PhD defence de Tiphaine Houdard

13 July 2022

Tiphaine Houdard defends her PhD in Computational Mechanics and Materials on July 13th, 22

Thermo-mechanical analyses of the solidification of electrofused alumina-zirconia-silica refractory blocks

Tiphaine Houdard conducted her PhD work in the 2MS and CFL CSM teams. She defends her PhD in Computational Mechanics and Materials on July 13th, 2022 in front of the following jury:

– Dr. Pierre Benigni, Universités Aix-Marseille et Toulon, IN2MP

– Prof. Emmanuel De Bilbao, Polytech Orléans, CEMHTI

– Dr. Julien Zollinger, Université de Lorraine – Institut Jean Lamour

– Mme Isabelle Cabodi, Saint Gobain Research

– Dr. Franck Pigeonneau, Mines Paris

– Dr. Charles-André Gandin, Mines Paris

Abstract:

Soldier blocks in glass furnaces are mainly composed of Al2O3 – ZrO2 – SiO2, they are therefore usually called AZS blocks. They are produced by electrofusion. This project is studying the solidification of those blocks after being poured in their mold. An objective is to anticipate the susceptibility of hot tears occurrence during the cooling down. A second objective is the prediction of the macro-porosity shape inside the riser.

First, a study of AZS material was done to validate the thermodynamic data obtained by TCOX10 database. It predicts the formation of mullite, that is not observed experimentally. Heat treatments validate however the thermodynamic results, and indicate a kinetic of crystallization. For this study, mullite is rejected to extract the relevant data for simulations such as the solidification path, the composition, the density and the enthalpy of the phases (zirconia, corundum and glassy phase, considered as the remaining liquid phase at low temperature).

Temperature history has been measured on industrial blocks and is simulated with THERCAST® software. To simulate the temperature history, it is required to tabulate the temperature properties of the different domains and to adjust numerical settings.

Block mechanical behavior is modelled to predict the hot tears formation, defects located at the edge of the blocks. After experimental observations, several criteria to predict the susceptibility of tear formation are compared. Cumulated strain over a given temperature range is considered as the most relevant. A study compares the values obtained depending on the temperature range or the visco-plastic consistency considered. After optimization, computations predict correctly the hot tears susceptibility for blocks at industrial scale.

To get a relevant shape of the macro-porosity, thermomechanical computations with THERCAST® show that the composition, which impacts deeply the liquid phase density, has to be known locally. Macrosegregation needs therefore to be implemented. An experimental map of chemical compositions is performed. The properties tabulations for concentrated oxide alloys required an adaptation of the PATH tool from PhysalurgY® library. by calling CIMLIB® solvers from THERCAST® for the thermal, hydraulic and solutal (for Al2O3, SiO2 and Na2O present in zirconia, corundum and glassy phases) computations, it is possible to model, considering an incompressible domain, the thermo-solutal convection during the cooling down, with sedimentation of crystals neglected. Numerical models anticipate the formation of segregated channels going upward from the skin to the core of the block and also the enhancement of glassy phase proportion inside the riser. However, the compressibility, to model the impact on the riser macro-porosity shape, had not been implemented.

Keywords: solidification, hot tears, macrosegregation, AZS, THERCAST, numerical modeling

PhD defence of Han Wang

8 July 2022

Han Wang defends his PhD in Computational Mechanics and Materials on July 8th, 22.

Study of the solidification of lipsticks and the resultant mechanical behavior. Application of numerical simulation to the forming process.

Han Wangconducted his PhD work in the 2MS and PSF teams under the supervision of Michel Bellet and Séverine A.E. Boyer. He defends his PhD in Computational Mechanics and Materials on July 8th, 2022 in front of the following jury:

– M. Roberto PANTANI Université de Salerne Rapporteur

– M. Philippe COUSSOT Ecole des Ponts Paris Tech Rapporteur

– Mme Véronique MICHAUD École Polytechnique Fédérale de Lausanne Examinatrice

– Mme Isabelle HÉNAUT IFP Energies Nouvelles Examinatrice

– Mme Florence DALLE Parfums Christian Dior Invitée

– M. Guillaume FRANCOIS Transvalor S.A. Invité

Abstract:

Lipsticks are formed by casting process. To understand and improve the industrial casting process, the solidification of model materials of lipsticks are studied by experimental characterization and numerical simulation. Experimentally, the crystallization kinetics of wax components on rapid cooling conditions is characterized and modelled by using the Differential Scanning Calorimetry (DSC). The mechanical constitutive model according to the progress of the solidification process is determined with different rheological testing protocols (Couette rheological tests and indentation tests). The crystallization morphology of different wax components is correlated to different cooling conditions. Numerically, the kinetics model of crystallization and the mechanical behavior law are associated and integrated in finite elements based numerical calculations. Thermomechanical phenomena of the casting process and probable causes of cast defects are analyzed by numerical simulation of the industrial process.

