Study Of Nuclear,Biological And Chemical Accident Modeling For Parallel Emergency Management

Nuclear,biological and chemical(NBC)accidents have posed a serious threat to the health of human beings and the development of the society.For example,the Bhopal accident(1984,India),the Chernobyl accident(1986,the former Soviet Union)and the Fukushima accident(2011,Japan)has brought huge damage to people.However,current management of NBC accident has some critical defects.For example,people cannot study the management of NBC accident from field study,so that the emergency management plan is decided according to governor’s experience and intuition,lacking scienfitic validation.Fortunately,NBC accident modeling for parallel emergency management is equipped with the ability to assist governors in emergency plan making,which can also meet the requirements of online decision,evacuation guidance and social management.The target of this study is to construct NBC accident models for parallel emergency management.After reviewing the limitations and development of current NBC accident modeling methods,this dissertation investigates the data-driven atmospheric dispersion model,accident source term estimation algorithm and large-scale pedestrian evacuation model in a novel way.NBC accident modeling and simulation methods based on Bayesian inference,no-linear system filtering,intelligent optimization algorithms and machine learning methods are proposed.Especially,some breakthroughs concerning NBC accident forward modeling and backward calculation are made in this dissertation.The contributions and deliverables are concluded as follows.(1)Proposing novel atmospheric dispersion models to improve the traditional ones.By deeply analyzing the theory and limitation of traditional dispersion method,this dissertation proposes data-driven atmospheric dispersion models to make prediction more accurate,including dynamic data-driven model based on particle filtering and static data-driven model based on artificial neural network.(2)Developing adaptive and intelligent accident source term estimation approaches.On the basis of data-driven NBC accident dispersion models,this dissertation proposes the source term estimation approaches using particle swarm optimization and classification neural network respectively to address the source estimation problem in different accident scenarios,improving the accuracy and efficiency of previous methods.In terms of the problems of varying release rate,long computational time and convergence of the algorithm,the proposed source estimation approaches are combined with Tikhonov regularization and expectation-maximization.(3)Constructing large-scale pedestrian evacuation model under NBC accident.Different from the common escaping,large-scale pedestrian evacuation may suffer from hazardous materials.Furthermore,the number of people in evacuation usually reaches million-level.Therefore,this dissertation developes a hazard evaluation model and combines social force model and cellular automata to meet the requirement of NBC accident.(4)Designing the prototype system of public safety monitoring and prediction for NBC accident.A framework and a prototype system of parallel emergency managent are constructed for tendency prediction,dynamic deduction,inverse calculation and pedestrian evacuation of NBC accident.Furthermore,using the proposed theories,algorithms and models,this dissertation designs and realizes a UAV-based platform for hazardous material monitoring and an atmospheric dispersion simulation system supporting global scale to support the prototype system.(5)Validating the feasibility of the prototype system via simulation experiment and field study.To validate whether the prototype system is work for NBC accident management,this dissertation uses the synthetic scenario made by commercial software,radioactivity data collected from Fukushima accident,gas tracing data from Indianapolis field study and real experiment conducted at a chemical industry park in Shanghai.The focus of this dissertation is the study of NBC accident modeling and simulation.On the basis of improving hazardous material atmospheric dispersion models,accident source term estimation approaches and pedestrian evacuation modeling methods are investigated.To better understand the emergency management of NBC accident,an atmospheric dispersion simulation system,a UAV-based monitoring platform are designed,which can be applied in the construction of parallel emergency management system.