Uncertainty Analysis In Qualitative Risk Assessment Based On Evidence Theory

The rapid development of modern science and technology gives birth to nuclear power stations,spacecrafts,and other modern engineering systems.While significantly motivating the development of human society,these complex systems are frequently visited by accidents.Their safety issues are still hanging a "sword of Damocles" upon human heads.Under such circumstance,Quantitative risk assessment(QRA)method emerged in order to conduct the safety analysis and evaluation of complex systems.Uncertainty analysis is a critical part of QRA.Existing uncertainty analysis in quantitative risk assessment is based on probability theory.However,considering the specific background of risk analysis,limitations of probability theory exist in terms of uncertainty representation.Evidence theory,which was developed in the last century,has a solid mathematical foundation and strong uncertainty analysis capabilities.It can be used to satisfy the specific demands of risk analysis.In this paper,uncertainty analysis methods in quantitative risk assessment is studied based on evidence theory.The main contents are as follows:(1)In accordance with the current engineering practice,uncertainty information contained in general data is analyzed.The general data uncertainty characterization method based on evidence theory is proposed.Belief information contained in likelihood probability is analyzed under the framework of the transferable belief model.Parameter estimation method based on possibility distribution is given.A group of nuclear power plants diesel engine failure data is analyzed based on the proposed methods.The uncertainties in failure rate of the diesel engine are characterized using evidence theory.Comparing to traditional probabilistic approach,input parameter uncertainty characterized by evidence theory shows more completeness.(2)General solution for "stress-strength" model using evidence theory is analyzed.Aiming at solving the computational problem,“design point”,which is a concept in structural reliability analysis,is introduced to construct an “auxiliary area” to simplify model calculation.After analyzing the relationship between "two-norm design points" and "infinity-norm design points",analytical and numerical methods for obtaining the infinity-norm design point are proposed.A simplified calculation method based on auxiliary areas is proposed.The evacuation success probability is calculated.The proposed method,which significantly reduces the number of joint focal elements that need to be processed,can reduce the amount of computation.(3)The combination of event tree and fault tree,which is widely used in current practice,shows defectiveness in handling dependent faults.Evidence networks are proposed to model scenarios.The converting methods from fault tree and event tree to evidence network are given.Uncertainties are propagated based on constructed evidence networks.The above methods are applied in the scenario analysis of a quantitative risk assessment of a tunnel.Uncertainties are propagated through the scenario logic models.(4)There can be more than one uncertainty representation method used for QRA inputs.The relationship between fuzzy numbers and probability theory,which are commonly used in current fault tree analysis,is analyzed.Uncertainties brought by dependencies and their characterization methods are introduced.One of evidence theory's many advantages is its compatibility with probability theory and possibility theory.A hybrid uncertainty propagation method based on evidence theory is proposed.A high integrity pressure protection system is analyzed using the proposed methods.The hybrid uncertainty propagation under different representations is achieved.