Development Of Adaptive Sampling Module In RISMC Analysis Tool

In order to overcome the shortcomings of traditional deterministic and probability methods,a dynamic safety analysis method based on Monte Carlo sampling and deterministic safety analysis software was proposed.This method belongs to the forward sampling method,that is,the uncertain information of the output space is obtained by sampling the input space of uncertain parameters and combining with a lot of deterministic simulation calculation.The forward sampling method does not make use of the calculation result information of sampling,so the calculation efficiency is low and the calculation time is long.The analysis efficiency of dynamic safety analysis method has been one of the key bottlenecks restricting its engineering application.Based on the forward sampling dynamic safety analysis method and the support vector machine method,this paper studies and proposes an adaptive method which can reverse the selection of new sampling data by a few forward calculation results.Firstly,based on the research of RISMC analysis tools at home and abroad,the forward sampling methods are summarized to improve the forward sampling accuracy,and the Latin hypercube sampling method is used as the basis of adaptive sampling.Based on the developed RISMC analysis tool,a design scheme of adaptive sampling module was proposed,and an adaptive sampling module was developed independently.Taking Qinshan Nuclear Power Plant as the research object,the simulation model of power outage accident is established by using RELAP5 program.Select the height parameters of the influencing factors to conduct adaptive sampling,and analyze and test the case of adaptive sampling.The test results show that the proposed adaptive sampling method based on support vector machine can determine the safety margin position in the output space based on the input space composed of uncertain parameters,automatically select the sampling parameters near the safety margin,continuously accurate safety margin position,and finally obtain the failure probability of the output space,that is,the probabilistic safety margin of accident risk guidance.In the error range,the calculation efficiency of the adaptive sampling method can be improved effectively compared with the forward sampling method.