Research On Reliability Analysis Method Of Digital Instrument & Control System For Nuclear Power Plant Based On Monte Carlo Method

As the nerve center of generation nuclear power plant(NPP),the Distributed Control System(DCS)plays the most important role in ensuring the safety,reliability and economical operation of NPPs.Compared with the analog control system,the structure of the DCS is more complex and there are a lot of interactions between hardware,software and operators.It will be difficult to achieve a satisfactory result if still using the traditional event-tree/fault-tree(ET/FT)approach to analyse its reliability.Therefore,new reliability modeling methods are required to conduct the reliability evaluation of DCS.In all reliability evaluation methods,Monte Carlo method has attracted more attention for it can effectively solve the Curse of Dimensionality.However,for DCS,system failure is usually a low probability event,and to obtain the higher accuracy,a large scale sampling must be carried out,which makes the lower computational efficiency.Therefore,it is significant to study the methods to improve the computational efficiency of Monte Carlo.In this paper,the digital reactor protection system is taken as research object.First,its structure and working principle are analyzed,and according to the simplified hypothesis of the actual system,a reliability block diagrams(RBD)model of the digital reactor protection system is established.Moreover,taking the signal trips the failure in main water or main steam system as an example,the reliability of the system is verified.Based on the work above,the theory of Monte Carlo reliability modeling and the reduction of sampling variance are studied.Using the hill climbing algorithm to solve the optimal value of k in the unified multiplier method,and then,the Lagrange multiplier method is introduced to construct the optimal sampling density function in the importance sampling method.At the same time,the analysis process is divided into two parts: pre sampling and formal sampling,an adaptive sampling algorithm is proposed.Moreover,advantages and disadvantages of the two improved algorithm are compared and analyzed.Finally,to improve the convergence order of Monte Carlo mean error,the construction of low discrepancy sequence and its randomization methods are researched,and the randomized low discrepancy sequence is applied to the adaptive sampling method,a random quasi Monte Carlo method based on the adaptive importance sampling is proposed.Moreover,the correctness and effectiveness of this improved algorithm are proved through simulations.