多要素语义关联的TBM卡机风险评估知识图谱构建方法

朱庆; 王艳君; 丁雨淋; 伍庭晨; 陈诗楚; 黄琦雨; 王晓勇; 韩伟锋

doi:10.3969/j.issn.0258-2724.20250430

多要素语义关联的TBM卡机风险评估知识图谱构建方法

doi: 10.3969/j.issn.0258-2724.20250430

西南交通大学地球科学与工程学院，四川成都 611752

基金项目: 国家自然科学基金项目（42371436）

详细信息

作者简介:
朱庆（1966—），男，教授，博士，研究方向为三维GIS与数字孪生，E-mail：zhuqing@swjtu.edu.cn

通讯作者:
丁雨淋（1986—），女，教授，博士，研究方向为三维GIS与虚拟地理环境，E-mail：dingyulin@swjtu.edu.cn

中图分类号: P208
计量
- 文章访问数: 101
- HTML全文浏览量: 43
- PDF下载量: 25
- 被引次数: 0
出版历程
- 收稿日期: 2025-08-24
- 修回日期: 2025-12-01
- 刊出日期: 2025-12-11

A Knowledge Graph Construction Method for Tunnel Boring Machine Jamming Risk Assessment Based on Multi-Factor Semantic Associations

Faculty of Geosciences and Engineering, Southwest Jiaotong University, Chengdu 611752, China

摘要

摘要:
复杂山区全断面隧道掘进机（tunnel boring machine，TBM）施工常面临断层破碎带及围岩大变形等不良地质条件，易引发卡机事故. TBM卡机风险评估涉及监测数据、经验规则等多源异构信息，现有研究多侧重于评估数据或模型的局部建模，缺乏对关键评估要素及其语义关系的系统刻画，导致风险表达结构碎片化，难以支撑复杂工况下的动态评估与分析. 针对上述问题，提出一种多要素语义关联的TBM卡机风险评估知识图谱构建方法，该方法通过整合“评估任务-评估数据-评估参数-评估模型-评估指标”5类核心要素构建本体结构，实现风险评估过程的语义关联表达；结合大语言模型的分步提示策略，从风险评估文献、施工标准及工程案例中抽取显性与隐性知识，并通过消融实验验证提示策略的有效性；进一步利用 Neo4j 图数据库实现知识图谱的结构化存储与管理. 实验结果表明：在构建的标注数据集上，知识抽取精确率为88.54%，召回率为83.42%，F1值为85.67%，相较于仅使用单一提示策略具有明显提升；典型工程案例分析显示，所提方法能够有效组织卡机风险相关要素，在风险信息表达完整性与语义组织方面具有良好表现，为卡机风险的结构化表达与评估提供了有效支撑.
- TBM施工 /
- 卡机风险 /
- 知识图谱 /
- 大语言模型 /
- 本体构建
Abstract:
In complex mountainous areas, tunnel boring machine (TBM) construction often faces adverse geological conditions, such as fault fracture zones and large deformation of surrounding rock, which can easily cause jamming accidents. TBM jamming risk assessment involves multi-source heterogeneous information such as monitoring data and empirical rules. Existing studies mostly focus on the local modeling of assessment data or models and lack a systematic characterization of key assessment factors and their semantic associations, leading to a fragmented risk representation structure, which makes it difficult to support dynamic assessment and analysis under complex conditions. To address the above issues, a construction method of a knowledge graph for TBM jamming risk assessment based on multi-factor semantic associations was proposed. This method constructed an ontology structure by integrating five types of core factors: "assessment task, assessment data, assessment parameter, assessment model, and assessment indicator", achieving semantic association representation of the risk assessment process. Combined with a stepwise prompting strategy of a large language model, explicit and implicit knowledge was extracted from risk assessment literature, construction standards, and engineering cases, and the effectiveness of the prompting strategy was verified through ablation experiments. Furthermore, the Neo4j graph database was utilized to realize structured storage and management of the knowledge graph. The experimental results indicate that on the constructed annotated dataset, the precision, recall, and F1 score of knowledge extraction are 88.54%, 83.42%, and 85.67%, respectively, showing a significant improvement compared with using a single prompting strategy. The analysis of typical engineering cases shows that the proposed method can effectively organize factors related to jamming risk, exhibits a good performance in the completeness of risk information representation and semantic organization, and provides effective support for the structured representation and assessment of jamming risk.
- TBM tunneling /
- jamming risk /
- knowledge graph /
- large language model /
- ontology construction

