FMM神经网络结合CART的感应电动机故障诊断方法  

A Scheme of Fault Diagnosis of Induction Motors Based on Hybrid FMM Neural Network and CART

作　　者：周柏清[1] 王刘涛[2] 任勇军[3] 

机构地区：[1]湖州职业技术学院信息工程分院,浙江湖州313000 [2]平顶山学院软件学院,河南平顶山467000 [3]南京信息工程大学计算机与软件学院,南京210044

出　　处：《电子器件》2016年第4期993-999,共7页Chinese Journal of Electron Devices

基　　金：国家自然科学基金项目(61300236)

摘　　要：针对感应电动机存在多种故障问题,提出一种融合模糊极小-极大(FMM)神经网络和分类回归树(CART)的电机故障诊断方法(FMM-CART),对转子断条、定子绕组和电压失衡三种常见电机故障进行诊断。通过采集电机三相的电流信号,并进行功率谱分析,提取特定谐波信号作为FMM-CART模型的输入特征。训练过的FMM神经网络根据输入特征计算置信因子,CART根据置信因子构建决策树,最终输出诊断结果。实验结果表明,FMM-CART能有效的诊断各种电机故障,且具有较少的检测时间和较低的网络复杂度。For the issues that there are many faults in induction motor, a scheme of fault diagnosis of induction mo- tors based on hybrid FMM neural network and CART is proposed, which is used to diagnose the rotor broken bars, stator winding and the voltage unbalance fault. It collects the current signal of three-phase, and makes the power spectrum analysis, selects the specific harmonic signal as the input of FMM-CART model. The trained FMM neural network calculates the confidence factors according to the input feature, CART builds decision tree according to the confidence factor. The experimental results show that, FMM-CART can diagnose all kinds of motor fault effectively, and has less time and lower the complexity of the network.

关 键 词：感应电动机故障诊断 三相电流信号 模糊极小-极大神经网络 分类回归树 功率谱分析 

分 类 号：TM346[电气工程—电机]