A Non-Intrusive Reduced-Order model for rapid analysis of thermal stratification in pressurizer surge line

Abstract

The pressurizer surge line serves to connect the pressurizer and primary circuit in a PWR (Pressurized Water Reactor) system. However, thermal stratification at its junction can induce distortion and stress, potentially damaging the pipes. Computational Fluid Dynamics (CFD) is a common numerical tool, but its time-intensive nature poses challenges for real-time assessment, especially with multiple parameter variations. To address this issue, we developed a rapid analysis method using a non-intrusive reduced-order model. The experimental design is optimized by incorporating the Generalized Subset Design to minimize sample requirements. The reduced-order model of the temperature field was derived using Proper Orthogonal Decomposition. Off-design cases were predicted using Linear, Radial Basis Function, and Radial Basis Function Neural Network interpolation techniques. The resulting temperature field was utilized for stress analysis in the pipe structure. Results indicate that linear interpolation performs best, with a maximum CvRMSE (Coefficient of Variation of the Root Mean Square Error) of 0.038 for temperature and a maximum RMSE(Root Mean Square Error) of −0.02% in predicting the maximum equivalent stress. The Radial Basis Function interpolation is slightly inferior to linear interpolation. It better fits the thermal stratification region but lacks accuracy in identifying its boundaries. This inaccuracy is more sensitive to equivalent stress, resulting in a maximum stress deviation of −0.08% for sharp boundaries. Additionally, the Radial Basis Function Neural Network is unsuitable for current study due to insufficient sample size, resulting in a maximum stress identification deviation of −3.8%. Finally, the POD coefficient is used as a independent variable to interpolate the maximum Von-Mises stress, and the relative errors were controlled within 5%. This study provides a rapid and accurate method to evaluate the temperature distributions and the maximum stresses.