基于1-DISVM的无刷直流电动机故障识别方法

Fault-recognition Method for BLDCM Based on Improved Unsupervised Learning Support Vector Machines

提出一种基于改进无监督学习支持向量机(1-DISVM)的无刷直流电动机故障识别方法。通过对无刷直流电动机正常以及驱动电路开关管断路、定子绕组端部断路、Hall传感器断线三种故障状态的仿真模拟,对仿真过程中得到的母线电流采样数据进行FFT频谱分析,作为输入特征向量用于支持向量机分类器的训练和故障识别。将改进无监督学习支持向量机用于无刷直流电动机的故障识别,并与无监督学习支持向量机(1-SVM)的故障识别结果进行比较,结果表明基于改进无监督学习支持向量机的无刷直流电动机故障识别方法具有更高的准确率。

This paper put forward a fault recognition method for BLDCM based on improved unsupervised learning support vector machines - 1-DISVM. Some simulation experiments were carried out around inverter system＇s single switch tube broken, Hall position sensor＇s wire broken, stator winding broken. The signals of the sampled bus current during the experiments were analyzed with FFT and then taken as the inputs for the classification algorithm＇s training and fault-recognition. By the comparison of the fault-recognition results be- tween 1-DISVM and common unsupervised learning support vector machines - 1-SVM, the result shows that 1-DISVM has higher precision in BLDCM＇s fault recognition.