«Previous Article|Table of Contents|Next Article»

长安大学学报（自然科学版）[ISSN:1006-6977/CN:61-1281/TN]

Issue:

2025年5期

Page:

172-185

Research Field:

交通工程

Publishing date:

Info

Title:

Construction method of metro operation fault knowledge graph

Author(s):

HUANG Hai-lai¹; 2;  SONG Rui¹

(1. School of Traffic and Transportation, Beijing Jiaotong University, Beijing 100044, China; 2. Shanghai Shentong Metro Group Co., Ltd., Shanghai 201103, China)

Keywords:

Key words:traffic engineering;  metro operation fault;  natural language processing;  knowledge graph;  fault aided decision

PACS:

U298.5

DOI:

10.19721/j.cnki.1671-8879.2025.05.015

Abstract:

A substantial corpus of metro operation fault knowledge is recognized as being dispersed throughout the unstructured safety texts of metro operation enterprises and is difficult to be reused. To address this challenge, a method for constructing a metro operation fault knowledge graph based on natural language processing technology and deep learning was proposed. This approach was designed to facilitate the extraction of high-quality fault information and enable a more in-depth analysis, thereby providing a valuable resource for metro fault disposal and prevention. First, a top-down ontological model framework for metro operation faults was established, and a text classification algorithm combining the generalized autoregressive pretraining for language understanding(XLNet), bidirectional gated recurrent unit(BiGRU), and self-attention mechanism, along with a pre-trained language model, was employed to extract fault description paragraphs from metro safety texts. Then, based on the bidirectional encoder representations from transformers(BERT)-bidirectional long short term memory networks(BiLSTM)-conditional random field(CRF)model, fault entities and relationships within the description paragraphs were further extracted to construct a knowledge graph. The experimental results of the BERT-BiLSTM-CRF model, BiLSTM-CRF model, and BiLSTM model for entity extraction were compared using operation safety text corpora from a metro operation enterprise in a first-tier city in China. The proposed method was used to construct a knowledge graph of metro operation failures, and graph queries were conducted based on instances. The research results demonstrate that the BERT-BiLSTM-CRF model achieves high comprehensive performance in the task of metro fault entity extraction, with an average the harmonic mean of precision and recall(F₁)reaching to 0.75, while for line entity recognition, the F₁ reaches to 0.98. The proposed method can effectively acquire knowledge related to metro operation failures from metro safety texts and construct a knowledge graph, thereby enabling the query of associated relationships. This fully explores fault-related knowledge that is difficult to obtain through manual analysis, achieving high-level digital analysis of metro safety texts.4 tabs, 11 figs, 36 refs.

References:

