Data-driven Stochastic 3D Microstructure modeling for learning mechanical properties
PhD thesis funded by a grant from French-German research agencies, in partnership with Ulm University
Centre de Morphologie Mathématique, Mines ParisTech (Fontainebleau) - Centre des Matériaux, Mines ParisTech (Évry)
The objective of this research project is twofold. First, we want to assess and understand the influence of the polycrystalline microstructure on fracture, and on the ductile and brittle response of γ-TiAl alloys. Second, and more generally, we want to implement tools allowing rapid characterization and exploration of the mechanical behavior of polycrystalline structures.
The research program centers around three main tasks, 3D and 4D imaging, microstructure modeling, and the prediction and exploration of fracture by statistical learning. The second task, which concerns the simulation of virtual microstructures, will take place at the University of Ulm through a parallel thesis, which is part of the same research project, while the other two tasks will be the subject of the thesis at the Mines.
The first task concerns the acquisition and analysis of images of polycrystalline microstructures and local orientations in 3D of samples of γ-TiAl materials as well as in-situ microtomography at the synchrotron allowing access to the cracking facies in 3D+time.
In the second task, polycrystalline microstructures will be characterized and virtually simulated using realistic random partition models, taking into account particle size distribution, grain shapes and crystallographic texture. On a smaller scale, the model will be completed by the addition of twining.
The third task will concern the prediction of the failure behavior, performed using a combined phase-field/FFT method, the predictions of which will be compared with experimental data from two materials with distinct mechanical behavior.