Society faces ever greater demands for energy that is affordable, reliable, and secure. Nuclear energy is well suited to meet these goals; however, a significant gap remains between the demand versus the ability to license and demonstrate inherent safety in new designs. Modeling nuclear systems is difficult and expensive, requiring coupled analysis across several different physics domains. Further, nuclear plant phenomena occur across multiple time scales with varying orders of magnitude which makes capturing coupled behavior challenging. To solve this, a new method of coupled analysis is proposed using Multi-Physics Surrogate Models (MPSM). MPSMs offer several advantages, including code-agnostic interchangeability and the ability to run faster, allowing a broader design space to be considered.

In this paper, we demonstrate a model of MPSM using surrogate models of three production codes: Griffin (a neutronics code), BISON (a fuel performance code), and Pronghorn (a thermo-hydraulics code). These codes are coupled to solve for the power distributions of a steady state SFR (Sodium Fast Reactor). We compare the performance of running the coupled codes natively against two surrogate approaches: creating surrogates for each subdomain and a single global surrogate of the coupled codes. Finaly, these models are used to run probabilistic analysis of surrogate performance and to build 95/95 confidence intervals.