Risk Analysis Of Reservoir Area Based On Bayesian Theory

With the development of the world economy and industrialization,industrial accidents and natural disasters occur frequently.In addition,the distance between the equipment in the industrial production areas is relatively close.They are closely connected and interact with each other.An initial accident can easily develop from the initial unit to the adjacent unit,which will cause a more serious accident.This phenomenon is called a domino effect accident.Although the frequency of domino accidents is low,the consequences of accidents are often more serious than a single accident.Because it involves a large number of units,the secondary accidents may be concurrent and the relationship between the accidents is complex and diverse.In particular,during the production,transportation,storage and processing of hazardous chemicals,the domino effect accident caused by the initial accident causes serious economic losses,casualties and adverse social impacts.This work mainly studies the possible safety hazards in storage of hazardous chemicals and explores the potential domino accident chain.The hazard factors,disaster models of the domino effect accident and the chain of potential accidents are analyzed.The causes of the disaster in a domino accident are complicated,the state of each node is complex,and there is a common cause failure.Based on the characteristics of domino effect accident,this work applies Bayesian network to analyze the major accidents caused by fire accidents in industrial plants.The secondary derivative disasters and the coupling relationship between various disasters are studied.This work analyzes the key factors of domino effect accidents,determines typical major accident disasters and establishes a dynamic risk assessment method for major accidents based on Bayesian networks and their secondary derived disasters according to observed evidence obtained in real time.Compared to traditional risk assessment methods(accident trees,event trees,etc.),Bayesian network has the following advantages:1)Bayesian network can not only predict the development process of the accident from the current time step to previous time step,but also identify the causes of the accident through the acquired observation data.2)The state of each node of Bayesian network is no longer limited to occurrence or non-occurrence.There may be many states of a node.3)By comparing prior with a posteriori analysis,the main causes of accident development can be determined.Based on the dynamic Bayesian theory,this work mainly carried out the following research:Based on the dynamic Bayesian network and fuzzy theory,the propagation mode of domino effect accident in the storage tank area of a chemical plant is studied,and the coupling relationship between fire and explosion accident is analyzed.The results show that fire and explosion accidents trigger each other,and the thermal radiation and explosion overpressure released by the initial accident act on adjacent units,which in turn triggers secondary accidents and higher-level accidents.Dynamic Bayesian network continuously updates the node probability based on the information of real-time feedback,and can obtain the failure probability of each node changing with time.There are many factors affecting the propagation of domino effect accidents with great uncertainties in each factor.In order to solve the uncertainty of domino effect accidents,fuzzy theory is used.The results of fuzzy theory analysis are well validated by the dynamic Bayesian network.The coupling relationship between the hazard factors and the accident consequences of thermal disasters in the cotton and linen warehouse are studied.The Bayesian network is used to analyze the causes of the fire and mildew in the cotton and linen warehouses and the coupling relationship between the disasters.The risk of thermal disasters and the configuration of safety protection measures in cotton and linen warehouses are analyzed by using typical accidents as an example to study the most common catastrophic modes in cotton and linen warehouses.The results show that the hazard factors of cotton and linen warehouse are complex.Bayesian network can clearly analyze the fire evolution of cotton and linen warehouse,quantitatively evaluate the risk of thermal disaster in cotton and linen warehouses,analyze the coupling relationship between fire and mildew and provides the basis and guidance for allocations of safety measures.