HTML全文

图 1 TBM卡机风险评估知识图谱构建流程

Figure 1. Construction process of knowledge graph for TBM jamming risk assessment

下载: 全尺寸图片幻灯片

图 2 TBM卡机风险评估要素知识体系（部分）

Figure 2. Knowledge system of TBM jamming risk assessment factors (partial)

下载: 全尺寸图片幻灯片

图 3 分步提示策略实现流程

Figure 3. Implementation process of stepwise prompting strategy

下载: 全尺寸图片幻灯片

图 4 Neo4j中卡机风险知识图谱的节点关系组织结构

Figure 4. Organization structure of nodes and relationships of jamming risk knowledge graph in Neo4j

下载: 全尺寸图片幻灯片

图 5 人工标注文本数据示例

Figure 5. Example of manually annotated text data

下载: 全尺寸图片幻灯片

图 6 TBM卡机风险图谱实例层结构（部分）

Figure 6. Instance layer structure of TBM jamming risk knowledge graph (partial)

下载: 全尺寸图片幻灯片

图 7 TBM卡机风险知识图谱可视化应用流程

Figure 7. Visualization application process of TBM jamming risk knowledge graph

下载: 全尺寸图片幻灯片

表 1 定义概念间关系

Table 1. Definition of relationships between concepts

关系类型	说明
Has_Task	表示任务与子任务之间的关系，描述不同的卡机风险事件
Required_Model	描述任务与模型之间的关系，表明任务使用或关联某个模型
Required_Data	表示任务与数据之间的关系，描述某个任务使用的数据
Required_ Parameter	表示任务与参数之间的关系，描述某个任务使用的数据参数
Required_Metric	表示任务与指标之间的关系，指示某个任务使用的评估指标
Mapping_to	表示输出参数与输入参数之间的一对一映射关系
Consist _of	表示输出参数与输入参数之间的一对多关系
Input	通常用于定义模型所需的输入参数
Output	表示模型生成的结果或输出参数
Subclass_of	表示实体间的上下级关系
Property_of	表示实体的属性信息
Instance_of	表示某个实例属于某个类，描述具体实例与类的隶属关系

下载: 导出CSV

表 2 基于规则的半结构化数据信息提取与融合（示例）

Table 2. Rule-based extraction and integration of semi-structured data information (example)

类别	实体	原始表述	匹配规则设计	处理操作
地质术语	软弱带	软弱夹层、不良地质体、软弱地带	模糊匹配关键词：软弱、夹层、不良地质	实体融合统一为软弱带
卡机原因分类	地质因素卡机	地质因素、地质条件卡机、地质卡机	上下文关键词卡机前缀包含地质相关词汇	分类归一化
属性名称冲突	施工区段	工区、盾构区间、施工段	标准化映射表：工区定义为施工区段；盾构区间定义为施工区间	属性名对齐
属性值冲突	推进速度	日进尺、掘进速度	正则替换单位统一为 m/d	单位转换 + 标准化

下载: 导出CSV

表 3 消融实验结果

Table 3. Results of ablation experiment %

提示策略	P	R	$\rho_{\mathrm{F1}} $
需求提示	72.13	65.45	68.63
需求提示 + 领域知识提示	81.27	75.91	78.48
需求提示 + 样例提示	86.35	81.22	83.70
完整的三步法提示策略	88.54	83.42	85.67

下载: 导出CSV

参考文献(34)

