approach for bearing fault diagnosis is present. Firstly, VMD is used to denoise the bearing signal; Secondly, KPCA(Kernel Principal Component Analysis) is applied for fault feature extraction and dimension reduction; Finally, the bearing fault is classified by OSELM. The experimental results show that the proposed ensemble method can solve the problem of bandwidth by calculating the center frequency of each mode and different modes can be effectively separated from the fault signals. This method can filter the fault signal and extract the fault feature, and is superior to the traditional single gradient descent learning method in terms of diagnosis rapidity. The proposed ensemble VMD-KPCA-OSELM approach is more suitable for the fault diagnosis of rolling bearings than the conventional single method.