Wassim ABDEL NOUR
Topology optimization of a distributor for a high efficiency heat exchanger
Heat exchangers are the cornerstone of many energy and industrial systems. For this reason, many studies aim to improve their effectiveness and / or reduce their mass and size. On the other hand, the use of gases with low impact for the ozone layer (low Global Warming Potential – GWP) are booming in various sectors including aeronautics. However, this substitution requires the construction of new standards for the manufacturing of thermal systems such as heat exchangers. The aim of this thesis is therefore to make a topological optimization of the distributors and collectors of a high efficiency multiphase flow heat exchanger, using a low GWP refrigerant.
Modeling and simulation of acoustical-thermomechanical couplings in complex fluids
The PhD is focusing the effect of power ultrasound on the behaviour of the materials. The constitutive equations of certain complex materials are currently assessed by rheological studies in particular in the nuclear industry for the management of the nuclear waste. Furthermore, for certain complex materials,it can be useful to control their behaviour through active mechanisms such as ultrasound. Indeed, recent studies have shown that the application of ultrasound helps to fluidize several types of systems. The complete understanding of the coupling between complex materials and acoustic solicitation is not yet acquired. Hence, this PhD aims to build a framework, with both theoretical and numerical analysis,to understand the phenomena in this type of material. The studies will be validated with the help of experiments carried out in parallel in other laboratories like CEA, ENS of Lyon or MIT. Besides, in parallel, I work for a curriculum for the French public administration named Corps des Mines during which a 10 months mission with Total is programmed.
Finite element methods for the two-phase aerothermal simulation of additive manufacturing applications such as cold spraying for coating materials
The objective of this thesis is to develop a 3D finite element solver in order to simulate two-phase compressible flows for the cold spray process. The results of the project will be used to qualify this technology for all potential applications. A first application consists in optimizing nozzle geometries (straight and bent), the second one will focus on the study of the effect of the projection angle and the projection distance with the current nozzle and new designs on the development of shock wave and its consequences on the projection efficiency.
George EL HABER
Using deep learning on localized area of interest for computational mechanics: classification of known environments or extrapolation for new situations
The purpose of this thesis is to develop deep learning methods so that information from previous computed simulation is utilized to speed up the solution of new problems. Various industrial problems are targeted in this thesis with focus on cavity filling application with viscous fluid. Two cases exist were in the first case, the targeted solution fits in an already defined deep learning model class. In the second case, were no model class is already determined, we wish to aggregate information resulting from fine simulation in a coarse finite element model.
Multiphase flow and high-fidelity simulation for Loast Foam casting
Lost Foam casting is a type of evaporative-pattern casting process used to make products with complex shapes. The objective of this thesis is to model and develop numerical tools for the prediction of the different phenomena that occur during this process. The work will be based on computational methods coupled between fluid dynamics and heat transfer using massively parallel finite elements developed in the laboratory.
Theoretical and experimental studies of phase separation in oxides to form nanoparticles in optical fibers
In order to enhance optical fibers optical properties, nanoparticles can be formed inside the core by phase separation. But the variations of the obtained nanoparticle composition regarding to their radius is not explained by the Classical Nucleation Theory (CNT). Therefore, in this thesis we aim to apply the Generalised Gibbs Approach (GGA) – which has only been applied to model systems until now, describing these composition variations – to a real oxide system. To conduct this study, chemical potentials of oxydes of the thermodynamic systems SiO2-AO (with A = Mg, Ca or Sr) will be investigated. Then a numerical tool will be developped to determine the kinetics of phase separation using the GGA. At the same time, optical glass fibers of composition of interest will be synthesised and caracterised to validate the GGA in real thermodynamical systems.
Improving confidence on CFD results by deep learning
We aim to evaluate the use machine learning technique on a validation basis of code_saturne, an open-source CFD code developed by EDF, in order to automatically detect in which configurations of model, mesh, and flow structures a given parameterization of the code could be optimal, and thus either suggest (in the context of a choice of models) or automatically adapt the tool settings (in the case of digital parameters influencing precision and performance). This is a medium to long term goal, as it may require extensive learning and the development of suitable quality metrics and heuristics. One intermediate goal would consist in using a learning process to better automate the comparison of the computational and experimental results on a validation basis, in order to detect any regressions induced by the models in the context where numerous modifications induce slight differences in the results produced. The ability to select the configuration giving the best results or the best cost / precision ratio would also be an interesting milestone.