Simulation/machine learning hybrid modelling (HSA)

Home > Research activities > IA2 program > HSA

Project description

Hybridization of physical simulation and machine learning methods for complex systems design and control.

Numerical simulation represents today a very important tool in the design and control of physical systems. However, in an increasingly complex context with several disciplines and stakeholders, simulation processes must evolve in order to improve the quality of the decisions made to design and control complex systems.
Launched in February 2020 for a duration of four years under the vast research program IA2 (Artificial Intelligence and Augmented Engineering), the HSA project aims to design new approaches based on artificial intelligence to hybridize conventional physical simulation processes by coupling them with machine learning. These new approaches will be applied on the various industrial use cases of the project, with a certain synergy that will allow a better generalization of these methods to other fields of physical modelling.

Expected results

The main objective of the project is to design new learning approaches to hybridize physical simulations. For instance, this should allow to:

Reduce the cost of the simulation, through the creation of substitution models (also called surrogate models) based on statistical learning.
Improve the quality of decisions in simulation based desgin.
Tackle physical problems that are difficult to solve with conventional modeling methods (such as inverse problems for example).