中文摘要：

针对液压泵振动信号非线性、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）、模糊熵和流行学习的液压泵故障特征提取方法。该方法将LCD、模糊熵和流行学习相结合。首先,利用LCD将振动信号分解成不同尺度下的内禀尺度分量（intrinsic scale component,ISC）并计算各分量的模糊熵,初步提取液压泵高维故障特征。其次,采用流行学习算法中较为典型的线性局部切空间排列（liner local tangent space alignment,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。液压泵的故障诊断实验表明,所提方法能够以较高的精度识别液压泵的各典型故障,具有一定的优势。

英文摘要：

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