logo

SCIENTIA SINICA Informationis, Volume 50, Issue 4: 511-526(2020) https://doi.org/10.1360/SSI-2019-0227

Research on the fault prediction method of an on-board subsystem in high-speed train control systems

More info
  • ReceivedOct 15, 2019
  • AcceptedJan 24, 2020
  • PublishedApr 13, 2020

Abstract

The train control system is the “nerve center” of the high-speed train information system, which is large-scale and comprises various components. The on-board subsystem is the core of the train control system and is key to ensuring traffic safety and improving operating efficiency. Currently, the fault data processing methods of the on-board subsystem remain manual, which primarily realizes the fault detection and location. It is difficult to achieve the fault mechanism level, and fault prediction cannot be realized effectively. In this paper, the on-board subsystem structure and the fault disposal status were analyzed. The existing problems have been summarized, and some concepts and algorithms to predict faults were introduced. Based on the on-board subsystem structure and each module's performance parameters, the system-level fault prediction model was established. Based on the practical operational data sets, the fault prediction based on the Bayesian network was carried out and verified under 20, 200, 2000 and 20000 sets, respectively. The prediction accuracy was 5%, 27%, 92% and 96.3%, respectively, under the condition of 2000 data sets. The hidden Markov model and neural network-based fault prediction solutions were compared with the proposed method. The results demonstrate the advanced performance of the Bayesian network-based solution in system-level fault prediction.


Funded by

国家自然科学基金重大项目(61490705)


References

[1] Wang H S, Wang J F. Key Technology of Train Control System of 200 350 km/h Trains. J China Railway Soc, 2009, 31: 107--110. Google Scholar

[2] He R, Ai B, Wang G. High-Speed Railway Communications: From GSM-R to LTE-R. IEEE Veh Technol Mag, 2016, 11: 49-58 CrossRef Google Scholar

[3] Han X, Tang T, Lv J D. Analysis of requirement-errors-caused failure of on-board subsystem of CTCS-3 train control system based on failure logs. J China Railway Soc, 2017, 39: 59--70. Google Scholar

[4] Pecht M, Jaai R. A prognostics and health management roadmap for information and electronics-rich systems. MicroElectron Reliability, 2010, 50: 317-323 CrossRef Google Scholar

[5] Ray A, Tangirala S. Stochastic modeling of fatigue crack dynamics for on-line failure prognostics. IEEE Trans Contr Syst Technol, 1996, 4: 443-451 CrossRef Google Scholar

[6] Li Y, Kurfess T R, Liang S Y. Stochastic Prognostics for Rolling Element Bearings. Mech Syst Signal Processing, 2000, 14: 747-762 CrossRef ADS Google Scholar

[7] Zhang S, Ganesan R. Multivariable Trend Analysis Using Neural Networks for Intelligent Diagnostics of Rotating Machinery. J Eng Gas Turbines Power, 1997, 119: 378-384 CrossRef Google Scholar

[8] Ali J B, Chebel M B, Saidi L, et al. Accurate bearing remaining useful life prediction based on Weibull distribution and artificial neural network. Mechanical Systems and Signal Processing, 2015, 56: 150-172. Google Scholar

[9] Liang X, Wang H F, Guo J. Bayesian network based fault diagnosis method for on-board equipment of train control system. Journal of the China Railway Society, 2017, 39(8):93-100. Google Scholar

[10] Zhao Y, Xu T H, Zhou Y P. Bayesian network based fault diagnosis system for vehicle on-board equipment of high-speed railway. Journal of the China Railway Society, 2014(11):48-53. Google Scholar

[11] Zhao Y, Xu T H. Text mining based fault diagnosis for vehicle on-board equipment of high speed railway signal system. J China Railway Soc, 2015, 8: 53--59. Google Scholar

[12] Wang Z X. CBR-based fault diagnosis system for vehicle equipment of high-speed railway. Dissertation for Master's Degree. Beijing: Beijing Jiaotong University, 2015. Google Scholar

[13] Chen X. The theory and method of fault discovery based on text information extraction for on-board vehicle equipment of high speed railway. Dissertation for Master's Degree. Beijing: Beijing Jiaotong University, 2017. Google Scholar

[14] Li Y. Application research on fault diagnosis of on-board equipment of train control system based on fuzzy theory and Bayesian network. Dissertation for Master's Degree. Lanzhou: Lanzhou Jiaotong University, 2015. Google Scholar

[15] Marcot B G. Common quandaries and their practical solutions in Bayesian network modeling. Ecological Modelling, 2017, 358: 1-9. Google Scholar

[16] Petitjean F, Buntine W, Webb G I. Accurate parameter estimation for Bayesian network classifiers using hierarchical Dirichlet processes. Mach Learn, 2018, 107: 1303-1331 CrossRef Google Scholar

