基于改进的有监督正交邻域保持嵌入的故障辨识

Fault Identification based on Improved Supervised Orthogonal Neighborhood Preserving Embedding

正交邻域保持嵌入(Orthogonal neighborhood preserving embedding,ONPE)是一种无监督的特征降维方法,且使用的是全局统一的邻域参数,在对高维故障特征集进行特征降维时,不能利用样本的类别标签信息和不能够根据样本空间分布的变化自适应调整邻域参数,使获得的低维特征仍出现混叠的情况。针对上述问题,提出了基于改进的有监督正交邻域保持嵌入(Improved supervised ONPE,IS-ONPE)特征降维的故障辨识方法。IS-ONPE利用样本的标签信息来调整样本点与点之间的距离以形成新的距离矩阵,通过新的距离矩阵进行邻域构建,同时利用局部集聚系数进行邻域参数的自适应调整,能够获得辨识度更高的低维特征。以低维特征作为支持向量机(Support Vector Machine,SVM)的输入来实现故障辨识。齿轮的故障辨识结果表明,所提出的方法能够提高故障辨识效果,具有一定优势。

The Orthogonal neighborhood preserving embedding （ONPE） is an unsupervised feature di- mension reduction method and only use global neighborhood parameter, when it is used to high - dimension fault feature for feature dimension reduction, it is incapacity of using sample class label information and adaptive adjust neighborhood parameter while the space distribution of samples changed. Aiming at the problems above, a fault identification method based on improved supervised orthogonal neighborhood preserving embedding （ IS - ONPE） for feature dimension reduction is proposed. In IS -ONPE, the distance between different points is ad- justed by utilizing class label information, thereby a new distance matrix is formed and the neighborhood is con- structed through this new distance matrix, at the same time, the neighborhood parameter are adaptive adjusted according to local cluster coefficient. With the low - dimensional feature as inputs of support vector machine （SVM） for identifying fault types. The experiment results of gear fault identification indicate that the proposed method can identification gear fault in high accuracy, it has some superiority.