Incipient Fault Diagnosis Method of Vehicle Power Supply Based on Feature Optimization and Deep Learning

被引：0
|
作者：
Li W. [1 ,2 ,3 ]
Han Y. [1 ,2 ]
Sun X. [4 ]
机构：
[1] College of Electrical and Information Engineering, Lanzhou University of Technology, Gansu, Lanzhou
[2] Key Laboratory of Gansu Advanced Control for Industrial Processes, Gansu, Lanzhou
[3] National Demonstration Center for Experimental Electrical and Control Engineering Education, Lanzhou University of Technology, Gansu, Lanzhou
[4] Lanzhou Power Supply Vehicle Research Institute Co., Ltd., Gansu, Lanzhou
来源：
Binggong Xuebao/Acta Armamentarii | 2022年 / 43卷 / 11期
关键词：
incipient fault diagnosis; recursive feature elimination; stacked auto-encoder; vehicle power supply;
D O I：
10.12382/bgxb.2021.0577
中图分类号：
学科分类号：
摘要：
The vehicle power supply is the main power source for the training and emergency management of military equipment. The accurate diagnosis of incipient faults can effectively prevent the occurrence of serious faults. However, monitoring data is often redundant, making it difficult to effectively extract the symptoms of incipient faults. Aiming at this issue, a new intelligent incipient fault diagnosis method is proposed based on a combination of recursive feature elimination (RFE) and stacked auto-encoders (SAE). The collected feature variables are sorted based on their importance by using RFE to eliminate redundancy and extract the optimal subset of feature features. Then, the feature subset is used as the input of the SAE deep network, and the incipient fault category is used as the output, resulting in the effective diagnosis of incipient faults in vehicle power supply. The simulation results show that compared with SAE and shallow neural networks, the diagnosis accuracy of the proposed method is significantly improved, reaching 95.4%. © 2022 China Ordnance Society. All rights reserved.
引用
收藏
页码：2935 / 2944
页数：9
相关论文
共 30 条
  • [21] WANG S R., Research on early faults and evolution laws of doubly-fed induction generator in grid-connected op-erations, (2018)
  • [22] XU X W., Analysis and solution of frequency instability fault of cummins diesel generator set [J], Plant Mainte-nance Engineering, 4, pp. 19-20, (2012)
  • [23] ZHUANG Z S., Analysis and treatment measures of common faults of diesel generator set, Mechanical & Electrical Technique of Hydropower Station, 42, 4, pp. 58-61, (2019)
  • [24] DING L S, GU M, JIANG Y H., The relationshi between common faults of diesel engine and the governor [J], Journal of Agricultural Mechanization Research, 6, pp. 218-220, (2005)
  • [25] LU X X, ZHAO D S., Trouble judgment and removal for electronic speed-adjustor, Small Internal Combustion Engine and Vehicle Technique, 38, 3, pp. 93-96, (2009)
  • [26] GUYON I, WESTON J, BARNHILL S, Et al., Gene selection for cancer classification using support vector machines[J], Machine Learning, 46, 1, pp. 389-422, (2002)
  • [27] BENGIO Y., Learning deep architectures for AI[J], Founda-tions and Trends in Machine Learning, 2, 1, pp. 1-127, (2009)
  • [28] YUAN F N, ZHANG L, SHI J T, Et al., Theories and applications of auto-encode neural networks: a litera-ture survey, Chinese Journal of Computers, 42, 1, pp. 203-230, (2019)
  • [29] DING S C, LI X Y, HANG J, Et al., Deep learning theory and its application to fault diagnosis of an electric ma-chine, Power System Protection and Control, 48, 8, pp. 172-187, (2020)
  • [30] BENGIO Y, LAMBLIN P, POPOVICI D, Et al., Greedy layer-wise training of deep networks [J], Advances in Neural Information Processing Systems, 19, pp. 153-160, (2007)
123