[1]	王亚锋. 高黎贡山隧道TBM不良地质条件下卡机脱困施工关键技术[J]. 隧道建设(中英文), 2021, 41(3): 441-448. doi: 10.3973/j.issn.2096-4498.2021.03.013 WANG Yafeng. Key technologies for TBM jamming release of Gaoligongshan tunnel under unfavorable geologies[J]. Tunnel Construction, 2021, 41(3): 441-448. doi: 10.3973/j.issn.2096-4498.2021.03.013
[2]	王文扬. TBM施工穿越破碎带卡机倾向性智能预测及工程应用[D]. 济南: 山东大学, 2023.
[3]	汪文强, 张新平, 宋振东. 数值模拟在高黎贡山隧道TBM施工段断层破碎带TSP探测中的应用[J]. 工程地球物理学报, 2020, 17 (6): 780-86. WANG Wenqiang, ZHANG Xinping, SONG Zhendong. Application of numerical simulation in TSP detection of fault fracture zone in TBM construction section of Gaoligongshan tunnel [J]. Journal of Engineering Geophysics, 2020, 17 (6): 780-786.
[4]	徐鹏, 黄俊, 周剑波, 等. 护盾式TBM穿越断层破碎带岩机相互作用三维数值模拟研究[J]. 现代隧道技术, 2020, 57(6): 63-69. doi: 10.13807/j.cnki.mtt.2020.06.008 XU Peng, HUANG Jun, ZHOU Jianbo, et al. Three-dimensional numerical simulation of rock-machine interaction during shield TBM tunnelling through fault fracture zones[J]. Modern Tunnelling Technology, 2020, 57(6): 63-69. doi: 10.13807/j.cnki.mtt.2020.06.008
[5]	邓鹏海. 深部软弱地层TBM掘进挤压变形卡机及防控过程FDEM数值模拟研究[D]. 武汉: 武汉大学, 2019.
[6]	陈湘生, 曾仕琪, 韩文龙, 等. 机器学习方法在盾构隧道工程中的应用研究现状与展望[J]. 土木与环境工程学报(中英文), 2022, 46(1): 1-13. doi: 10.11835/j.issn.2096-6717.2022.069 CHEN Xiangsheng, ZENG Shiqi, HAN Wenlong, et al. Machine learning methods in shield tunneling engineering: a state-of-the-art review and prospects[J]. Journal of Civil and Environmental Engineering, 2022, 46(1): 1-13. doi: 10.11835/j.issn.2096-6717.2022.069
[7]	HASANPOUR R, ROSTAMI J, SCHMITT J, et al. Prediction of TBM jamming risk in squeezing grounds using bayesian and artificial neural networks[J]. Journal of Rock Mechanics and Geotechnical Engineering, 2020, 12(1): 21-31. doi: 10.1016/j.jrmge.2019.04.006
[8]	GUO D, MENG F, WU H, et al. Risk assessment of shield tunneling crossing building based on variable weight theory and cloud model[J]. Tunnelling and Underground Space Technology, 2024, 145: 105593. doi: 10.1016/j.tust.2024.105593
[9]	FOULADGAR M M, YAZDANI-CHAMZINI A, ZAVADSKAS E K. Risk evaluation of tunneling projects[J]. Archives of Civil and Mechanical Engineering, 2012, 12: 1-12.
[10]	PAULHEIM H. Knowledge graph refinement: a survey of approaches and evaluation methods[J]. Semantic Web, 2017, 8(3): 489-508. doi: 10.3233/sw-160218
[11]	ZHONG L, WU J, LI Q, et al. A comprehensive survey on automatic knowledge graph construction[J]. ACM Computing Surveys, 2023, 56(4): 1-62. doi: 10.1145/3618295
[12]	张雪英, 张春菊, 吴明光, 等. 顾及时空特征的地理知识图谱构建方法[J]. 中国科学: 信息科学, 2020, 50(7): 1019-1032. ZHANG Xueying, ZHANG Chunju, WU Mingguang, et al. Spatiotemporal features based geographical knowledge graph construction[J]. Scientia Sinica (Informationis), 2020, 50(7): 1019-1032.