Keywords: lipstick, casting process, crystallization kinetics, mechanical behavior, crystallization morphology, numerical simulation

PhD defence of Ali-Malek Boubaya

5 July 2022

Ali-Malek Boubaya will defend his PhD in Computational Mathematics, High Performance Computing and Data on July 5, 22.

New interface capturing and boiling framework modeling for industrial cooling

Ali-Malek Boubaya conducts his PhD work in the CFL team. He will defend his PhD in Computational Mathematics, High Performance Computing and Data, on June 24th, 22 in front of the following jury:

Pr. Stefanie ELGETI, Technischen Universitat Wien

Pr. Franck BOYER, Université de Toulouse

Pr. Médéric ARGENTINA, Université Côte d’Azur

Pr. Elie HACHEM, CEMEF Mines Paris

Dr. Aurelien LARCHER, CEMEF Mines Paris

Abstract:

This thesis is part of the desire to have a numerical framework dealing with complex industrial cases. For this reason, it is necessary to develop new solvers that enrich the library of methods currently used in the team. This desire relies on the will to consider a new approach of interface capturing enabling the obtaining of new information about the physics of the system and bringing numerical robustness.

The phase-field method is based on an energetic approach of the interface separating the phases that which is not the case for the currently used interface capturing method. Furthermore, the method owns good properties to model the capillary action, involved in the wetting, and consider contact angles. Directed by a perspective of industrial applications, this interface capturing aims to be unified with other physical models for a multiphysics characteristic and a multiple applications to the industrial numerical simulation.

Keywords: Mathematics, Numerical analysis, Comutational Fluid Dynamics, High Performance Consulting, Thermic

PhD defence of Rémy Gérard

4 July 2022

Rémy Gérard will defend his PhD in Computational Mathematics, High Performance Computing and Data on July 4, 22.

Fast, flexible surface to surface thermal radiative transfer for sintering applications

Rémy Gérard onducts his PhD work in the CFL team. He will defend his PhD in Computational Mathematics, High Performance Computing and Data, on July 4th, 22 in front of the following jury:

– M. Joan Baiges Aznar, Universitat Politècnica de Catalunya, Barcelona

– M. Charbel Abchi, Notre Dame University, Liban

– M. Assaad, Mines Paris – Centre Efficacité énergétique des Systèmes

– M. Elie Hachem, Mines Paris, CEMEF

– M. Aurélien Larcher, Mines Paris, CEMEF

Abstract:

Properly taking into account thermal radiative transfer matters a great deal to properly model numerous industrial processes. This thesis focuses on setting up a proper modelling for it in certain industrial configurations where some hypotheses are met: grey, homogeneous materials and non-transmitting medium. The Surface-to-surface approach we have chosen allows for such an accurate solving of thermal transfer. We make use of immersed meshes and implicit object definitions by the levelset function to allow for an adaptable model and an implicit definition of the radiating surfaces. We couple radiative transfer with other forms of thermal transfer in a P1 finite elements methods. After validating our model on numerous simple test cases, we set up ray tracing to accelerate obstruction computation and organize the radiating facets into a kd-tree. We conclude by ensuring our solver is highly scalable on parallel computing and show simulation cases of real industrial processes.

Thermal radiation between two cubes separated by a vacuum

Keywords: Thermal Radiation, S2S, Finite Elements, Kd Tree, Immersed Mesh, Ray Tracing

Phd defence of Mohamed Mahmoud

30 June 2022

Mohamed Mahmoud defens his PhD in Computational Mechanics and Materials on June 30, 22.

A 3-D Multi-physics computational model for thin sheet metal forming processes : Application to deep drawing and magnetic pulse forming processes

Mohamed Mahmoud conducted his PhD work in the CSM team under the supervision of François Bay and Daniel Pino Munoz. Il defends his PhD in Computational Mechanics and Materials on June 30th, 22 in front of the following jury:

– Prof. Diego Celentano, Pontificia Universidad Católica de Chile

– MC Tudor Balan, ENSAM Metz

– Prof. Guillaume Racineux, Institut de Recherche en Génie Civil et Mécanique – UMR6183, Ecole Centrale Nantes

– IR Christine Beraudo, Transvalor

Abstract:

The core of this work is focused on the development of an efficient 3D computational tool for modeling thin sheet metal forming processes using advanced remeshing and parallel computations techniques.

