Surrogate Model of Severe Accident Analysis Code for SBO Aiming Probabilistic Safety Margin Analysis

摘要

A surrogate model (ROM) for severe accident analysis (SBO) is developed using the low-rank approximation with SVD in order to reduce the calculation costs of a severe accident code for safety margin analysis.ROM can rapidly reconstruct many accident progression results using 25 samples obtained by the RELAP5/SCDAPSIM. Only 9 bases out of 25 bases are used to construct the surrogate model. The boundary between the core damage and core intact regions estimated by ROM reproduces that obtained by the RELAP5/SCDAPSIM within the accuracy of approximately 300 s.In order to improve the accuracy, two issues will be addressed. The first one is the improvement on interpolation of expansion coefficients. The second one is the optimum number of samples and sampling methods to reach sufficient reproduction accuracy.