| 160 | 2 | 112 |
| 下载次数 | 被引频次 | 阅读次数 |
目前转向架故障诊断研究仅仅考虑了较为理想的列车运行状态，未考虑车轮多边形的影响，而车轮多边形激励的存在能显著降低故障诊断的精度。为此，提出了一种基于LSTM网络和卷积神经网络的深度学习方法。该方法由LSTM以及一维卷积神经网络组成，并引入了注意力机制用于强调训练数据特征，提高转向架故障信号识别的准确率。以CRH380动车组转向架为例，引入车轮多边形激励，采用所提方法对不同工况下的关键部件故障进行了分类。通过与已有方法对比发现，所提方法由于引入了注意力机制，较大程度克服了车轮多边形带来的噪声干扰，提高了转向架故障诊断的精度。
Abstract：Current research on bogie fault diagnosis only considers ideal train operating conditions, without considering the influence of wheel polygons, the excitation of which significantly reduces the accuracy of fault diagnosis. For this reason, a deep learning method based on LSTM network and convolutional neural network is proposed. The method consists of a long short-term memory network as well as a one-dimensional convolutional neural network, and introduces an attention mechanism to emphasize the features of training data, thereby improving the accuracy of bogie fault signal recognition. Taking the bogie of CRH380 rolling stock as an example, the wheel polygon excitation is introduced, and the proposed method is used to classify the key component faults under different working conditions. By comparing with the existing methods, it is found that the proposed method, due to the introduction of the attention mechanism, overcomes the noise interference caused by wheel polygons to a greater extent, and improves the accuracy of bogie fault diagnosis.
[1]粟丽源.基于卷积神经网络的高速列车转向架故障诊断方法研究[D].成都：西南交通大学，2019.
[2]梁开伟.基于多重卷积循环神经网络的高速列车转向架故障诊断方法研究[D].成都：西南交通大学，2020.
[3]ZUO Y,WANG N,JIA L,et al. Fully Decomposed Singular Value and Fixed Dictionary Extreme Learning Machine for Bogie Fault Diagnosis[J]. IEEE Transactions on Intelligent Transportation Systems,2022,23(8):10262-10274.
[4]LI Y,JIN W. Application of Full Vector IMF Entropy in Fault Diagnosis of High Speed Train[J]. Journal of Vibration,Measurement and Diagnosis,2021,41(5):874-879.
[5]秦娜，金炜东，黄进.基于EEMD的高速列车转向架故障诊断[J].计算机工程，2013,39(12):1-4.
[6]WANG Z,JIA L,KOU L,et al. Spectral Kurtosis Entropy and Weighted SaE-ELM for Bogie Fault Diagnosis under Variable Conditions[J]. Sensors(Basel, Switzerland),2018,18(6):1705.
[7]HUANG D,LI S,QIN N,et al. Fault Diagnosis of High-Speed Train Bogie Based on the Improved-CEEMDAN and 1-D CNN Algorithms[J]. IEEE Transactions on Instrumentation and Measurement,2021(70):1-11.
[8]王忆佳，曾京，罗仁.高速列车车轮多边形化对车辆动力学性能的影响[J].四川大学学报（工程科学版），2013,45(3):176-182.
[9]王红兵，李国芳，王泽根.车轮多边形对车辆动力学性能影响分析[J].铁道标准设计，2020,64(6):165-171.
[10]尹振坤，吴越，韩健.高速列车车轮多边形磨耗对轮轨垂向力的影响[J].铁道学报，2017,39(10):26-32.
[11]POPP. System dynamics of railway vehicles and track[J].ARCHIVE OF APPLIED MECHANICS,2003,72(11-12):949-961.
[12]李凤林，杜红梅，樊懿葳.基于形态学滤波的车轮多边形故障诊断方法[J].机械，2023,50(8):39-46.
[13]高润.列车车轮失圆动态检测方法与技术研究[D].北京：北京交通大学，2021.
[14]朱海燕，胡华涛，尹必超.轨道车辆车轮多边形研究进展[J].交通运输工程学报，2020,20(1):102-119.
[15]赵新利，吴越，郭涛.车轮多边形磨耗统计规律及关键影响因素分析[J].振动测试与诊断，2020,40(1):48-53.
[16]胡晓依.动车组车轮多边形磨耗形成与发展过程仿真研究[J].中国铁道科学，2021,42(2):107-115.
[17]翟婉明.车辆-轨道耦合动力学Vehicle-track coupling dynamics[M].北京：科学出版社，2007.
[18]HOCHREITER S, SCHMIDHUBER J. Long Short-Term Memory[J]. Neural Computation,1997,9(8):1735-1780.
[19]ALEX. Supervised Sequence Labelling with Recurrent Neural Networks[M]. Berlin:Springer,2012.
[20]VIET. Skin Lesion Classification on Imbalanced Data Using Deep Learning with Soft Attention[J]. SENSORS,2022,22(19):7530.
[21]LIU J W,LIU J W,LUO X L. Research progress in attention mechanism in deep learning[J]. Chinese Journal of Engineering,2021,43(11):1499-1511.
基本信息:
DOI：
中图分类号:U279
引用信息:
[1]张懿,杨旭锋,方修洋.一种考虑车轮多边形的转向架故障诊断深度学习方法[J].机械,2024,51(09):37-44.
基金信息:
中央高校基本科研业务费专项资金（2682022ZTPY079）; 四川省科技计划项目（2021YFG0178）