基于贝叶斯网络的尾矿库溃坝风险动态评价  

Dynamic risk evaluation of tailings dam failure based on Bayesian network

作　　者：崔旭阳 叶义成[1,2] 胡南燕 周栋 谭文侃 黄兆云 Cui Xuyang;Ye Yicheng;Hu Nanyan;Zhou Dong;Tan Wenkan;Huang Zhaoyun(College of Resources and Environmental Engineering,Wuhan University of Science and Technology,Wuhan Hubei 430081,China;Industrial Safety Engineering Technology Research Center of Hubei Province,Wuhan Hubei 430081,China;Hubei Jingshen Safety Technology Co.,Ltd.,Yichang Hubei 443000,China)

机构地区：[1]武汉科技大学资源与环境工程学院,湖北武汉430081 [2]湖北省工业安全工程技术研究中心,湖北武汉430081 [3]湖北景深安全技术有限公司,湖北宜昌443000

出　　处：《化工矿物与加工》2022年第10期25-30,共6页Industrial Minerals & Processing

基　　金：湖北省重点研发计划(2020BCA082);湖北省中央引导地方科技发展专项(2019ZYYD060)。

摘　　要：尾矿库溃坝事故致因要素复杂,各风险因素耦合并随时间演化,传统的静态评价忽略了被评价系统的动态变化问题。基于此,首先考虑人员、环境、管理及尾矿库自身风险要素,对尾矿库系统进行风险辨识,构建尾矿库溃坝风险指标体系;然后基于Kendall和谐系数确定指标重要度排序,使用数据包络法和熵值法计算各指标的权重,嵌入尾矿库运行期各个时间段的时间权重对原始贝叶斯网络作加权改进,将模型横向扩展成具有多个时间片的动态贝叶斯网络,通过区间数排序改进条件概率的推理计算,利用改进的动态加权贝叶斯网络模型评估尾矿库溃坝的风险趋势;最后将改进后的模型与传统模型进行对比分析,结果表明,改进后的评估模型考虑了尾矿库溃坝的动态发展过程,评估结果更符合工程实际。Because of the complex factors causing failure of the tailings dam,which are inter-coupling and changing with time,the dynamic change of the system to be evaluated was ignored during the traditional static evaluation.First of all,such risk factors as personnel,environment,management and the tailings pond were considered during risk identification on the tailings handling system thus the risk index system for the failure of tailings dam was able to be constructed.Afterwards,the factor importance ranking was determined based on the Kendall's coefficient of concor-dance,the weight of each factor can be calculated using the DEA and the entropy method,the weighting improvement on the primitive Bayesian network was made by inserting the time weight of each stage during operation of the tailings pond,the model was horizontally extended into a dynamic Bayesian network with many time slots,the reasoning calculation of conditional probability was improved by sorting interval numbers,the risk tendency of the tailings pond failure was evaluated by use of the improved dynamic weighted Bayesian network model.Finally,comparative analysis on the improved model and the traditional model was carried out.The results from the improved evaluation model reflected the dynamic failure process of the tailings dam and preferably conformed to the engineering facts.

关 键 词：尾矿库 溃坝 贝叶斯网络 风险评价 时间权重 区间数排序 Kendall和谐系数 

分 类 号：TV649[水利工程—水利水电工程]