改进FTA-DBN在矿山典型重大风险预测中的应用  

Enhancing the application of FTA DBN for predicting major risks in mining operations

作　　者：陆卫东 孙胤洲[3] 杨婷 张钧博 金欢 王越 王家澳 LU Weidong;SUN Yinzhou;YANG Ting;ZHANG Junbo;JIN Huan;WANG Yue;WANG Jiaao(School of Safety Science and Engineering,Xinjiang University of Engineering,Urumqi 830023,China;Xinjiang Key Laboratory of Coal Mine Disaster Intelligent Prevention and Emergency Response,Xinjiang Institute of Engineering,Urumqi 830023,China;College of Safety Science and Engineering,Liaoning Technical University,Huludao 125105,Liaoning,China;School of Electrical and Control Engineering,Liaoning Technical University,Huludao 125105,Liaoning,China;Faculty of Business Administration,Liaoning Technical University,Huludao 125105,Liaoning,China)

机构地区：[1]新疆工程学院安全科学与工程学院,乌鲁木齐830023 [2]新疆工程学院新疆煤矿灾害智能防控与应急重点实验室,乌鲁木齐830023 [3]辽宁工程技术大学安全科学与工程学院,辽宁葫芦岛125105 [4]辽宁工程技术大学电气与控制工程学院,辽宁葫芦岛125105 [5]辽宁工程技术大学工商管理学院,辽宁葫芦岛125105

出　　处：《安全与环境学报》2025年第4期1286-1297,共12页Journal of Safety and Environment

基　　金：新疆维吾尔自治区重点研发计划项目(2022B03031-1)。

摘　　要：为有效预防矿山重大事故的发生,降低矿山事故风险,以故障树分析法(Fault Tree Analysis,FTA)为理论基础,揭示矿山重大风险节点的耦合机制,然后建立通用的矿山重大风险评估动态贝叶斯网络(Dynamic Bayesian Network,DBN)。从人、机、环、管四方面筛选影响因素,构建故障树模型,涵盖45个基本事件。通过专家评价语言模糊转化及改进的相似聚合法确定DBN模型参数,以7级语言量表收集6位不同权重专家意见,基于经处理得到的各基本事件先验概率,构建DBN模型进行正向推理。将时间划分为九个时间片,在无证据输入下,发现部分节点风险概率随时间上升,矿山风险总体呈上升趋势。反向诊断假设矿山典型重大风险预测风险状态概率100%,计算节点后验概率及变异风险(Risk of Variability,ROV)值并排序,确定人员技术水平差、文化水平低等为主要因素,产品储存量过多、车辆违规操作等为关键因素。最后以某矿山为例开展分析验证工作。研究表明:所构建模型能够基于输入证据准确预测出矿山重大风险概率的变化;通过分析新疆某矿山,成功对关键风险因素进行识别,并对这些风险因素进行排序,从而识别出系统的薄弱环节,并实现风险监控,决策者因此可以迅速做出反应,减少事故风险。To effectively prevent major mine accidents and mitigate risks,this paper employs Fault Tree Analysis(FTA)to uncover the coupling mechanisms of significant risk nodes in mining operations.Subsequently,a comprehensive Dynamic Bayesian Network(DBN)model is developed for assessing major mine risks.Influencing factors are categorized into four areas:human,machine,environment,and management,to develop a fault tree model encompassing 45 basic events.The parameters of the DBN model are established through fuzzy transformation of expert evaluation language,utilizing an enhanced similarity aggregation method.Input from six experts,each assigned different weights,is gathered using a 7-level linguistic scale.After processing this information,the prior probabilities for each basic event are determined.The DBN model is then analyzed using forward reasoning,dividing the time frame into nine distinct time slices.In the absence of evidence input,it was observed that the risk probabilities of certain nodes increase over time,indicating an overall upward trend in mine risk.For reverse diagnosis,assuming a 100%probability of the predicted risk state for typical major mine risks,the posterior probabilities of the nodes are calculated along with their Risk of Variability(ROV)values,which are then ranked accordingly.Factors such as inadequate technical expertise and low educational levels among personnel are identified as primary contributors,while excessive product storage and illegal vehicle operations are recognized as key issues.Finally,a case study of a mine is performed for analysis and validation.The results indicate that the constructed model can accurately predict fluctuations in major mine risk probabilities based on the evidence provided.Through sensitivity analysis,key risk factors are effectively identified and ranked,highlighting system vulnerabilities for risk monitoring.This enables decision-makers to respond promptly,thereby reducing the likelihood of accidents.

关 键 词：安全工程 矿山工程 风险分析 风险预测 贝叶斯网络 

分 类 号：X948[环境科学与工程—安全科学]