基于LSTM网络在时变转速鼠笼电机断条诊断中的应用  

Application of LSTM Network in the Diagnosis of Time-Varying Speed Squirrel-Cage Motor Breaking

作　　者：李梦含 李垣江[1] 夏炎 王延波 LI Menghan;LI Yuanjiang;XIA Yan;WANG Yanbo(College of Electronic Information,Jiangsu University of Science and Technology,Zhenjiang 212003)

机构地区：[1]江苏科技大学电子信息学院,镇江212003

出　　处：《计算机与数字工程》2021年第10期2078-2082,共5页Computer & Digital Engineering

摘　　要：针对鼠笼电机在时变转速状态下运行时破坏了电机电流信号特征分析(Motor Current Signature Analysis,MCSA)的使用条件,使MCSA方法没办法诊断出时变情况下电机断条故障的问题。基于长短期记忆(Long Short-Term Memory,LSTM)网络提出一种高效准确的鼠笼电机断条诊断方法。首先通过采集故障鼠笼电机电流信号对LSTM网络进行训练,应用训练好的网络预估下一时间状态故障电机的电流值,然后通过对比采集信号和预估信号检测出故障,最后该方法通过时间域电流信号直接进行检测,并且从机器学习角度解决电机断条故障。结果表明,即使在短时数据条件依然能够诊断出早期断条故障。When the squirrel-cage motor is operated under the time-varying speed state,the condition of the motor current signal characteristic analysis(MCSA)is destroyed,so that the MCSA method cannot diagnose the problem of the motor broken strip in the case of time-varying.Based on long short-term memory(LSTM)network,an efficient and accurate diagnosis method for squirrel-cage motor broken bars is proposed.The method firstly trains the LSTM network by collecting the faulty squirrel motor current signal,the trained network is applied to estimate the current value of the next time state fault motor,and then the fault is detected by comparing the collected signal with the estimated signal.Finally,the method directly detects the current signal in the time domain,avoids frequency domain conversion and signal filtering,and solves the motor broken bar fault from the perspective of machine learning.The experimental results show that the validity and superiority of the proposed method can be verified even if the short-term data conditions can still diagnose the early broken faults.

关 键 词：鼠笼电机断条 LSTM网络 机器学习 故障诊断 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]