LPM

Data-driven ML methods for numerical simulations have shown promise in overcoming the computational costs of such simulations, but struggle with scalability and versatility. The project aims to develop a Foundational Large Physics Model for fluid simulation, i.e., a single pre-trained model based on the transformer architecture capable of solving $3$D simulations on various geometries, under a wide range of flow conditions, only requiring light fine-tuning to optimally adapt the model to each specific problem.