基于局部特征尺度分解排列熵和线性局部且空间排列的故障特征提取方法

Fault Feature Extraction Method Based on Local Characteristic-scale Decomposition Permutation Entropy and Liner Local Tangent Space Alignment

针对机械振动信号非线性、非平稳性以及故障特征难以提取的问题,提出了基于局部特征尺度分解(local characteristic-scale decomposition,LCD)、排列熵和线性局部切空间排列(Liner local tangent space alignment,LLTSA)的机械故障特征提取方法。该方法将LCD、排列熵和LLTSA相结合。首先,利用LCD将机械振动信号分解成不同尺度下的内禀尺度分量(intrinsic scale component,ISC)并计算各分量的排列熵,初步提取高维故障特征。其次,采用LLTSA对故障特征进行二次特征提取,得到维数低、敏感度高且聚类性好的低维特征。最后,采用支持向量机(support vector machine,SVM)对提取特征进行评估。滚动轴承的故障诊断实验表明,所提方法能够以较高的精度识别滚动轴承的各典型故障,具有一定的优势。

Aiming at the fact that the mechanical vibration signal would exactly display non-stationary characteristics and fault features hard to extracted, a mechanical fault feature extraction method based on local characteristic-scale decomposition （LCD） permutation entropy and liner local tangent space alignment （LLTSA） was proposed. The proposed method combined the LCD, permutation entropy and LLTSA. Firstly, the vibration signals was decomposed into several ISCs （ intrinsic scale component） and permutation entropy of each ISC was calculated, and the high-dimension fault feature was preliminarily extracted. Secondly, LLTSA was applied to compress the high-dimension features into low-dimension features which have better discrimination. Finally, the support vector machine （SVM） was employed to evaluate the feature extraction method. Experiment results of rolling bearing show that the proposed method can classify different fault type of roiling bearing exactly and has certain superiority.