Title: Efficient thermal numerical simulation and decision-making for ceramics in new satellite generation
Date: 2022-03-15 14:39
Authors: Elie Hachem / Aurélien Larcher
Job_Location: Sophia Antipolis
Job_Duration: 3 years
Job_Employer: Mines Paris / Thales / CNES
Thales Alenia Space has been using technical ceramics in its optical instruments for over a decade. The most used ceramic is Si3N4 (silicon nitride) in the constitution of telescope structures, because it has very good properties of thermo-elastic and water stability. However, understanding the manufacturing process to better control its stages, and to reduce the risks associated with the experimental process remains a challenge. In particular, the most complex stage from a scientific point of view is sintering with temperatures up to 2000 ° C and pressures up to 10 bar inside a dedicated oven.
The interest in space industry is twofold:
(1) controlling the environment in the furnace will provide the information necessary for simulating the behavior of the part during sintering, and the prediction of its final state.
(2) a better understanding of various parameters that influence the operation of sintering furnaces will support the development of new furnaces, in particular for increasingly large sizes.
In this thesis, we propose a novel coupling between computational fluid dynamics (CFD) and machine learning to tackle these objectives. First, we are particularly interested in volume immersion approaches for 3D conjugate heat transfer computation.
Indeed, this makes it possible to predict very precisely the thermal history of the parts in their environment without using an exchange coefficient or simple boundary conditions, but by calculating exactly the strong solid fluid coupling in particularly the radiative heat transfer with complex loading using a novel approach of surface-to-surface (S2S) radiation in the context of an immersed volume method. The mesh is automatically and anisotropically adapted to the load interfaces, making it possible to improve the precision of calculations carried out in a monolithic way (heat transfer in both fluid and solid domains) and whatever their level of detail and therefore of complexity.
Doing so, Deep Reinforcement Learning in which neural networks are used in the context of decision-making problems for optimization and control, can and will benefit from such a framework to assess and then to optimize different configurations (position of the loads, orientation, temperature regulator…). The project will benefit from the computational resources at CEMEF: two GPU Tesla V100 and ~3000 cores as well as from experimental results made by Thales Alenia Space.
• Master of Science or equivalent in applied mathematics, physics, or mechanical engineering, with competences in fluid dynamics, statistics, or scientific computing
• Good experience in programming (C, C++) and in data post-processing and analysis
• Excellent writing skills, fluent in English
• Rigorous, autonomous and motivated by working at the edge between basic research and industrial applications
The proposed work will be carried out in the research group Computing & Fluids located at the CEMEF Research Center of MINES ParisTech in Sophia-Antipolis, France.
The selected candidate will be awarded a PhD grant of excellence from the CNES (Centre national d’études spatiales) in collaboration with Thales Alenia Space.
This grant covers a three-year doctoral contract (including benefits) that should start in October 2022.