基于变分模态分解与多尺度熵的滚动轴承故障诊断方法

Rolling Bearing Fault Diagnosis Method Based on Variational Mode Decomposition and Multiscale Entropy

针对滚动轴承故障振动信号具有非平稳、非线性特征以及提取特征困难等问题,提出一种基于变分模态分解(Variational Mode Decomposition,VMD)的多尺度熵(Multi-scale entropy,MSE)的特征向量提取方法,并输入拉普拉斯支持向量机(Laplacian support vector machines,LapSVM)中进行滚动轴承故障识别。该方法首先利用VMD分解的多尺度熵对原始振动信号进行特征向量的提取,然后与基于VMD样本熵以及VMD时域统计量(峭度、歪度)对比说明该方法的优势,最后将上述特征向量输入到LapSVM分类器中进行识别对比。试验数据分析结果表明,所提方法在诊断精度、计算速度上大大提高。

Aiming at the problem that the vibration signal of rolling bearing has nonstationarity,non-linearity and difficulty in extracting characteristic.a new rolling bearing fault diagnosis approach based on VMD Multi-scale entropy and LapSVM was put forward for extracting feature vectors.and the LapSVM is utilized to recognize the rolling bearing fault identification.In this method,the VMD and MSE was applied to extract the feature vector from the original vibration signals.then,Compared with VMD sample entropy and VMD time domain statistics（kurtosis,crooked）,the advantages of this method are illustrated.Finally,Feature vectors that obtain from the above methods was entered LapSVM to identify the rolling bearing fault categories.The results of the experimental data show that the proposed method is greatly improved in the diagnosis accuracy and calculation speed.