Pierre
TRONTIN

Pierre Trontin was graduated from ISAE-Supaero. After completing his PhD thesis at Onera on the numerical simulation of two-phase flows with interfaces, he was appointed as a research engineer at Onera. He worked for 10 years at the Multi-Physics Department for Energy (DMPE), in particular on the modeling of in-flight icing. He joined UCB Lyon 1 and the Fluid Mechanics and Acoustics Laboratory (LMFA) in 2021. He works on the simulation of two-phase flows.

Multiphase flows involving several immiscible phases such as a suspension of droplets or solid particles, a plume of bubbles, the impact of a jet, the disintegration of a liquid sheet, the runback and wetting of a film on a wall, or the accretion of ice, are involved in many sectors. Applications include industry (fuel atomization in a combustion chamber, fluidized bed reactor, paint spreading, aquaplaning, aircraft anti-icing, ...), health (nebulization and transport of medicinal aerosols, ...) or environment (spraying, flood forecasting ...). These flows are by nature multi-physical coupling many effects such as dynamics, thermics, mass and heat transfers (phase changes) or fluid/wall interactions. In addition to these couplings, there is a multi-scale nature with no obvious separation between large and small scales.
 

        
Figure 1: Falling liquid film destabilization into rivulets (reprinted from the PhD thesis of J. Lallement)
 

Figure 2: Sheared liquid sheet fragmentation (reprinted from IJMF 2010).

The numerical simulation of a realistic multiphase flow involves a large number of degrees of freedom, which makes the simultaneous resolution of all the scales involved expensive or even prohibitive. The large-scale simulation is an approach where only the largest scales are solved with a subgrid model to take into account the action of small scales on the larger ones. Modeling the closure of subgrid scale terms remains a major challenge for two-phase applications where additional difficulties appear compared to the single-phase case, notably related to additional couplings with interfacial scales. In the so-called structural approach, the subgrid contributions are constructed by evaluating the unresolved fields from the resolved ones (ADM or Approximate Deconvolution Model for example). In this approach, we propose to use artificial intelligence tools for the construction of subgrid terms. From high-fidelity direct numerical simulations (DNS) where all scales are resolved, a neural network is trained to reconstruct subgrid information (an interfacial area density for two-phase flows, for example) as a function of the resolved fields (such as the liquid volume fraction). Preliminary results have been obtained in the context of a flat sheared liquid sheet in a periodic domain.                                                      

Bibliography:

- L. Bennani, P. Trontin, R. Chauvin, and P. Villedieu. A non-overlapping optimized Schwarz method for the heat equation with non linear boundary conditions and with applications to deicing. Comput. & Math. with Appl., 80(6) :1500–1522, 2020.
- P. Trontin, J. Lallement, and P. Villedieu. A conservative Saint-Venant type model to describe the dynamics of thin partially wetting films with regularized forces at the contact line. ESAIM : Procs., 69 :79–103, 2020.
- R. Chauvin, L. Bennani, P. Trontin, and P. Villedieu. An implicit time marching galerkin method for the simulation of icing phenomena with a triple layer model. Finite Elements in Analysis and Design, 150 :20–33, 2018.
- P. Trontin, S. Vincent, J.L. Estivalezes, and J.P. Caltagirone. Direct numerical simulation of a freely decaying turbulent interfacial flow. Int. J. Mult. Flow., 36 :891–907, 2010.