[1] 黄慧建,范溢峻,苏晗翀,等.高温高湿环境下地铁列车辅助变流器高故障率的设备优化方案[J].城市轨道交通研究,2024,27(9):255-259.
HUANG Hui-jian, FAN Yi-jun, SU Han-chong, et al. Equipment optimization plan for high failure rate of metro train auxiliary converters in high temperature and high humidity environment[J]. Urban Rail Transit Research, 2024, 27(9): 255-259.
[2]张红星,刘 超,宋君君,等.某地铁列车网络系统中继器故障分析及改进[J].铁道机车车辆,2024,44(3):154-158.
ZHANG Hong-xing, LIU Chao, SONG Jun-jun, et al. Fault analysis and improvement of the repeater in a metro train network system[J]. Railway Locomotive & Car, 2024, 44(3): 154-158.
[3]占玉林,李嘉鑫,张秉鹤,等.考虑荷载类型及结构参数的轨道交通U形梁剪力滞效应[J].长安大学学报(自然科学版),2024,44(2):57-67.
ZHAN Yu-lin, LI Jia-xin, ZHANG Bing-he, et al. Shear lag effect of u-shaped beams in urban rail transit considering load types and structural parameters[J]. Journal of Chang'an University(Natural Science Edition), 2024, 44(2): 57-67.
[4]ZHOU Z P, GOH Y M, SHI Q Q, et al. Data-driven determination of collapse accident patterns for the mitigation of safety risks at metro construction sites[J]. Tunnelling and Underground Space Technology, 2022, 127: 1-17.
[5]汪益敏,罗 跃,于 恒,等.人员密集型地铁车站安全风险评价方法[J].交通运输工程学报,2020,20(5):198-207.
WANG Yi-min, LUO Yue, YU Heng, et al. Safety risk assessment method for metro stations with high passenger density[J]. Journal of Transportation Engineering, 2020, 20(5): 198-207.
[6]ZHOU Z P, LIU S, QI H N. Mitigating subway construction collapse risk using Bayesian network modeling[J]. Automation in Construction, 2022, 143: 104541.
[7]QIAO D S, ZHOU X B, YE X J, et al. Security risk assessment of submerged floating tunnel based on fault tree and multistate fuzzy Bayesian network[J]. Ocean and Coastal Management, 2024, 258: 107355.
[8]YANG Y H, LIU Y X, ZHOU M X, et al. Robustness assessment of urban rail transit based on complex network theory: A case study of the Beijing Subway[J]. Safety Science, 2015, 79: 149-162.
[9]ZHANG Y, HAN J Z, LIU J, et al. Safety prediction of rail transit system based on deep learning[C]//IEEE.2017 IEEE/ACIS 16th International Conference on Computer and Information Science(ICIS). New York: IEEE, 2017: 851-856.
[10]SHUN W, LV Y M,YUAN P, et al. Metro traffic flow prediction via knowledge graph and spatiotemporal graph neural network[J]. Journal of Advanced Transportation, 2022, 2022: 1-13.
[11]CHI H N, WANG B Y, GE Q B, et al. Knowledge graph-based enhanced transformer for metro individual travel destination prediction[J]. Journal of Advanced Transportation, 2022, 2022: 1-9.
[12]唐伟文,郭晟楠,陈 炜,等.融合时序知识图谱的路段级交通事故风险预测[J].模式识别与人工智能,2023,36(8):721-732.
TANG Wei-wen, GUO Sheng-nan, CHEN Wei, et al. Section-level traffic accident risk prediction with temporal knowledge graph fusion[J]. Pattern Recognition and Artificial Intelligence, 2023, 36(8): 721-732.
[13]齐如煜,尹章才,顾江岩,等.高精地图的知识图谱表达[J].武汉大学学报(信息科学版),2024,49(4):651-661.
QI Ru-yu, YIN Zhang-cai, GU Jiang-yan, et al. Knowledge graph representation of high-precision maps[J]. Geomatics and Information Science of Wuhan University, 2024, 49(4): 651-661.
[14]毛慧慧,赵小乐,杜圣东,等.基于时序知识图谱嵌入的短期地铁客流量预测[J].计算机科学,2023,50(7):213-220.
MAO H H, ZHAO X L, DU S D, et al. Short-term metro passenger flow prediction based on temporal knowledge graph embedding[J]. Computer Science, 2023, 50(7): 213-220.
[15]TANG J J, ZHANG J H, JIN J C, et al. Knowledge science, engineering and management[M]. Cham: Springer Nature Switzerland, 2023.
[16]XIU C, SUN Y C, PENG Q Y. Modelling traffic as multi-graph signals: Using domain knowledge to enhance the network-level passenger flow prediction in metro systems[J]. Journal of Rail Transport Planning & Amp; Management, 2022, 24: 100342.
[17]ZENG J, TANG J J. Combining knowledge graph into metro passenger flow prediction: A split-attention relational graph convolutional network[J]. Expert Systems with Applications, 2023, 213: 118790.
[18]WANG S, LV Y M, PENG Y, et al. Metro traffic flow prediction via knowledge graph and spatiotemporal graph neural network[J]. Journal of Advanced Transportation, 2022, 2022: 1-13.
