基于大语言模型的数字孪生模型语义重构方法

doi:10.16183/j.cnki.jsjtu.2025.240

上海交通大学学报 ›› 2026, Vol. 60 ›› Issue (8): 1290-1298.doi: 10.16183/j.cnki.jsjtu.2025.240

基于大语言模型的数字孪生模型语义重构方法

卞文超^a, 温晓健^a, 王新厚^b, 鲍劲松^a()

^a 机械工程学院, 东华大学上海 201620
^b 纺织学院, 东华大学上海 201620

收稿日期:2025-07-17 修回日期:2025-10-10 接受日期:2025-11-12 出版日期:2026-08-28 发布日期:2026-09-02
通讯作者: 鲍劲松,教授,博士生导师,电话(Tel.): 021-67792583;E-mail:bao@dhu.edu.cn.
作者简介:卞文超(2000—),硕士生,从事智能制造研究.
基金资助:
国家自然科学基金(52475513)

Semantic Reconstruction Method of Digital Twin Model Based on Large Language Model

BIAN Wenchao^a, WEN Xiaojian^a, WANG Xinhou^b, BAO Jinsong^a()

^a College of Mechanical Engineering Donghua University, Shanghai 201620, China
^b College of Textiles, Donghua University, Shanghai 201620, China

Received:2025-07-17 Revised:2025-10-10 Accepted:2025-11-12 Online:2026-08-28 Published:2026-09-02

摘要/Abstract

摘要：

信息技术的快速发展和工业4.0的深入推进,正驱动数字孪生技术在制造业智能化转型中的广泛应用.然而现有制造业数字孪生模型在语义层面仍存在局限性,表现为对物理资产的多维属性描述不够全面,动态交互能力不足且缺乏统一的语义标准规范.为此,提出一种基于大语言模型的数字孪生模型语义重构方法.该方法利用数字孪生定义语言构建语义信息资产库,并融合大语言模型与知识图谱技术,实时生成和补充资产语义信息,实现数字孪生的动态语义重构.以活塞加工产线为对象进行试验验证,结果表明,相较于传统手动建模方法,基于大语言模型的语义重构方法在处理复杂任务时效率与稳定性显著提升,能够快速响应动态变化的生产需求,并在可视化展示方面具有优势.

关键词: 数字孪生, 语义重构, 大语言模型, 数字孪生定义语言, 知识图谱

Abstract:

The rapid development of information technology and the deep advancement of Industry 4.0 are driving the widespread application of digital twin technology in the intelligent transformation of manufacturing. However, existing digital twin models in manufacturing still have limitations at the semantic level, as manifested in inadequate multidimensional attribute descriptions of physical assets, insufficient dynamic interaction capabilities, and lack of unified semantic standards and specifications. To address these issues, this paper proposes a semantic reconstruction method for digital twin models based on large language models (LLMs). This method utilizes a digital twin definition language to build a semantic information asset library and integrates large language models with knowledge graph technology to generate and enrich asset semantic information in real time, thereby enabling dynamic semantic reconstruction of digital twins. The experimental results focusing on a piston processing production line demonstrate that, compared to traditional manual modeling methods, the proposed semantic reconstruction method based on LLM significantly improves efficiency and stability in handling complex tasks, respond rapidly to dynamically changing production demands, and has advantages in visual presentation.

Key words: digital twin, semantic reconstruction, large language model (LLM), digital twin definition language (DTDL), knowledge graph

中图分类号:

TH181

卞文超, 温晓健, 王新厚, 鲍劲松. 基于大语言模型的数字孪生模型语义重构方法[J]. 上海交通大学学报, 2026, 60(8): 1290-1298.

BIAN Wenchao, WEN Xiaojian, WANG Xinhou, BAO Jinsong. Semantic Reconstruction Method of Digital Twin Model Based on Large Language Model[J]. Journal of Shanghai Jiao Tong University, 2026, 60(8): 1290-1298.