One of the main topics lies in the implementation of a prism division algorithm and a prismatic solid-shell element formulation in the tetrahedral-based FORGE3 software. A partitioning algorithm has been adapted in order to enable distributed memory computation.

The proposed methodologies offer a numerical tool that is capable of handling various sheet metal forming applications such as: (1). Unconstrained cylindrical bending problem for a highly bending-dominant thin structure plastic deformation; (2). Deep drawing process in which the anisotropic plastic behavior of sheet metal comes more prominent and affects the accuracy of the predicted earing profile; (3) Electromagnetic forming processing with the direct and indirect forming process which is a direct application of the multi-physics interaction between the mechanical solver and Electromagnetic solver.

Comparisons with standard mixed finite element formulations have been performed and show the superiority of solid-shell elements for most thin sheet metal forming processes with dominant high bending effects. These computations also enable a large reduction of computational time while retaining high accuracy. Moreover, a new remeshing strategy for enabling the tetrahedral remesher to generate prism-compatible meshes for the new element has been developed and will open the way for additional sheet metal forming applications.

Keywords: Solid-shell finite elements, Reduced integration, Anisotropic plasticity, Sheet Metal, Electromagnetic Forming, Deep Drawing

PhD defence of Hassan Ghraieb

24 June 2022

Hassan Ghraieb will defend his PhD in Computational Mathematics, High Performance Computing and Data on June 26, 22.

On the coupling of deep reinforcement learning and computational fluid dynamics

Hassan Ghraieb conducts his PhD work in the CFL team. He will defend his PhD in Computational Mathematics, High Performance Computing and Data, on June 24th, 22 in front of the following jury:

– Prof. Ramon Codina, Department of Civil and Environmental Engineering, Universitat Politècnica de Catalunya, Spain

– Prof. Anil Anthony Bharath, Imperial College, London, UK

– Dr. Nissrine Akkari, SafranTech, France

– Dr. Anca Bleme, Sorbonne University, France

– Prof. Elie Hachem, CEMEF Mines Paris, France

– Dr. Philipe Méliga, CEMEF Mines Paris, France

– IR Jonathan Viquerat, CEMEF Mines Paris, France

Abstract:

This thesis gauges the relevance of deep reinforcement learning (DRL) techniques for the optimal control of fluid mechanical systems. Reinforcement learning (RL) is the process by which an agent learns by trial and error interactions with its environment the succession of actions maximizing its cumulative reward over time. In a deep reinforcement learning context (deep RL or DRL), the agent is a deep neural network based on the neural circuits formed by neurons in the human brain. The coupling between state-of-the-art DRL algorithms and computational fluid dynamics (CFD) solvers and their implementation in a high performance computing context make for the novelties and main objective of the thesis. The CFD resolution framework used to compute the reward provided to the DRL agent relies on the Variational Multiscale (VMS) stabilized finite element method. The latter introduces an a priori decomposition of the numerical solution into large and small-scale components, the general picture being that only the large scales and resolved at the discrete level, while the effect of the small scales is modeled after consistently derived source terms proportional to the residual of the large scale solution. Regarding the DRL algorithms, two different frameworks are considered. The first one has the agent interact only once per episode with its environment to learn the mapping from a constant input state to an optimal action (hence, single-step episodes, and by extension, single-step DRL), and is thus relevant to open loop control, where a desired output is optimized under pre-determined actuation parameters (for instance, a constant inlet velocity). The second one has the agent interact multiple time per episode to learn a more complex state-action relation (hence, multi-step DRL) and is more relevant to closed-loop control, where the output is optimized by continuously adjusting the design parameters to flow measurements. Several test-cases in two and three dimensions (both in laminar and turbulent flow regimes) are successfully tackled and presented to assess the relevance, accuracy and performance of the proposed methodologies, with particular emphasis put on drag reduction and thermal control applications. The obtained results emphasize the high potential of the DRL-CFD framework, and are expected to contribute to further progress towards improved and faster design and control of industrial fluid mechanical systems.

Keywords: Deep Reinforcement Learning, Neural networks, Computational fluid dynamics, Flow control, Thermal control