[13]	朱庆, 王所智, 丁雨淋, 等. 铁路隧道钻爆法施工智能管理的安全质量进度知识图谱构建方法[J]. 武汉大学学报(信息科学版), 2022, 47(8): 1155-1164. ZHU Qing, WANG Suozhi, DING Yulin, et al. A method of safety-quality-schedule knowledge graph for intelligent management of drilling and blasting construction of railway tunnels[J]. Geomatics and Information Science of Wuhan University, 2022, 47(8): 1155-1164.
[14]	车万翔, 窦志成, 冯岩松, 等. 大模型时代的自然语言处理: 挑战、机遇与发展[J]. 中国科学: 信息科学, 2023, 53(9): 1645-1687 doi: 10.1360/SSI-2023-0113 CHE Wanxiang, DOU Zhicheng, FENG Yansong, et al. Towards a comprehensive understanding of the impact of large language models on natural language processing: challenges, opportunities and future directions[J]. Scientia Sinica (Informationis), 2023, 53(9): 1645-1687. doi: 10.1360/SSI-2023-0113
[15]	许强, 崔圣华, 黄维, 等. 面向工程地质领域的滑坡知识图谱构建方法研究[J]. 武汉大学学报(信息科学版), 2023, 48(10): 1601-1615. doi: 10.13203/j.whugis20230245 XU Qiang, CUI Shenghua, HUANG Wei, et al. Construction of a landslide knowledge graph in the field of engineering geology[J]. Geomatics and Information Science of Wuhan University, 2023, 48(10): 1601-1615. doi: 10.13203/j.whugis20230245
[16]	ABU-SALIH B. Domain-specific knowledge graphs: a survey[J]. Journal of Network and Computer Applications, 2021, 185(5): 103076. doi: 10.1016/j.jnca.2021.103076
[17]	PAN S R, LUO L H, WANG Y F, et al. Unifying large language models and knowledge graphs: a roadmap[J]. IEEE Transactions on Knowledge and Data Engineering, 2024, 36(7): 3580-3599. doi: 10.1109/TKDE.2024.3352100
[18]	王雅琳, 邹江枫, 王凯, 等. 基于本体引导的注塑知识图谱构建及缺陷溯因应用[J]. 电子与信息学报, 2022, 44(5): 1521-1529. doi: 10.11999/JEIT211416 WANG Yalin, ZOU Jiangfeng, WANG Kai, et al. Injection molding knowledge graph based on ontology guidance and its application to quality diagnosis[J]. Journal of Electronics and Information Technology, 2022, 44(5): 1521-1529. doi: 10.11999/JEIT211416
[19]	HOGAN A, BLOMQVIST E, COCHEZ M, et al. Knowledge graphs[J]. ACM Computing Surveys, 2021, 54(4): 1-37.
[20]	杜志强, 李钰, 张叶廷, 等. 自然灾害应急知识图谱构建方法研究[J]. 武汉大学学报(信息科学版), 2020, 45(9): 1344-1355. DU Zhiqiang, LI Yu, ZHANG Yeting, et al. Knowledge graph construction method on natural disaster emergency[J]. Geomatics and Information Science of Wuhan University, 2020, 45(9): 1344-1355.
[21]	许振浩, 邵瑞琦, 林鹏, 等. 隧道不良地质识别: 方法、现状及智能化发展方向[J]. 地球学报, 2024, 45(1): 5-24. doi: 10.3975/cagsb.2023.030702 XU Zhenhao, SHAO Ruiqi, LIN Peng, et al. Adverse geology identification in tunnel: method, research status and intelligent development direction[J]. Acta Geoscientica Sinica, 2024, 45(1): 5-24. doi: 10.3975/cagsb.2023.030702
[22]	张娜, 李建斌, 荆留杰, 等. 基于隧道掘进机掘进过程的岩体状态感知方法[J]. 浙江大学学报(工学版), 2019, 53(10): 1977-1985. doi: 10.3785/j.issn.1008-973X.2019.10.015 ZHANG Na, LI Jianbin, JING Liujie, et al. Prediction method of rockmass parameters based on tunnelling process of tunnel boring machine[J]. Journal of Zhejiang University (Engineering Science), 2019, 53(10): 1977-1985. doi: 10.3785/j.issn.1008-973X.2019.10.015
