Research And Application Of Risk Assessment And Risk Reduction For Tailings Dam Failure

As an artificial risk source of debris flow with high potential energy,tailings dam failure cannot only threaten the life of residents downstream,but also endanger the surrounding falicities and equipment,and lead to environment pollution.Focused on the tailings dam failure,methods and models,such as evidence-based,machine learning,three-dimensional risk matrix,bow-tie model,cloud model,and interactive multi-criteria decision-making,are used to the research and application of the risk assessment and risk reduction for tailings dam failure.The conclusions in the paper are summarized as follows:(1)Hazard identification and evolutionary relationship analysisBased on the evidence,hazard lists containing a total of 116 kinds of hazards of tailings dam failure are constructed.Based on adversarial interpretive structure modeling method(AISM),a set of 19-level hierarchical directed topological diagrams of ’result-oriented UP type’ and’cause-oriented DOWN type’ are illustrated to characterize the evolution relationship among hazards.Based on cross-impact matrix multiplication applied to classification(MICMAC),the driving-dependence value coordinate map is drawn to realize the quantification and visualization of the evolution relationship among hazards.(2)Risk characterizationBased on the probability,intensity and exposure,a three-dimensional risk matrix model of tailings dam failure is established,to realize the quasi-quatitative assessment of four risk levels(level Ⅰ,level Ⅱ,level Ⅲ,level Ⅳ).Considering the multi-classification prediction requirement of the probability dimension,the sparrow search algrorithm(SSA)is used to optimize the support vector machine(SVM).The SSA-SVM prediction model of probability level of tailings dam failure is established,which could predict the four levels of probability.Based on the prediction accuracy,mean square error,precision(macro/micro),recall(macro/micro),F1 score(macro/micro)and other model evaluation indicators,the SSA-SVM proposed in the paper is compared and analyzed.The results show that SSA-SVM model has obvious advatages in the probability prediction of talings pond failure,and can be used as a new method for the probability prediction.Considering the regression prediction demand of tailings discharge volume within the intensity dimension,the GWO algorithm is used to optimize the support vector regression(SVR).The GWO-SVR prediction model of the tailings discharge volume is established,which realizes the prediction of tailings discharge volume.Based on the predicted value,the maximum discharge distance of tailings is further obtained.Furthermore,according to the two aforementioned parameters,the energy released due to tailings dam failure is measured,and the intensity level is divided.In terms of the exposure dimension,the four types of disaster-bearing bodies,including personnel,economy,environment and society,are taken into accout.By establishing corresponding exposure assessment model,applying comprehensive weighted factor analysis method,the exposure index and level of tailings dam failure is calculated and divided.(3)Risk reduction systemA new risk reduction system,named ’X-3-4-5’ system,is proposed in the paper.Specifically,by considering multiple hazards and ’X’ evolution paths(X means many),separating risk development process into 3 stages(influencing factors couple to hazards,hazards evolve into accidents,and accidents result in disaster),confirming 4 risk levels of tailings dam failure,introducing a 5-level of treatment measures(elimination,substitution,engineering including isolation,management including monitoring,and individual protection),the ’X-3-4-5’ risk reduction system is finally constructed in the paper.Based on 5 levels,a total of 166 measures are put forward.Based on the bow tie model,visualization of hazards and their evolutionary relationships,consequences,and risk reduction measures are achieved;Quantitative ranking of risk reduction measures based on cloud model-interactive multi-criteria decision-making method are realized.(4)Engineering applicationThe above research results are applied to Dadongbei tailings pond.The results show that the risk level of the tailings pond is Level Ⅳ,belonging to low risk.It is consistent with the actual operation status of the tailings pond.Meanwhile,a risk reduction bow-tie model is constructed,and 54 measures are provided.To sum up,a new set of risk management method for tailings dam failure is formed in the paper.It can serve as a guideline for the risk reduction of tailings dam failure,and can be used to construct a method library,hazard library,measure library,and computer system for the spatiotemporal(vertical and horizontal)dynamic risk assessment of tailings dam failure.