工业大模型驱动的航空智能制造AI孪生控制方法体系

doi:10.3901/JME.260228

机械工程学报 ›› 2026, Vol. 62 ›› Issue (5): 61-73.doi: 10.3901/JME.260228

• 特邀专栏：信息驱动的总装拉动生产模式、技术及应用 • 上一篇下一篇

扫码分享

工业大模型驱动的航空智能制造AI孪生控制方法体系

许鸿伟^1,2, 刘丽兰^1,2, 张洁³, 秦威⁴, 邢宏文⁵, 汪玮⁵, 刘思仁⁵, 吕佑龙³

1. 上海大学机电工程与自动化学院上海 200444;
2. 上海市智能制造及机器人重点实验室上海 200444;
3. 东华大学人工智能研究院上海 201620;
4. 上海交通大学机械与动力工程学院上海 200240;
5. 上海飞机制造有限公司上海 201324

收稿日期:2025-05-16 修回日期:2025-11-04 出版日期:2026-03-05 发布日期:2026-04-23
作者简介:许鸿伟,男,1995年出生,博士,博士后,助理研究员。主要研究方向为复杂系统制造过程质量控制、工业大数据分析与智能人机协作。E-mail:hongwei_xu@shu.edu.cn
刘丽兰,女,1975年出生,博士,教授,博士研究生导师。主要研究方向为数字孪生、航空航天装备质量控制。E-mail:lancy@shu.edu.cn
邢宏文(通信作者),男,1984年出生,研究员。主要研究方向为航空智能制造、先进装配技术。E-mail:xinghongwen@comac.cc
基金资助:
国家自然科学基金青年科学基金(52505565)、上海市科学技术委员会优秀技术带头人计划(23XD1431500)、中国博士后科学基金(2025M770789)和国家资助博士后研究人员计划(GZB20250346)资助项目。

AI Twin Control Method System for Aeronautical Intelligent Manufacturing Driven by Industrial Large Models

XU Hongwei^1,2, LIU Lilan^1,2, ZHANG Jie³, QIN Wei⁴, XING Hongwen⁵, WANG Wei⁵, LIU Siren⁵, Lü Youlong³

1. School of Mechatronic Engineering and Automation, Shanghai University, Shanghai 200444;
2. Shanghai Key Laboratory of Intelligent Manufacturing and Robotics, Shanghai 200444;
3. Institute of Artificial Intelligence, Donghua University, Shanghai 201620;
4. School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai 200240;
5. Shanghai Aircraft Manufacturing Co., Ltd., Shanghai 201324

Received:2025-05-16 Revised:2025-11-04 Online:2026-03-05 Published:2026-04-23

摘要/Abstract

摘要： 针对航空智能制造中数据-知识协同效率低、装配偏差溯源困难、工艺参数优化滞后及虚实交互验证不足等关键问题，本研究提出一种以工业大模型为认知引擎的AI孪生控制方法体系，构建覆盖“感知-诊断-决策-验证”的数字孪生闭环控制框架。通过知识图谱网络化关联建模，实现多源异构数据的语义融合与动态推理，打造航空制造工业大模型语料库，形成具备自主演化能力的专业知识底座；开发面向多场景智能决策闭环的工业大模型算法库，利用贝叶斯因果推断解析装配偏差的多层级耦合诱因，结合增量集成学习实现多源耦合偏差的动态演化预测，基于迁移强化学习突破跨场景参数优化瓶颈；最终通过数字孪生技术构建虚实双向驱动的验证闭环。以某型号民用客机机身壁板装配为验证对象的结果表明，所提方法能够显著提升长桁自动装配精度，长桁一次装调成功率较传统方法提升24%；实现钻铆质量实时检测，连续钻铆缺陷识别准确率达98%。该研究通过构建与演化领域知识底座，深度驱动了从偏差因果溯源到孪生验证的全流程闭环，实现了制造决策从经验驱动到模型认知驱动的范式跃迁。