图/表 11

图1

图2

图3

图4

表1

表2

表3

表4

图5

图6

图7

参考文献 18

[1]	LUO D, THEVENIN S, DOLGUI A. A state-of-the-art on production planning in Industry 4.0[J]. International Journal of Production Research, 2023, 61(19): 6602-6632. doi: 10.1080/00207543.2022.2122622 URL
[2]	ONAJI I, TIWARI D, SOULATIANTORK P, et al. Digital twin in manufacturing: Conceptual framework and case studies[J]. International Journal of Computer Integrated Manufacturing, 2022, 35(8): 831-858. doi: 10.1080/0951192X.2022.2027014 URL
[3]	LÖHNER R, AIRAUDO F, ANTIL H, et al. High-fidelity digital twins: Detecting and localizing weaknesses in structures[J]. International Journal for Numerical Methods in Engineering, 2024, 125(21): e7568.
[4]	TUHAISE V V, TAH J H M, ABANDA F H. Technologies for digital twin applications in construction[J]. Automation in Construction, 2023, 152: 104931. doi: 10.1016/j.autcon.2023.104931 URL
[5]	陶飞, 张贺, 戚庆林, 等. 数字孪生模型构建理论及应用[J]. 计算机集成制造系统, 2021, 27(1): 1-15. doi: 10.13196/j.cims.2021.01.001
	TAO Fei, ZHANG He, QI Qinglin, et al. Theory of digital twin modeling and its application[J]. Computer Integrated Manufacturing Systems, 2021, 27(1): 1-15.
[6]	刘世民, 孙学民, 陆玉前, 等. 知识驱动的加工产品数字孪生拟态建模方法[J]. 机械工程学报, 2021, 57(23): 182-194. doi: 10.3901/JME.2021.23.182
	LIU Shimin, SUN Xuemin, LU Yuqian, et al. A knowledge-driven digital twin modeling method for machining products based on biomimicry[J]. Journal of Mechanical Engineering, 2021, 57(23): 182-194. doi: 10.3901/JME.2021.23.182
[7]	ZHANG H, QI Q L, TAO F. A multi-scale modeling method for digital twin shop-floor[J]. Journal of Manufacturing Systems, 2022, 62: 417-428. doi: 10.1016/j.jmsy.2021.12.011 URL
[8]	PARK K T, YANG J H, NOH S D. VREDI: Virtual representation for a digital twin application in a work-center-level asset administration shell[J]. Journal of Intelligent Manufacturing, 2021, 32(2): 501-544. doi: 10.1007/s10845-020-01586-x
[9]	CAVALIERI S, GAMBADORO S. Digital twin of a water supply system using the asset administration shell[J]. Sensors, 2024, 24(5): 1360. doi: 10.3390/s24051360 URL
[10]	BOTH M, KÄMPER B, CARTUS A, et al. Automated monitoring applications for existing buildings through natural language processing based semantic mapping of operational data and creation of digital twins[J]. Energy and Buildings, 2023, 300: 113635. doi: 10.1016/j.enbuild.2023.113635 URL
[11]	PANG Y H, HE Q, JIANG G Q, et al. Spatio-temporal fusion neural network for multi-class fault diagnosis of wind turbines based on SCADA data[J]. Renewable Energy, 2020, 161: 510-524. doi: 10.1016/j.renene.2020.06.154 URL
[12]	WANG X D, HU X F, REN Z J, et al. Knowledge-graph-based multi-domain model integration method for digital-twin workshops[J]. The International Journal of Advanced Manufacturing Technology, 2023, 128(1/2): 405-421. doi: 10.1007/s00170-023-11874-4
[13]	ZHU W H, LIU H Y, DONG Q X, et al. Multilingual machine translation with large language models:Empirical results and analysis[C]// Findings of the Association for Computational Linguistics: NAACL 2024. Mexico City, Mexico: ACL, 2024: 2765-2781.
[14]	WANG T, FAN J M, ZHENG P. An LLM-based vision and language cobot navigation approach for Human-centric Smart Manufacturing[J]. Journal of Manufacturing Systems, 2024, 75: 299-305. doi: 10.1016/j.jmsy.2024.04.020 URL
[15]	CHEN J P, LUO H W, HUANG S H, et al. Autonomous human-robot collaborative assembly method driven by the fusion of large language model and digital twin[J]. Journal of Physics: Conference Series, 2024, 2832(1): 012004. doi: 10.1088/1742-6596/2832/1/012004
[16]	ZHANG J B, ZHU J, GUO Z H, et al. More intelligent knowledge graph: A large language model-driven method for knowledge representation in geospatial digital twins[J]. International Journal of Applied Earth Observation and Geoinformation, 2025, 139: 104527. doi: 10.1016/j.jag.2025.104527 URL
[17]	ISO 55000: 2024 Asset management—Vocabulary, overview and principles[S].
[18]	VDI 2770 Blatt 1: 2020-04 Operation of process engineering plants—Minimum requirements for digital manufacturer information of the process industry—Fundamentals[S].

