Bandeau image

ML4CFD: ML for CFD

ML4CFD: ML for CFD

  • The project
  • Contacts
Chapo
Scientific Machine Learning (ML) applied to Computational Fluid Dynamics (CFD) simulation.
Contenu
Ancre
The project
Corps de texte

The project is the result of a long term ongoing collaboration between IFPEN and the TAU team of Inria.
IFPEN is a multi-disciplinary French research institute dedicated to new energy and environment technologies. It has launched a new project, named ACAI (Acceleration of Computations through Artificial Intelligence), that  coordinates several data scientists and applied researchers to combine state of the art machine learning with high performance computing in CFD, Computational Mechanics or Subsurface Reactive Transport Simulations.
The INRIA-Saclay TAU team (TAckling the Underspecified) is well known for its activities in machine learning, stochastic optimization, and more generally, artificial intelligence. One of its major topics is the application of machine learning methods to scientific computing problems.

 

This research project aims to enhance the performance of CFD simulations using  machine learning.

 

Employing ML models in CFD simulations with complex physics (e.g, combustion, reactive phenomena, multiphase flows, etc.), can help to speedup existing algorithms, e.g., by creating surrogate models for complex phenomena. The goal of this project is to investigate both approaches:

  • improving spatio-temporal schemes, preconditioning linear solvers and forecast constrained dynamics
  • replacing demanding computations in presence of spatial discontinuities, small scale phenomena or extremely fast events.

These approaches might allow a more adequate adaptation of the sizes of the spatio-temporal evolution steps by taking into account the estimation of future interactions.