Thermal-hydraulic Reliability Evaluation of Passive System Based on Adaptive MCMC and Subset Simulation

For the commonly used subset simulation (SS) method based on modified Metropolis Hastings (MMH) algorithm, the acceptance rate of candidate samples decreases and the error of failure probability estimate increases at lower failure probability level, and the algorithm’s robustness is also poor. For the accurate, efficient and robust evaluation of thermal-hydraulic reliability of passive system using SS, a new Markov chain Monte Carlo (MCMC) method based on adaptive conditional sampling (aCS) was introduced and a reliability evaluation method based on adaptive MCMC and SS was proposed. Taking the experimental facility of a secondary side passive residual heat removal system as an example, the performance comparison of SS based on MMH (MMH-SS) and SS based on aCS (aCS-SS) was given at different failure probability levels. Calculation results show that, under lower failure rate level, aCS-SS can make the acceptance rate of candidate samples stably near the target value, and accuracy and robustness of the failure probability estimate are higher than those of MMH-SS.