[17] Hasan S, Ukkusuri S V. Reconstructing activity location sequences from incomplete check-in data: a semi-Markov continuous-time Bayesian network model. IEEE Transactions on Intelligent Transportation Systems, 2017, 19(3): 687-698. Google Scholar

[18] Cai B, Liu Y, Xie M. A Dynamic-Bayesian-Network-Based Fault Diagnosis Methodology Considering Transient and Intermittent Faults. IEEE Trans Automat Sci Eng, 2017, 14: 276-285 CrossRef Google Scholar

[19] Sun S, Zhang C, Yu G. A Bayesian Network Approach to Traffic Flow Forecasting. IEEE Trans Intell Transp Syst, 2006, 7: 124-132 CrossRef Google Scholar

[20] Xue-Wen Chen , Anantha G, Xiaotong Lin G. Improving Bayesian Network Structure Learning with Mutual Information-Based Node Ordering in the K2 Algorithm. IEEE Trans Knowl Data Eng, 2008, 20: 628-640 CrossRef Google Scholar

[21] Bańbura M, Modugno M. MAXIMUM LIKELIHOOD ESTIMATION OF FACTOR MODELS ON DATASETS WITH ARBITRARY PATTERN OF MISSING DATA. J Appl Econ, 2014, 29: 133-160 CrossRef Google Scholar

[22] Xu J L, Wang H J, Long B. Fault prediction of complex systems based on Bayesian network. Systems Engineering and Electronics, 2008, 30(4):780-784. Google Scholar

[23] Wu G, Qiu Y J, Wang G R. Map matching algorithm based on hidden Markov model and genetic algorithm. Journal of Northeastern University (Natural Science), 2017, 38(4):472-475. Google Scholar

  • Figure 1

    The structure of 300T on-board subsystem in train control system

  • Figure 2

    BN-based fault prediction

  • Figure 3

    (Color online) Faults produced in 2015 by different modules of on-board subsystem in train control system

  • Figure 4

    Fault model of on-board subsystem

  • Figure 5

    Simplified fault model of on-board subsystem

  • Figure 6

    (Color online) Result of BN structure learning. (a)–(d) Sample = 2, 20, 200, 2000, respectively

  • Figure 7

    (Color online) Results of signal test point parameter learning. (a) Energy carrier signal; (b) TCR working voltage; (c) SDU working voltage; (d) VDX working voltage; (e) radio operating frequency

  • Figure 8

    (Color online) Results of fault prediction based on the BN. (a)–(d) Sample = 20, 200, 2000, 20000, respectively

  • Figure 9

    Result of fault prediction based on the HMM

  • Figure 10

    (Color online) Result of fault prediction based on the NN. (a) Training confusion matrix; (b) validation confusion matrix; (c) testing confusion matrix; (d) final confusion matrix

  • Table 1   Test points in functional units
    Functional unit Test point Parameter
    (1) $t_1$ BTM working voltage
    $t_2$ Energy carrier signal
    $t_3$ Up-link signal
    $t_4$ Power sense signal
    (2) $t_5$ SDU working voltage
    $t_6$ ODO working voltage
    $t_7$ Radar working voltage
    (3) $t_8$ TCR working voltage
    $t_9$ Antenna frequency
    (4) $t_{10}$ VDX working voltage
    $t_{11}$ Mining voltage
    (5) $t_{12}$ STU-V working voltage
    $t_{13}$ Radio operating frequency
  • Table 2   Bayesian nodes and discrete values
    Discrete value State $t_2$ $t_5$ $t_8$ $t_{10}$ $t_{13}$ $F_1$
    1 Normal 27.095 (MHz) 18 (V) 110 (V) 24$\sim$120 (V) 915 (MHz) 1
    2 Deviation $[ - 5\% ,~ + 5\% ]$ $[ - 5\% ,~ + 5\% ]$ $[ - 30\% ,~ - 25\% ]$ $[ - 25\% ,~ + 30\% ]$ $[ - 5\% ,~ + 5\% ]$
    3 Abnormal $< - 5\% ,~ > 5\%$ $< - 5\% ,~ > 5\%$ $ < - 30\% ,~ > 25\% $ $< - 25\% ,~ > 30\%$ $< - 5\% ,~ > 5\%$ 2
  • Table 3   Predict accuracy under different algorithms
    Algorithm Prediction accuracy (2000 set) (%)
    Hidden Markov model 80.1
    Neural network 67.4
    Bayesian network 92

Copyright 2020 Science China Press Co., Ltd. 《中国科学》杂志社有限责任公司 版权所有

京ICP备18024590号-1       京公网安备11010102003388号