关键词: 工业大模型, 数字孪生, 偏差累积传递, 贝叶斯因果推断, 工艺参数优化

Abstract: To address key challenges in aviation intelligent manufacturing, such as low data-knowledge collaboration efficiency, difficulties in tracing assembly deviation sources, lagging process parameter optimization, and insufficient virtual-real interactive verification, this study proposes an AI twin control methodology framework with industrial large models as the cognitive engine, and constructs a digital twin closed-loop control framework covering the entire "perception-diagnosis-decision-verification" process. Through networked associative modeling of knowledge graphs, semantic fusion and dynamic reasoning of multi-source heterogeneous data are realized, an industrial large model corpus for aviation manufacturing is established, and a professional knowledge base with autonomous evolution capabilities is formed. An industrial large model algorithm library for multi-scenario intelligent decision closed-loops is developed: Bayesian causal inference is used to analyze multi-level coupled causes of assembly deviations; incremental ensemble learning is integrated to achieve dynamic evolution prediction of multi-source coupled deviations; and transfer reinforcement learning is applied to break through the bottleneck of cross-scenario parameter optimization. Finally, a virtual-real bidirectional driven verification closed-loop is built using digital twin technology. Verification results based on the fuselage panel assembly of a certain type of civil aircraft show that the proposed method significantly improves the automatic assembly accuracy of stringers, with the one-time assembly and adjustment success rate of stringers increased by 24% compared with traditional methods. It also enables real-time inspection of drilling and riveting quality, achieving an accuracy rate of 98% in identifying continuous drilling and riveting defects. By constructing and evolving a domain-specific knowledge base, this study deeply drives the full-process closed-loop from deviation causal tracing to twin verification, and realizes a paradigm shift in manufacturing decision-making from experience-driven to model cognition-driven.

Key words: industrial large models, digital twin, deviation accumulation and transmission, Bayesian causal inference, process parameter optimization

中图分类号:

TH166

许鸿伟, 刘丽兰, 张洁, 秦威, 邢宏文, 汪玮, 刘思仁, 吕佑龙. 工业大模型驱动的航空智能制造AI孪生控制方法体系[J]. 机械工程学报, 2026, 62(5): 61-73.

XU Hongwei, LIU Lilan, ZHANG Jie, QIN Wei, XING Hongwen, WANG Wei, LIU Siren, Lü Youlong. AI Twin Control Method System for Aeronautical Intelligent Manufacturing Driven by Industrial Large Models[J]. Journal of Mechanical Engineering, 2026, 62(5): 61-73.

