The main objective of ASSAS is to develop a proof-of-concept for a basic-principles severe accident simulator, that will feature a generic western-type PWR. To do so, the reference European severe accident simulation code ASTEC will be interfaced with Tecnatom’s commercial simulation environment, TEAM_SUITE®. R&D actions will be necessary for the simulator to reach expected performances (especially real time execution), and to explore how it could be adapted to other reactor designs and simulation codes in the future.
The consortium will explore how machine-learning may be used to develop fast and robust surrogate models that could replace severe accident codes partly or globally in the simulator. Different strategies and techniques will be tested in parallel to maximise chances of success, considering the challenges related to severe accident modelling. A comprehensive database of accidental scenario simulations will be created to train the AI models. State-of-the-art regression techniques coupled with sparse dimension reduction, as well as an interpolation engine between pre-calculated sequences will be developed and evaluated.
In addition, efficient programming and mesh optimisation will be used to optimise ASTEC’s performances.