基于半监督邻域自适应LLTSA算法的故障诊断

Fault diagnosis based on semi-supervised neighborhood adaptive LLTSA

为了有效利用振动信号进行故障诊断,提出了一种基于半监督邻域自适应线性局部切空间排列(SSNALLTSA)算法的故障诊断方法。从多域提取振动信号的混合特征,构建原始高维特征集。利用半监督邻域自适应线性局部切空间排列算法对原始特征集进行维数约简,提取出辨识性较高的敏感特征子集。将得到的低维特征输入SVM分类器进行识别,判断故障类型。液压泵故障诊断实验结果表明,该算法克服了LLTSA无监督和使用全局统一邻域参数的不足,可更有效地寻找数据的低维本质流形,提高了识别准确率,具有一定优势。

; In order to diagnose faults effectively using vibration signals, a fault diagnosis method based on the semi- supervised neighborhood adaptive linear local tangent space alignment （LLTSA） was proposed. Firstly, the mixed features of vibration signals were extracted in multi-domain to construct the original high-dimensional feature set. Then, the algorithm of semi-supervised neighborhood adaptive LLTSA was used to reduce the dimension of the original feature set and to extract the sensitive feature subset with the higher identifiability. Finally, lower dimensional features were input into a SVM classifier to recognize fault types. The fault diagnosis test results of hydraulic pumps indicated that the proposed algorithm overcomes drawbacks of LLTSA without supervision and using globally unified neighborhood parameters; lower dimensional intrinsic manifolds of data can be more effectively found with this algorithm ; it improves the recognition accuracy and has certain advantages.