| 35 | 0 | 35 |
| 下载次数 | 被引频次 | 阅读次数 |
施工安全风险的多因素耦合与专家评价的模糊性，是安全评估中的重点难题。为解决复杂施工场景下致因诊断与风险量化问题，提出一种集成云贝叶斯网络及改进D-S证据理论的双向推理评估模型，构建涵盖人-机-环-管的四维三层贝叶斯网络拓扑，引入标准云模型与群体共识算子，将语义评价转化为具有二阶不确定性的先验概率；针对多专家证据间的认知冲突，设计融合单焦元三角散度与经典冲突系数的综合冲突度量方法，并利用贝叶斯网络正逆向推理机制，实现风险预测、致因诊断与敏感性分析的协同输出。以衡永高速公路高墩施工工区为工程背景进行验证，正向推理预测该工区发生安全事故的风险概率为0.658，属于高风险等级；反向致因诊断揭示违规操作行为与安全防护佩戴缺失为最可能的致因组合，其灵敏度指标分别高达0.32与0.25。该模型为施工安全管理中的风险预警与精准管控决策提供了可量化的技术支撑。
Abstract：The multi-factor coupling of construction safety risk and the fuzziness of expert evaluation are two major problems in safety assessment. A two-way reasoning evaluation model integrating cloud Bayesian network model and improved D-S evidence theory is proposed to solve the problem of cause traceability and risk quantification in complex construction scenarios. A four-dimensional three-layer Bayesian network topology covering human-machine-environment-management is constructed. The standard cloud model and group consensus operator are introduced to transform the specific semantic evaluation into a priori probability with second-order uncertainty. Aiming at the cognitive conflict between multi-expert evidence， a comprehensive conflict measurement method combining single-focus element triangular divergence and classical conflict coefficient is designed to achieve effective reconciliation of high-conflict evidence. By using the forward and backward reasoning mechanism of Bayesian network， the collaborative output of risk prediction， cause diagnosis and sensitivity analysis is realized. Taking the high pier construction area of Hengyong Expressway as the engineering background for verification， the forward reasoning predicts that the risk probability of safety accidents in the work area is 0.658， which belongs to the high risk level ; the reverse cause diagnosis revealed that the violation operation behavior and the lack of safety protection wearing were the most likely cause combinations， and the sensitivity indexes were as high as 0.32 and 0.25， respectively. The model provides quantifiable technical support for risk early warning and precise control decision-making in construction safety management.
[1]董磊， 吴福居， 史健勇， 等. 基于大语言模型的施工安全多模态知识图谱的构建与应用[J]. 计算机工程与应用， 2025， 61(9): 325-333.
[2]张鑫，吴珂琪，肖晖，等.基于改进YOLOv7模型的公路建设施工现场安全帽佩戴检测系统[J/OL].公路工程，2026:1-10(2026-03-12)[2026-04-25].https://link.cnki.net/urlid/43.1481.U.20260312.1114.002.
[3]苏伟， 王祺顺， 杨瑞松， 等. 基于改进MobileNet-V3算法的高速公路施工现场风险行为实时检测研究[J]. 公路工程， 2025， 50(3): 250-257.
[4]粟双全， 杨飞， 杨瑞松， 等. 高速公路施工现场AI视频监控系统技术及应用[J]. 公路工程， 2025， 50(4): 223-231.
[5]李宏， 马金魁， 田坤云， 等. 瓦斯抽采钻孔封孔作业风险辨识与评估[J]. 中国安全科学学报， 2023， 33(12): 85-91.
[6]FANG Y， ZHANG Y Y， TU H J， et al. Application of dunhuang caisson patterns in cultural and creative product design based on artificial intelligence generated content technology[J]. Engineering Applications of Artificial Intelligence， 2026， 177(Pt 1): 114816.
[7]YANG R S， XU P， WU Y M. Evolution of freezing in drainage ditches of high-speed railway tunnels in cold regions[J]. Cold Regions Science and Technology， 2026， 243: 104778.
[8]REN J， HAN G L， GAO X， et al. Water quality assessment in a megacity river: water chemical analysis integrated with AHP-Entropy weighted fuzzy comprehensive evaluation model[J]. Journal of Environmental Sciences， 2026， 162: 121-130.
[9]WANG B Z. Optimizing fault tree traversal order for efficient minimal cut set generation: a lightweight graph neural network approach[J]. Reliability Engineering & System Safety， 2026， 272(Pt 3): 112648.
[10]朱庆， 蒋蕙琪， 吴浩宇， 等. 大语言模型驱动的钻爆法隧道塌方风险研判智能体构建方法[J/OL].西南交通大学学报，2026:1-12(2026-04-16)[2026-04-25].https://link.cnki.net/urlid/51.1277.U.20260416. 1026.002.
[11]SU Y K， GAO X H， ZHENG Z Z. Unsupervised deep learning and Bayesian network–based identification of key risk factors in construction quality defects[J]. Journal of Construction Engineering and Management， 2026， 152(7): 04026083.
[12]许鹏， 杨瑞松， 胡清海， 等. 基于水体冻结机制的寒区隧道排水沟冻害预测[J]. 冰川冻土， 2025， 47(4): 1004-1014.
[13]MA Q， DING L， JI J， et al. Performance degradation assessment of domino effect safety barriers in multi-hazard environments based on a two-stage Markovian model using dynamic Bayesian network[J]. Process Safety and Environmental Protection， 2026， 212: 108883.
[14]PERIN A， HOSSEINI S A， ZIO E， et al. Enhanced Bayesian network framework with imprecise probabilities for modelling degradation in concrete structures of radioactive waste disposals accounting for uncertainties due to climate change[J]. Structural Safety， 2026， 121: 102718.
[15]MA C， SHAH A M， CHEN W， et al. Comparing the sleep benefits of nature-based Interventions: evidence from a Bayesian network meta-analysis[J]. iScience， 2026， 29(4): 115513.
[16]GRAMIFAR M， KABIR G， KHAN S A. Credit risk assessment using Bayesian networks and machine learning approaches[J]. Advanced Engineering Informatics， 2026， 74(Pt A): 104638.
[17]许鹏， 杨瑞松， 刘平飞， 等. 寒区隧道衬背脱空积水冻胀对衬砌结构的影响及预测方法[J]. 长安大学学报(自然科学版)， 2025， 45(4): 107-117.
[18]FAN M， YANG Z Q， LI J J， et al. Study on the Bayesian network probability model for soil seismic liquefaction with incomplete data[J]. Bulletin of Earthquake Engineering， 2026， 24(6): 3701-3729.
[19]DENG Y S， YANG S R， CUI T， et al. An integrated Dempster-Shafer approach for evaluating development potential of urban underground space[J]. Results in Engineering， 2026， 29: 109613.
[20]天津市建工工程总承包有限公司. 建筑施工安全检查标准: JGJ 59-2011[S]. 北京: 中国建筑工业出版社， 2011.
[21]孙妮妮， 陈泽华， 牛昱光， 等. 基于云模型重叠度的相似性度量[J]. 计算机应用， 2015， 35(7): 1955-1958， 1964.
[22]旷文坚，彭祖武.基于云模型的微型桩加固边坡施工安全风险评估研究[J].湖南交通科技，2025，51(3):119-125.
基本信息:
中图分类号:TU714
引用信息:
[1]陈海威,曾亚林,杨飞,等.基于云贝叶斯网络的建筑施工安全风险双向推理评估模型研究[J].公路工程().
基金信息:
湖南省交通科技创新计划项目（202501-2，202220）
2026-07-02
2026-07-02
2026-07-02