<p>Due to strong interactions among various phases and among the phases and fluid motions, multiphase flows (MPFs) are so complex that lots of efforts have to be paid to predict its sequential patterns of phases and motions. The present paper applies the physical constraints inherent in MPFs and enforces them to a physics-informed neural network (PINN) model either explicitly or implicitly, depending on the type of constraints. To predict the unobserved order parameters (OPs) (which locate the phases) in the future steps, the conditional neural processes (CNPs) with long short-term memory (LSTM, combined as CNPLSTM) are applied to quickly infer the dynamics of the phases after encoding only a few observations. After that, the multiphase consistent and conservative boundedness mapping (MCBOM) algorithm is implemented the correction the predicted OPs from CNP-LSTM so that the mass conservation, the summation of the volume fractions of the phases being unity, the consistency of reduction, and the boundedness of the OPs are strictly satisfied. Next, the density of the fluid mixture is computed from the corrected OPs. The observed velocity and density of the fluid mixture then encode in a physics-informed conditional neural processes and long short-term memory (PICNP-LSTM) where the constraint of momentum conservation is included in the loss function. Finally, the unobserved velocity in future steps is predicted from PICNP-LSTM. The proposed physics-informed neural processes (PINPs) model (CNP-LSTM-MCBOM-PICNP-LSTM) for MPFs avoids unphysical behaviors of the OPs, accelerates the convergence, and requires fewer data. The proposed model successfully predicts several canonical MPF problems, i.e., the horizontal shear layer (HSL) and dam break (DB) problems, and its performances are validated.</p>

Dans cette thèse, l’étude des interactions entre des particules ultrafines émises par les pots d’échappement et l’écoulement de sillage créé par le véhicule émetteur a été réalisée principalement selon une approche numérique. Une campagne expérimentale a été conduite à des fins de validation. L’objet de la thèse vise à comprendre l’impact des particules issues des pots d’échappement sur l’environnement proche tant du côté piéton que du côté des passagers des véhicules suiveurs. Pour cela, l’écoulement du fluide a été traité avec une approche eulérienne type URANS (Unsteady Reynolds Average Navier-Stokes) combinée à un suivi lagrangien pour les nanoparticules. En effet, cette thèse est conduite en parallèle d’un projet collaboratif financé par l’ADEME (CAPTIHV) dont le but est d’évaluer la qualité de l’air des habitacles des véhicules automobiles, et en particulier de l’infiltration des particules ultrafines issues du trafic environnant. L’étude de la dispersion des particules fines en écoulements turbulents nécessite une analyse fine des structures turbulentes qui s’y développent. Notre étude numérique a donc consisté, en premier lieu, à analyser cette dispersion dans le cas d’un écoulement de sillage classique à l’aval d’un cylindre. Cela nous a permis de caractériser la dynamique d’interactions de nanoparticules solides de carbone avec les structures tourbillonnaires en considérant l’impact de la turbulence et de la diffusion brownienne. Cela a permis d’évaluer l’influence des principaux mécanismes influençant la dispersion. Les résultats de ces simulations nous ont permis de sélectionner les mécanismes/forces importants pouvant influencer la dispersion de telles particules dans le sillage d’un véhicule automobile ; Cela nous a facilité la mise en place et l’analyse des simulations relativement plus complexes de l’aérodynamique du corps d’Ahmed à culot droit en présence des nanoparticules simulant les suies des gaz d’échappement. Les interactions des particules ultrafines avec les structures tourbillonnaires se créant dans le sillage des véhicules ont été évaluées à partir de profils de concentrations et les coefficients de dispersions transversales. La dernière étape a consisté en une campagne d’essais en soufflerie qui nous a permis de caractériser les champs de vitesses moyens et turbulents ainsi que les champs de concentrations particulaires à l’aval du véhicule pour valider les résultats numériques / In this thesis, the study of the interactions between ultrafine particles emitted by the exhaust pipes and the wake flow generated by the emitting vehicle was carried out mainly using a numerical approach. An experimental campaign was conducted for validation purpose. The goal of the thesis is to understand the impact of exhaust particles on the surrounding environment on both the pedestrian and the passengers of the following vehicles. For this purpose, the fluid flow was resolved with an Eulerian type URANS model (Unsteady Reynolds Average Navier-Stokes) combined to the Lagrangian approach for the nanoparticles trajectories calculation. This thesis is conducted simultaneously with a collaborative project funded by ADEME (CAPTIHV) whose purpose is to assess the air quality of automotive car cabins, and particulate infiltration from the surrounding traffic in particular of ultrafine particles. The study of the dispersion of fine particles in turbulent flows requires a fine analysis of the turbulent structures that develop in such flows. Our numerical study therefore consisted, first, in analyzing this dispersion in the case of a classic wake flow downstream of a cylinder. This enabled us to characterize the interaction of solid carbon nanoparticles with vortical structures evaluating at the same time the impact of turbulence and Brownian diffusion. This allowed determining the influence of the main mechanisms influencing nanoparticles dispersion. In a second step, we replaced the cylinder configuration by a simplified geometry of a motor vehicle, Ahmed body configuration. Therefore, simulations with and without of particles presence have been conducted and have allowed to highlight the swirls structures and to characterize the particles dispersion through particle concentration profiles and the particles dispersion coefficients. The results of these simulations allowed us determining the important mechanisms / forces that can influence the dispersion of such particles in the wake of a ground vehicle; this facilitated the implementation and analysis of relatively more complex simulations of the aerodynamics of the square back Ahmed body in the presence of nanoparticles simulating soot from the exhaust gases. The interactions of ultrafine particles with vortical structures appearing in the wake of vehicles were evaluated from concentration profiles and transverse dispersion coefficients. The final step was a wind tunnel experimental campaign that allowed us to characterize the average and turbulent velocity fields as well as the particle concentration fields downstream of the vehicle to validate the numerical results