Generation and Use of Reduced Order Models for Safety Applications Using RAVEN

INTRODUCTION The Risk-Informed Safety Margin Characterization (RISMC) [1] Pathway (as part of the Light Water Sustainability (LWRS) Program [2]) aims to develop simulation-based tools and methods to assess risks for existing Nuclear Power Plants (NPPs). This Pathway, by developing new methods, is extending the Probabilistic Risk assessment (PRA) stateof-the-practice methods [3] which have been traditionally based on logic structures such as Event-Trees (ETs) and Fault-Trees (FTs) [4]. In more detail, the RISMC approach uses stochastic frameworks (i.e., RAVEN [5]) coupled with deterministic codes that model specific physical aspects of the plant (e.g., thermo-hydraulic and thermo-mechanic using RELAP5-3D [6] or RELAP-7 [7], and GRIZZLY [8] respectively). One research direction is use of surrogate models, also known as Reduced Order Models (ROMs), as possible substitutes for one or more of the needed physical aspects. The use of ROMs can greatly reduce the computational cost of a single multi-physics simulation run. This advantage is relevant when many simulation runs need to be performed according to the desired stochastic analysis (usually through a stochastic sampling process).