[19]ZENG J, TANG J J. Combining knowledge graph into metro passenger flow prediction: A split-attention relational graph convolutional network[J]. Expert Systems with Applications, 2023, 213: 118790.
[20]林海香,胡娜娜,何 乔,等.基于建筑信息模型数据驱动的铁路设备运维多模态知识图谱构建[J].同济大学学报(自然科学版),2024,52(2):166-173.
LIN Hai-xiang, HU Na-na, HE Qiao, et al. Construction of multi-modal knowledge graph for railway equipment operation and maintenance based on building information modeling data[J]. Journal of Tongji University(Natural Sciences), 2024, 52(2): 166-173.
[21]CHEN F, YAN H, MA X P, et al. Construction and application of knowledge graph for urban rail fire accident[M]. Singapore: Springer Singapore, 2022.
[22]DIAO X R, WANG Y H, SUN W H, et al. Proceedings of the 6th international conference on electrical engineering and information technologies for rail transportation(EITRT)2023[M]. Singapore: Springer Nature Singapore, 2024.
[23]LIU Zi-yu, LI Ying, ZHAO Li-xia, et al. Construction of intelligent query system for metro electromechanical equipment faults based on the knowledge graph[J]. Journal of Intelligent & Amp, Fuzzy Systems, 2021, 41: 4351-4368.
[24]ZENG Y, QIN Y, LIU D, et al. Railway train device fault causality model based on knowledge graph[C]//IEEE. 2020 International Conference on Sensing, Diagnostics, Prognostics, and Control(SDPC). New York: IEEE, 2020: 1-8.
[25]LIU C,YANG S W. Using text mining to establish knowledge graph from accident/incident reports in risk assessment[J]. Expert Systems with Applications, 2022, 207: 117991.
[26]许 慧,李树秀,邢 镔.基于知识图谱的轨道交通运营风险事件智能分析研究[J].铁道标准设计,2024,68(8):34-42,49.
XU Hui, LI Shu-xiu, XING Bin. Intelligent analysis of urban rail transit operation risk events based on knowledge graph[J]. Railway Standard Design, 2024, 68(8): 34-42, 49.
[27]朱广宇,张 萌,裔 扬.基于知识图谱的城市轨道交通突发事件演化结果预测[J].电子与信息学报,2023,45(3):949-957.
ZHU Guang-yu, ZHANG Meng, YI Yang. Prediction of evolutionary outcomes of urban rail transit emergency events based on knowledge graph[J]. Journal of Electronics & Information Technology, 2023, 45(3): 949-957.
[28]HUO X,YIN Y,JIAO L D, et al. A data-driven and knowledge graph-based analysis of the risk hazard coupling mechanism in subway construction accidents[J]. Reliability Engineering & Amp, System Safety, 2024, 250: 110254.
[29]罗 丽.地铁运营安全事故知识图谱的构建及分析[D].北京:中国矿业大学,2022.
LUO Li. Construction and analysis of knowledge graph for metro operation safety accidents[D]. Beijing: China University of Mining and Technology, 2022.
[30]OLIAEE A H, DAS S, LIU J L, et al. Using bidirectional encoder representations from transformers(BERT)to classify traffic crash severity types[J]. Natural Language Processing Journal, 2023, 3: 100007.
[31]WANG YC. A Chinese text classification model based on XLNet and BiGRU[C]//IEEE. 2024 7th International Conference on Advanced Algorithms and Control Engineering(ICAACE). New York: IEEE, 2024: 337-341.
[32]RAHIM M A, RAHMAN M, ISLAM S, et al. Deep learning-based vehicular engine health monitoring system utilising a hybrid convolutional neural network/bidirectional gated recurrent unit[J]. Expert Systems with Applications, 2024, 257: 125080.
[33]刘建伟,刘俊文,罗雄麟.深度学习中注意力机制研究进展[J].工程科学学报,2021,43(11):1499-1511.
LIU Jian-wei, LIU Jun-wen, LUO Xiong-lin. Research progress of attention mechanism in deep learning[J]. Chinese Journal of Engineering Science, 2021, 43(11): 1499-1511.
[34]CUI Y M, ZHAO M. PERT: Pre-training BERT with permuted language model[EB/OL].[2025-09-04]. https://arxiv.org/abs/2203.06906.pdf.
[35]VASWANI, ASHISH, SHAZEER, et al. Attention is all you need[EB/OL]. [2025-09-04]. https://doi.org/10.48550/arXiv.1706.03762.
[36]朱 媛.基于多任务学习的序列标注式因果关系抽取[D].长春:吉林大学,2022.
ZHU Yuan. Causal relation extraction via sequence labeling based on multi-task learning[D]. Changchun: Jilin University, 2022.

Memo

Memo:

-

Common functions

Last Update: 2025-09-30