• CN:11-2187/TH
  • ISSN:0577-6686

机械工程学报 ›› 2025, Vol. 61 ›› Issue (12): 315-326.doi: 10.3901/JME.2025.12.315

• 交叉与前沿 • 上一篇    

扫码分享

基于数字孪生的生产安全风险实时评估方法

袁堂晓1, 吴亲亲1, 柳林燕1, 汪惠芬1, 田双2   

  1. 1. 南京理工大学机械工程学院 南京 210094;
    2. 重庆长安望江工业集团有限公司装备及工艺研究所 重庆 401133
  • 收稿日期:2024-07-22 修回日期:2024-12-16 发布日期:2025-08-07
  • 作者简介:袁堂晓,男,1993年出生,博士研究生。主要研究方向为决策场景的风险评估。E-mail:yuantx@njust.edu.cn;柳林燕(通信作者),女,1985年出生,博士,副教授,硕士研究生导师。主要研究方向为数字化设计制造、智能制造、数字孪生、知识库与知识管理。E-mail:liulinyan@njust.edu.cn
  • 基金资助:
    国家自然科学基金资助项目(51705256)。

Real-time Evaluation Method for Production Safety Risks Based on Digital Twin

YUAN Tangxiao1, WU Qinqin1, LIU Linyan1, WANG Huifen1, TIAN Shuang2   

  1. 1. School of Mechanical Engineering, Nanjing University of Science and Technology, Nanjing 210094;
    2. Institute of Equipment and Process Research, Chongqing Chang'an Wangjiang Industrial Group Co., Ltd., Chongqing 401133
  • Received:2024-07-22 Revised:2024-12-16 Published:2025-08-07

摘要: 在危险产品自动化生产时,设备失能等异常情况将导致风险事故。因此,为有效预防人员伤亡和减少财产损失,基于现场状况监测的风险实时评估迫在眉睫。首先,从时空域交叉的角度表征了车间实时风险。其次在实时风险表征基础上,构建一个综合物理、映射和预仿真的数字孪生三重空间实时风险评估框架;采用Agent建模方法,对风险要素-状态环境、灾害事件和承灾者分别进行建模;详细设计多源数据与事故机理在这三重空间内的交互逻辑。最后,将所提方法应用在某火工品生产单元中,验证所提出的方法在实际应用场景下的有效性。该方法集成生产行为、设备状态、环境等多个维度的信息,能够实时评估风险,突破传统方法的局限,为自主低风险火工品生产探索了新的技术路径,为黑灯工厂的实现提供技术支持。

关键词: 实时风险评估, 数字孪生, Agent模型, 生产安全

Abstract: In the automated production of dangerous products, equipment failures and other exceptional situations can lead to risk accidents. Therefore, effective real-time risk assessment based on on-site condition monitoring is crucial to prevent personnel injuries and reduce property losses. Firstly, the workshop's real-time risks are characterized from the perspective of spatiotemporal cross-analysis. Secondly, based on this real-time risk characterization, a comprehensive framework for real-time risk assessment in a digital twin environment that integrates physical modeling, mapping, and pre-simulation is constructed. Utilizing an agent-based modeling approach, the risk factors-state environment, disaster events, and vulnerable entities-are modeled separately. The interaction logic between multi-source data and accident mechanisms within these three spatial dimensions is designed in detail. Finally, the proposed method is applied in a specific pyrotechnic product production unit, demonstrating its effectiveness in practical application scenarios. This method integrates information across multiple dimensions, including production behavior, equipment status, and environmental conditions, enabling real-time risk assessment. It overcomes the limitations of traditional methods and explores new technological pathways for autonomous low-risk production of pyrotechnic products, providing technical support for the realization of dark factory environments.

Key words: real-time risk assessment, digital twin, agent-based model, production safety

中图分类号: