Soutenance de thèse Elena LOPEZ TOMAS- “Towards efficient numerical techniques for real time image-based thermal and flow computational upscaling in composites manufacturing.”

RÉSUMÉ:

In the context of real time control of structures during production or processes adaptability for the reduction of defaults, simulation and properties’ homogenization based on real images seems to be promising techniques. However, they involve three challenges: I) the treatment of real images, II) the use of adapted homogenization techniques to the involved physics, III) the “real time constraint” to be truly usable. Conventional asquisition technologies create enormous amounts of data that need to be treated in order to achieve the “real time” objective. For this, it is pertinent to develop techniques to extract the required information from that data for decision making in real time. Concerning thermo-mechanical homogenization, the involved physics are similar in the micro and macro scales and classical homogenization can be used. However, in the case of flows in porous media (impregnation), the physics are different and new techniques must be proposed. Combining model reduction techniques and high performance computing, the computation cost is alleviated but still unable to achieve the real time objective. The use of “machine learning” techniques will allow to obtain the solution fields in real time.

JURY:

Directeur de thèse:

M. BINETRUY Christophe, Professeur des universités, École Centrale de Nantes

Co-encadrants:

Mme ABISSET-CHAVANNE Emmanuelle, Maître de conférences, École Centrale de Nantes
M. CHINESTA Francisco, Professeur des universités, École Centrale de Nantes

Rapporteurs:

M. BOISSE Philippe, Professeur des universités, INSA de Lyon
M. CUETO Élias, Professeur Université de Saragosse (Espagne)

Examinateurs:

M. KEUNINGS Roland, Professeur Université Catholique de Louvain (Belgique)
M. UPADHYAY Ram, Ingénieur, General Electric Global Research Center, Niskayuna (États-Unis)