Multiscale analysis and inverse design of uncertain meso-structures (Meso-AID)

We aim to synthesize a variety of microstructures that cover potential candidate materials for material design tasks. In order to account for the intrinsic stochasticity of materials, variations of the microstructure need to be considered. Further, low-dimensional descriptors that on the one hand describe and characterize the microstructure and, on the other hand, enable the recovery of the microstructure from the descriptor alone.

The core activities in this part are at Université Libre de Bruxelles by doctoral student Sri Harsha under the guidance of Prof. Thierry J. Massart.

The inverse design of materials relies heavily on simulations. Many in silico prototypes must be studied before the desired properties can be achieved. Therefore, many simulations must be performed. We start out by performing high-fidelity simulations for select microstructures using optimized sampling procedures. Based on the available precious data, robust and reliable reduced order models (ROM) will be trained. These models are then deployed to boost the amount of quality data by order of magnitude for the subsequent stages of the project.

The core activities in this part are at University of Stuttgart by doctoral student Tanmay Chitnis under the guidance of Prof. Felix Fritzen.

Ultimately, the development of micromechanically motivated machine learned models is anticipated. We focus on Deep Material Networks (DMN) that will accept the microstructure descriptors as additional inputs. Thereby, they can generalize the constitutive behavior in the materials space, by which we denote the manifold of candidate microstructures to be explored.

The core activities in this part are at Université de Liège under the guidance of Prof. Ludovic Noels.

Contribution by Felix Fritzen and Tanmay Chitnis are funded by Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) within the WEAVE program – FR2702-11/1 – project number 530808823, and under Germany‘s Excellence Strategy – EXC-2075 – project number 390740016.