Jennifer EL ZAHABI
Advanced numerical modeling and simulation of mass and heat transfer for Lost Foam casting process
Lost Foam casting is a foundry process aiming to produce high finish quality parts with complex shapes. It consists in using a model made of expanded polystyrene, embedded in a sand mold, which will vaporize when pouring molten metal alloy. When advancing, the metal front can become destabilized with irregular areas causing incomplete vaporization, preventing the evacuation of residues and therefore trapping polystyrene bubbles that can weaken the structure once solidified. The objective of this thesis is then to develop a model and numerical tools of prediction of the phenomena mentioned above. The work will be based on calculation methods coupled between fluid dynamics and heat transfer using massively parallel finite elements developed in the laboratory and supported with experiments.
Fluid-Structure Interaction and Machine Learning for risk evaluation of intracranial aneurysms
Intracranial aneurysms are pathological dilations of blood vessels taking the shape of bulges or balloons mostly on wall sides or at arterial bifurcations. These anomalies bear the risk of rupture and subsequent subarachnoidal hemorrhage, which is associated with high mortality and morbidity rates. As a result, when detecting an aneurysm on an MRI scanner, the question of potential clinical intervention is raised. In the current stage, decisions concerning intracranial aneurysm surgery are mostly based on the observation of shape, size and location of the aneurysm in the brain along with the patient’s age and medical record. However, no characteristics of the hemodynamics in the aneurysm are used in hospitals. Fluid mechanical simulations can yet provide insights into the flow patterns and stress distributions, thus helping the evaluation of aneurysm rupture risks. In the underlying project, a high-fidelity solver will be assembled considering fluid-structure interactions between the vessels and the blood flow, in order to model hemodynamics and wall stresses as accurately as possible. Several constitutive laws will be implemented to model the behavior of arterial walls. In the cases where treatment is required, stents may be employed to divert the flow from the aneurysm sack and aforementioned flow simulations can serve as a basis for designing an optimal patient-specific stent. This will be implemented as part of this project through deep reinforcement learning.
Pablo JEKEN RICO
High-fidelity FSI Simulations of Intracranial Aneurysms and DRL-based Optimization of Flow Diverters
Intracranial aneurysms have a considerable incidence rate in adults and are now often discovered before showing any symptoms. To assess the growth potential, risk and treatment of aneurysms, fluid mechanical simulations of the vessels can be done, which provide insights into the flow patterns and stress distributions. In the cases where treatment is required stents may be employed to divert the flow from the aneurysm sack. A good choice of the stent can drastically reduce the aneurysms exposure and therefore its risk of rupture. In the underlying project, a high-fidelity solver will be assembled considering fluid-structure interactions between the vessels, blood and an optional flow diverter. Several hemodynamic models will be implemented and tested for best performance by comparing simulation results with 4D MRA data. For this study, detailed patient data from 85 people will be used to validate the work. The solver is then coupled with a DRL (Deep Reinforcement Learning) framework to compute a stent configuration tailored to each individual patient. With the thesis project, we aim to provide physicians with a reliable and robust tool that supports the complex decision making in the analysis and treatment of aneurysms.