Début de thèse : 09/12/2024
Fin de thèse : 08/12/2027
Date de soutenance prévue : –
Résumé
To investigate the possibility of integrating learning models as part of the CFD workflow. Traditional numerical solvers come with a high computaional cost. The goal is to design surrogate models to approximate solutions to PDEs faster. To facilitate this problem, we employ, physics-informed machine learning models which have both a data-driven component, as well as, a PDE component to adhere to solutions that make sense in the framework of the physical system.
Our work is in the context of, but not limited to, approximating flow fields around automotive vehicles. This is particularly challenging because we aim to account for the effect of changing, parametrized shape. More formally, it translates to, solving boundary-value-problems with a changing boundary.
Mots clés
Machine learning, partial differential equations, physics-informed learning, surrogate modelling, computational fluid dynamics
Partenaires ou/et Financeurs
Stellantis
Objectifs de développement durable concernés
Publications
- Approximation of functions satisfying partial dierential equations is paramount for simulation of physical uid ows and other problems in physics. Recent studies related to physics-informed machine learning [1] have proven useful as a data-driven complement to numerical models for Partial Dierential Equations (PDEs). However, their eciency and convergence depend on the availability of vast training […]
- Approximation of functions satisfying partial dierential equations is paramount for simulation of physical uid ows and other problems in physics. Recent studies related to physics-informed machine learning [1] have proven useful as a data-driven complement to numerical models for Partial Dierential Equations (PDEs). However, their eciency and convergence depend on the availability of vast training […]
- Many situations lead to sparse data or uncertain data. Gaussian processes (GPs) offer a mathematically funded framework for learning, optimizing, and other decision-making activities. This poster, presented at the HCERES* audit of the LIMOS laboratory in 2025, summarizes several contributions of the St-Etienne applied mathematics team of the LIMOS made between 2023 and 2025. * […]
- The problem of estimating fluid flow around obstacles has a wide range of applications such as airfoil design in aerospace engineering [6] car shape optimization or the design of combustion engine (where the flow is internal) in the in automotive engineering [1], the control of processes in the food industry [13], the understanding of bacterial […]
- […]
- […]