试验类型	平均响应时间/s			T_a/s	S_a/%
试验类型	替换	增加	删除	T_a/s	S_a/%
语义重构法	11.2	8.4	8.1	9.2	87
手动方法	93	74	67	78	96
模糊指令	20	14	11	15	81

试验类型	T_a/s	S_a/%
语义重构法	26	79
手动方法	243.7	94

试验类别	T_a/s	S_a/%
设备变更	7.3	94
模糊指令	10	91
工艺调整	19	88

试验模型	T_a/s	S_a/%
完整模型	7.3	94
变体A无知识图谱	6.1	71
变体B无多目标优化	7.1	82
变体C无LLM	25.4	62

基于大语言模型的数字孪生模型语义重构方法

Semantic Reconstruction Method of Digital Twin Model Based on Large Language Model

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 11

参考文献 18

相关文章 15

编辑推荐

Metrics

本文评价

[1]	齐麟龙, 刘立全, 王喆, 张子绅, 朱颖, 夏唐斌. 基于数字孪生的改进A^*HNSA算法在船舶分段吊运任务的应用[J]. 上海交通大学学报, 2026, 60(8): 1266-1278.
[2]	温晓健, 严哲, 胡佐治, 袁轶, 鲍劲松, 张丹. 一种生成式船舶薄板产线数字孪生建模方法[J]. 上海交通大学学报, 2026, 60(8): 1279-1289.
[3]	季昆, 吕超凡, 吕健豪, 鲍劲松, 马彦军. 一种大语言模型驱动的船舶三维小样增强式建模方法[J]. 上海交通大学学报, 2026, 60(6): 996-1007.
[4]	刘鸣飞, 凌威, 王森, 鲍劲松. 融合几何与工艺语义的风力发电机装配工艺重用方法[J]. 上海交通大学学报, 2026, 60(5): 809-823.
[5]	. 基于结构感知图注意力网络的少样本知识图谱补全[J]. J Shanghai Jiaotong Univ Sci, 2026, 31(4): 1024-1033.
[6]	. 通过显式推理建模的医学语言模型混合监督微调方法[J]. J Shanghai Jiaotong Univ Sci, 2026, 31(3): 660-670.
[7]	叶海波, 余科, 牛荣兵, 李思威. 基于大模型的战术语音指控系统作战应用研究[J]. 空天防御, 2026, 9(1): 98-107.
[8]	唐爱盼, 徐隽骁, 凌爱军, 等. 海上设施数字孪生系统测试、评估、验证与确认（TEVV）技术研究[J]. 海洋工程装备与技术, 2026, 13(1): 123-128.
[9]	张育圣, 许永辉, 周宇琦, 杜江, 魏长安. 基于大模型的协议模板和对象模型的智能关联方法[J]. 空天防御, 2025, 8(6): 94-102.
[10]	贺益雄, 代永刚, 赵兴亚, 于德清, 黄立文. 河口深槽可航宽度变化水域航行决策方法[J]. 上海交通大学学报, 2025, 59(4): 489-502.
[11]	张思霈, 徐天洋, 康传华, 慈慧鹏, 刘瑞. 数字平行战场：理论综述与发展展望[J]. 空天防御, 2025, 8(3): 29-39.
[12]	王刚, 杨科, 权文, 郭相科, 赵小茹. 基于数字孪生的防空反导数智平行战场构建方法研究[J]. 空天防御, 2025, 8(3): 1-13.
[13]	潘飞, 王骁龙, 蔡云泽, 等. 基于期刊的全球海洋工程领域研究前沿知识图谱分析(2015—2024)[J]. 海洋工程装备与技术, 2025, 12(3): 132-144.
[14]	李龙跃, 王文豪, 皮雳, 贾忠慧, 赵慧珍. 防空反导作战模拟推演分析方法综述[J]. 空天防御, 2025, 8(1): 48-53.
[15]	胡志强, 刘鸣飞, 李琦, 李心雨, 鲍劲松. 基于多源异构数据的风机多模态装配工艺知识图谱建模[J]. 上海交通大学学报, 2024, 58(8): 1249-1263.