基于局部切空间排列和K-最近邻分类器的转子故障诊断方法

Rotor Fault Diagnosis Methods Based on Local Tangent Space Alignment and K-Nearest Neighbor

为了解决大型机械设备故障数据难以准确快速提取的问题,提出了一种基于局部切空间排列(LTSA)和K-最近邻分类器的转子故障诊断模型。首先基于转子的振动信号构造一个高维多征兆矩阵,利用LTSA提取高维矩阵的低维特征向量,映射在可视空间里;然后将提取的低维特征向量输入K-最近邻分类器进行故障模式识别。试验和数据降维仿真过程表明,该模型的准确度和快速性均优于LTSA和神经网络以及LTSA和支持向量机组成的故障诊断模型。

In order to solve the problem that the large mechanical equipment failure data is diffi- cult to accurately extract, this paper put forward a kind of rotor fault diagnosis models based on I.TSA and KNN. The vibration signals of rotor structure were used to construct dimensional matrix, then the low dimensional feature vector of high dimension matrix in the LTSA was extracted,and pro- jected into the visual space. And the extracted low dimensional feature vectors were put into the KNN in order to do fault pattern recognition. Finally, experimental and data dimension reduction simula tion process shows that the accuracy and rapidity of the method with LTSA and KNN are better than the fault diagnosis model by neural network and support vector machine.