导出引用管理器 EndNote|Reference Manager|ProCite|BibTeX|RefWorks

链接本文: http://www.cjmenet.com.cn/CN/10.3901/JME.260228

http://www.cjmenet.com.cn/CN/Y2026/V62/I5/61

参考文献

[1] 张永亮，张辉，李智，等. 飞机柔性装配技术研究现状与发展趋势[J]. 制造技术与机床，2025(6):73-84. ZHANG Yongliang，ZHANG Hui，LI Zhi，et al. Research status and development trend of aircraft flexible assembly technology[J]. Manufacturing Technology & Machine Tool，2025(6):73-84.
[2] ZHU J Y，QIN W，HU J H，et al. Influential process nodes identification strategy for aircraft assembly system based on complex network and improved PageRank[J]. Adv Eng Inform，2023，58:1-8.
[3] 常笑，贾晓亮. 知识驱动的飞机大修数字孪生车间主动管控方法[J]. 计算机集成制造系统，2021，27(2):390-402. CHANG Xiao，JIA Xiaoliang. Knowledge-driven active management and control method for digital twin workshop of aircraft overhaul[J]. Computer Integrated Manufacturing Systems，2021，27(2):390-402.
[4] 陶飞，张辰源，刘蔚然，等. 数字工程及十个领域应用展望[J]. 机械工程学报，2023，59(13):193-215. TAO Fei，ZHANG Chenyuan，LIU Weiran，et al. Digital engineering and its application prospects in ten fields[J]. Journal of Mechanical Engineering，2023，59(13):193-215.
[5] 宋学官，来孝楠，何西旺，等. 重大装备形性一体化数字孪生关键技术[J]. 机械工程学报，2022，58(10):298-325. SONG Xueguan，LAI Xiaonan，HE Xiwang，et al. Key technologies of shape-performance integrated digital twin for major equipment[J]. Journal of Mechanical Engineering，2022，58(10):298-325.
[6] 施佳宏，刘晓军，刘庭煜，等. 面向生产线仿真的数字孪生逻辑模型构建方法[J]. 计算机集成制造系统，2022，28(2):442-54. SHI Jiahong，LIU Xiaojun，LIU Tingyu，et al. Construction method of digital twin logic model for production line simulation[J]. Computer Integrated Manufacturing Systems，2022，28(2):442-54.
[7] 张洁，高亮，秦威，等. 大数据驱动的智能车间运行分析与决策方法体系[J]. 计算机集成制造系统，2016，22(5):1220-1228. ZHANG Jie，GAO Liang，QIN Wei，et al. Big data-driven operation analysis and decision-making method system for smart workshop[J]. Computer Integrated Manufacturing Systems，2016，22(5):1220-1228.
[8] SUN Y N，QIN W，ZHUANG Z L，et al. An adaptive fault detection and root-cause analysis scheme for complex industrial processes using moving window KPCA and information geometric causal inference[J]. J Intell Manuf，2021，32(7):2007-2021.
[9] HU J H，ZHU J Y，SUN Y N，et al. Network-based two-stage robust scheduling strategy for the aircraft assembly system with uncertain duration[J]. Comput Ind Eng，2024，188:1-15.
[10] 陶飞，刘蔚然，刘检华，等. 数字孪生及其应用探索[J]. 计算机集成制造系统，2018，24(1):1-18. TAO Fei，LIU Weiran，LIU Jianhua，et al. Digital twin and its application exploration[J]. Computer Integrated Manufacturing Systems，2018，24(1):1-18.
[11] 陶飞，张萌，程江峰，等. 数字孪生车间——一种未来车间运行新模式[J]. 计算机集成制造系统，2017，23(1):1-9. TAO Fei，ZHANG Meng，CHENG Jiangfeng，et al. Digital twin workshop—A new operation mode for future workshops[J]. Computer Integrated Manufacturing Systems，2017，23(1):1-9.
[12] 陶飞，刘蔚然，张萌，等. 数字孪生五维模型及十大领域应用[J]. 计算机集成制造系统，2019，25(01):1-18. TAO Fei，LIU Weiran，ZHANG Meng，et al. Five-dimensional model of digital twin and its application in ten fields[J]. Computer Integrated Manufacturing Systems，2019，25(1):1-18.
[13] LI W，CHEN L L，CHEN X H，et al. A data-and- knowledge-driven framework for automated generation of asphalt pavement maintenance strategies[J]. Int J Pavement Eng，2025，26(1):1-14.
[14] EGHBALI P，SATIR O B，BECCE F，et al. Causal associations between scapular morphology and shoulder condition estimated with Bayesian statistics[J]. Comput Meth Prog Bio，2025，263:1-11.
[15] XU C，HAYASHI N，INUIGUCHI M. Enhancing reinforcement learning controllers with GAN-generated data and transfer learning[J]. Sice J Control Meas，2025，18(1):1-15.
[16] CHEN J P，HUANG S H，WANG X W，et al. Perception-decision-execution coordination mechanism driven dynamic autonomous collaboration method for human-like collaborative robot based on multimodal large language model[J]. Robot Cim-Int Manuf，2026，98:1-15.
