Project Description
PhysimAI is an integrative ANITI program aimed at exploring the use of AI for the physical simulation of engineering systems in order to accelerate and improve their design.
The program involves significant industrial challenges, as physical simulation is currently a computationally expensive step, and its efficiency has a direct impact on the quality and speed of the design process.
It also presents major scientific challenges: data is not available in large volumes, the learning problem is extremely high-dimensional, and unlike traditional simulations, AI inherently does not follow physical laws, raising concerns about the reliability of the results produced.
PhysimAI is structured around several research axes:
- Designing more efficient AI models for high-dimensional data, specifically tailored to mesh-structured data used in physical simulation.
- Accounting for epistemic uncertainty associated with these models.
- Introducing physical biases during the learning process or in the model design.
- Hybridizing traditional simulation codes with AI.
These scientific directions will be explored using an open-source development methodology to structure the research and facilitate its transfer to industry.
Academic Partners:
Industrial Partners:
English 
French