摘要
论文提出了一种基于长短期记忆网络(LSTM)的动车组牵引电机轴承健康状态评估与预测的方法.首先在动车组牵引电机轴承的时域、频域以及时频域内,进行退化特征提取,并以相关性、单调性和鲁棒性为原则,进行敏感特征选取;然后,采用深度学习网络,对所选取的退化特征进行特征融合,在此基础上,计算最小量化误差(MQE)并构建健康指数(HI)退化曲线;最后,基于LSTM的动车组牵引电机轴承的健康状态,进行评估与预测,同时运用人工神经网络(ANN),对牵引电机轴承的健康状态进行评估预测.进而并对两种评估预测的结果进行比较分析,结果表明:采用LSTM的评估预测精度较ANN的高,而且性能更优,更适合于动车组牵引电机健康状态的评估与预测.
In response to the difficulty in assessing and predicting the bearing health status of the multiple unit train traction motor,this paper proposes a method to assess and predict the bearing health status of the multiple unit train transaction motor based on the long and short-term memory(LSTM).According to this method,the time domain and frequency domain of the multiple unit train traction motor bearing,and the degeneration characteristics of the multiple unit train traction motor bearing within the time and frequency domain are extracted.Meanwhile,the sensitive characteristics are selected with relevance,monotony and robustness as the principles.Secondly,the deep learning network is used to pursue characteristic integration of the selected degeneration characteristics.On that basis,the minimum quantization error is computed,and the health index degenerated curve is constructed.At last,the bearing health status of the multiple unit train traction motor is assessed and predicted based on the LSTM while the artificial neural network(ANN)is used to assess and predict the bearing health status of the multiple unit train traction motor.Moreover,the assessment and prediction results by the two different methods are comparatively analyzed.The results indicate that use of the LSTM for assessment and prediction is more accurate than ANN,and has better performance,which is more appropriate for the evaluation and prediction of the health state of traction motor for EMU.
作者
胡文涛
孟建军
HU Wen-tao;MENG Jian-jun(Mechatronics T&R Institute,Lanzhou Jiaotong University,Lanzhou 730070,China)
出处
《兰州交通大学学报》
CAS
2021年第2期95-100,共6页
Journal of Lanzhou Jiaotong University
基金
甘肃省高等学校科研项目:(2017D-09)(2018C-10)。
关键词
车辆工程
健康状态评估
长短期记忆网络
轴承
vehicle engineering
health status
assessment
long-short term memory
bearing