[17] 郭飞燕，刘检华，邹方，等. 数字孪生驱动的装配工艺设计现状及关键实现技术研究[J]. 机械工程学报，2019，55(17):110-32. GUO Feiyan，LIU Jianhua，ZOU Fang，et al. Research on status and key implementation technologies of digital twin-driven assembly process design[J]. Journal of Mechanical Engineering，2019，55(17):110-32.
[18] 黄彬彬，张映锋，黄博，等. 数字孪生驱动的复杂产品智能运维服务体系与核心技术[J]. 机械工程学报，2022，58(12):250-60. HUANG Binbin，ZHANG Yingfeng，HUANG Bo，et al. Digital twin-driven intelligent operation and maintenance service system and core technologies for complex products[J]. Journal of Mechanical Engineering，2022，58(12):250-260.
[19] 任磊，贾子翟，赖李媛君，等. 数据驱动的工业智能:现状与展望[J]. 计算机集成制造系统，2022，28(7):1913-1939. REN Lei，JIA Zizhai，LAI Liyuanjun，et al. Data-driven industrial intelligence:Status and prospects[J]. Computer Integrated Manufacturing Systems，2022，28(7):1913-39.
[20] 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. TAO Fei，ZHANG He，QI Qinglin，et al. Theory and application of digital twin model construction[J]. Computer Integrated Manufacturing Systems，2021，27(1):1-15.
[21] 闫纪红，姬思阳. 大数据驱动的车间数字孪生模型构建方法[J]. 机械工程学报，2023，59(12):62-77. YAN Jihong，JI Siyang. Big data-driven construction method of workshop digital twin model[J]. Journal of Mechanical Engineering，2023，59(12):62-77.
[22] 王昊琪，李旭鹏，李浩，等. 人机环境动静融合的数字孪生车间快速建模方法[J]. 航空制造技术，2024，67(11):14-22. WANG Haoqi，LI Xupeng，LI Hao，et al. Rapid modeling method of digital twin workshop with dynamic and static integration of human-machine-environment[J]. Aeronautical Manufacturing Technology，2024，67(11):14-22.
[23] JI X F，GONG F M，WANG N L，et al. Pixel-level semantic parsing in complex industrial scenarios using large vision-language models[J]. Inform Fusion，2025，116:1-14.
[24] JOSE S，NGUYEN K T P，MEDJAHER K，et al. Advancing multimodal diagnostics:Integrating industrial textual data and domain knowledge with large language models[J]. Expert Syst Appl，2024，255:1-13.
[25] LIAN Z，ZHOU Z J，HU C H，et al. Interpretable large-scale belief rule base for complex industrial systems modeling with expert knowledge and limited data[J]. Adv Eng Inform，2024，62:1-15.
[26] 李浩，陶飞，王昊琪，等. 基于数字孪生的复杂产品设计制造一体化开发框架与关键技术[J]. 计算机集成制造系统，2019，25(6):1320-36. LI Hao，TAO Fei，WANG Haoqi，et al. Digital twin-based integrated development framework and key technologies for complex product design and manufacturing[J]. Computer Integrated Manufacturing Systems，2019，25(6):1320-36.
[27] 庄存波，刘检华，熊辉，等. 产品数字孪生体的内涵、体系结构及其发展趋势[J]. 计算机集成制造系统，2017，23(4):753-68. ZHUANG Cunbo，LIU Jianhua，XIONG Hui，et al. Connotation，architecture and development trend of product digital twin[J]. Computer Integrated Manufacturing Systems，2017，23(4):753-68.
[28] XU Q H，GAO P J，WANG J L，et al. AKGNN-PC:An assembly knowledge graph neural network model with predictive value calibration module for refrigeration compressor performance prediction with assembly error propagation and data imbalance scenarios[J]. Adv Eng Inform，2024，60:1-11.
[29] ZHOU B，LI X Y，LIU T Y，et al. CausalKGPT:Industrial structure causal knowledge-enhanced large language model for cause analysis of quality problems in aerospace product manufacturing[J]. Adv Eng Inform，2024，59:1-16.
[30] NODA A，ISOZAKI T. Practically effective adjustment variable selection in causal inference[J]. J Phys- Complexity，2025，6(1):1-13.
[31] XU H W，QIN W，SUN Y N，et al. Attention mechanism-based deep learning for heat load prediction in blast furnace ironmaking process[J]. J Intell Manuf，2024，35(3):1207-20.
[32] XU H W，ZHANG Q H，SUN Y N，et al. A fast ramp-up framework for wafer yield improvement in semiconductor manufacturing systems[J]. J Manuf Syst，2024，76:222-33.
[33] ZHANG Q，XU C，LI J，et al. LLM-TSFD:An industrial time series human-in-the-loop fault diagnosis method based on a large language model[J]. Expert Syst Appl，2025，264:1-23.