[23]	刘泉声, 黄兴, 刘建平, 等. 深部复合地层围岩与TBM的相互作用及安全控制[J]. 煤炭学报, 2015, 40(6): 1213-1224. doi: 10.3772/j.issn.1009-5659.2018.22.014 LIU Quansheng, HUANG Xing, LIU Jianping, et al. Interaction and safety control between TBM and deep mixed ground[J]. Journal of China Coal Society, 2015, 40(6): 1213-1224. doi: 10.3772/j.issn.1009-5659.2018.22.014
[24]	刘斌, 李术才, 李建斌, 等. TBM 掘进前方不良地质与岩体参数的综合获取方法[J]. 山东大学学报(工学版), 2016, 46(6): 105-112. doi: 10.6040/j.issn.1672-3961.0.2016.478 LIU Bin, LI Shucai, LI Jianbin, et al. Integrated acquisition method of adverse geology and rock properties ahead of tunnel face in TBM construction tunnel[J]. Journal of Shandong University (Engineering Science), 2016, 46(6): 105-112. doi: 10.6040/j.issn.1672-3961.0.2016.478
[25]	张骏骁. 语义约束的滑坡灾情评估优势信息智能选取方法[D]. 成都: 西南交通大学, 2019.
[26]	WANG Y, CHENG Y, QI Q, et al. IDS-KG: an industrial dataspace-based knowledge graph construction approach for smart maintenance[J]. Journal of Industrial Information Integration, 2024, 38: 100566. doi: 10.1016/j.jii.2024.100566
[27]	POLAK M P, MORGAN D. Extracting accurate materials data from research papers with conversational language models and prompt engineering[J]. Nature Communications, 2024, 15(1): 1569. doi: 10.1038/s41467-024-45914-8
[28]	ZHU Y, WANG X, CHEN J, et al. LLMs for knowledge graph construction and reasoning: recent capabilities and future opportunities[J]. World Wide Web, 2024, 27(5): 58. doi: 10.1007/s11280-024-01297-w
[29]	HENDRIK S, ALBERT W, VICTOR S, et al. Interactive and visual prompt engineering for adhoc task adaptation with large language models[J]. IEEE Transactions on Visualization and Computer Graphics, 2022, 29(1): 1146-1156. doi: 10.1109/tvcg.2022.3209479
[30]	LIU P, YUAN W, FU J, et al. Pre-train, prompt, and predict: a systematic survey of prompting methods in natural language processing[J]. ACM Computing Surveys, 2023, 55(9): 1-35. doi: 10.1145/3560815
[31]	王淑营, 李雪, 黎荣, 等. 基于知识图谱的高速列车知识融合方法[J]. 西南交通大学学报, 2024, 59(5): 1194-1203. doi: 10.3969/j.issn.0258-2724.20220193 WANG Shuying, LI Xue, LI Rong, et al. Knowledge graph-based knowledge fusion method for high-speed trains[J]. Journal of Southwest Jiaotong University, 2024, 59(5): 1194-1203. doi: 10.3969/j.issn.0258-2724.20220193
[32]	LING J, LI X, LI H, et al. Hybrid NLP-based extraction method to develop a knowledge graph for rock tunnel support design[J]. Advanced Engineering Informatics, 2024, 62: 102725. doi: 10.1016/j.aei.2024.102725
[33]	朱庆, 李茂粟, 丁雨淋, 等. 滑坡灾情数据多层级语义检索方法[J]. 西南交通大学学报, 2020, 55(3): 467-475 doi: 10.3969/j.issn.0258-2724.20180695 ZHU Qing, LI Maosu, DING Yulin, et al. Multi-level semantic retrieval method for landslide disaster data[J]. Journal of Southwest Jiaotong University, 2020, 55(3): 467-475. doi: 10.3969/j.issn.0258-2724.20180695
[34]	LEWIS P, PEREZ E, PIKTUS A, et al. Retrieval-augmented generation for knowledge-intensive NLP tasks[J]. Advances in Neural Information Processing Systems, 2020, 33: 9459-9474.