THE SECOND SAUDI INTERNATIONAL CONFERENCE ON NUCLEAR POWER ENGINEERING (SCOPE-2)

Nuclear safety analysis, particularly in the context of severe accidents (SAs), requires modeling complex physical and stochastic phenomena. The Risk-Oriented Accident Analysis Methodology (ROAAM+) framework, under development at KTH, addresses this by integrating full-scale mechanistic models with computationally efficient surrogate models (SMs). However, many existing SMs lack the capacity to capture the inherently dynamic and time-dependent nature of severe accident scenarios. To address this, this study explores the use of machine learning (ML) techniques, focusing on neural operators, specifically Deep Operator Network (DeepONet), to develop SMs that can emulate the behavior of the MELCOR code – a comprehensive tool used for severe accident simulations.