工业大模型驱动的航空智能制造AI孪生控制方法体系

AI Twin Control Method System for Aeronautical Intelligent Manufacturing Driven by Industrial Large Models

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

本文评价

[1]	杨毅晟, 李明, 杨泽源, 闫喜强, 严思杰, 丁汉. 数字技术赋能风洞全生命周期体系[J]. 机械工程学报, 2026, 62(7): 97-113.
[2]	王金龙, 宣泳辉, 冀秀坤, 鲍永杰, 石泽宇. 基于单向降阶高保真代理模型的CFRP机翼力学状态数字孪生研究[J]. 机械工程学报, 2026, 62(7): 126-138.
[3]	罗忠, 李洪雨, 李雷, 董胤哲. 数字孪生驱动的转子系统变工况难测点响应实时预测方法[J]. 机械工程学报, 2026, 62(7): 221-233.
[4]	杨泽浩, 董威, 黄思翰, 阴艳超, 董李扬, 郑祖杰. 基于数字孪生和强化学习的智能生产线实时调度仿真优化方法[J]. 机械工程学报, 2026, 62(5): 12-25.
[5]	吴锦辉, 陈永昶, 吴亚男, 郝会乾, 袁翔, 陶友瑞. 数字孪生驱动的反射面天线运行状态监测与虚拟调试方法[J]. 机械工程学报, 2026, 62(5): 263-273.
[6]	杨亮亮, 龚壮壮, 何西旺, 王沐晨, 闵强, 阚子云, 宋学官. 面向结构损伤识别的数字孪生建模[J]. 机械工程学报, 2026, 62(4): 328-341.
[7]	刘素艳, 董一林, 乔一鸣, 马增强. 基于改进蜣螂优化算法的轴承数字孪生模型的动态故障注入研究[J]. 机械工程学报, 2026, 62(3): 446-457.
[8]	王世博, 葛世荣, 管增伦, 周士林, 王丽杰, 李雪峰, 王赟, 袁晓明, 马广军. 面向数字孪生的采煤机滚筒实时截割载荷模型研究[J]. 机械工程学报, 2026, 62(3): 479-491.
[9]	刘海波, 邓平, 迟庆宇, 刘天然, 刘阔, 李特, 黄祖广, 刘行健, 薄其乐, STEVEN Y LIANG, 王永青. 智能加工中数据使能技术与应用[J]. 机械工程学报, 2026, 62(2): 407-444.
[10]	朱海华, 付泰然, 李霏, 刘长春, 蔡祺祥, 唐敦兵. 基于数字孪生的复杂产品装配物料过程齐套时间预测[J]. 机械工程学报, 2025, 61(8): 384-398.
[11]	李睿智, 陈悦敏, 闫纪红. 面向轨迹动态感知与自主决策的工业机器人数字孪生建模方法研究[J]. 机械工程学报, 2025, 61(7): 269-283.
[12]	胡炳涛, 钟锐锐, 冯毅雄, 杨晨, 王天跃, 洪兆溪, 谭建荣. 人-信息-物理互联环境下数字车间制造能力建模与自适应调度[J]. 机械工程学报, 2025, 61(3): 23-39.
[13]	马帅, 冷杰武, 陈祝云, 李巍华, 李波, 刘强. 基于数字孪生和深度迁移学习的电主轴热误差建模方法[J]. 机械工程学报, 2025, 61(3): 52-66.
[14]	王玉静, 李祎然, 康守强, 刘连胜, 李玉庆, 孙宇林. 基于数字孪生的不同工况下谐波减速器故障诊断方法[J]. 机械工程学报, 2025, 61(18): 12-26.
[15]	王柏村, 宋词, 苑艺修, 周慧颖, 鲍劲松, 黄思翰, 刘蔚然, 刘庭煜, 阮兵, 陶飞, 谢海波, 杨华勇. 面向人本智造的人体运动数字孪生研究与应用进展[J]. 机械工程学报, 2025, 61(15): 21-39.