News

A framework that generalizes several existing schemes for parametric PDEs and achieves generalization in out-of-distribution, multiscale problems. More details: https://arxiv.org/abs/2405.19019

We are currently looking for M.Sc. students to work on the topic of "Inverse Materials Design".  This is an interdisciplinary topic that combines computational modeling of physical systems with adavnaced machine learning techniques. More details can be found here

Highlights • We propose a probabilistic, data-driven closure model for Reynolds-Averaged Navier-Stokes simulations. • It is trained with high-fidelity data of mean velocity/pressure and quantifies aleatoric uncertainties in the model form. • It can automatically identify regions where closure is…

Abstract: Random media and the process-structure-property chain generally define complex, high-dimensional and stochastic materials systems, posing a challenging setting for any prediction or optimization task. In this thesis, we pursue a Bayesian approach for learning and predicting the behavior…

Abstract: Solving high-dimensional, nonlinear systems is a key challenge in engineering and computational physics. We propose novel physics-aware machine learning models that rely both on physical knowledge as well as a small amount of data and are, after an initial training phase, able to solve…

Application deadline: March 8th 2023. Details about the position as well as the application process can be found here

Deep Operator Networks (DeepONets) offer a powerful, data-driven tool for solving parametric PDEs by learning operators, i.e. maps between infinite-dimensional function spaces. In this work, we employ physics-informed DeepONets in the context of high-dimensional, Bayesian inverse problems.…

While the forward and backward modeling of the process-structure-property chain has received a lot of attention from the materials’ community, fewer efforts have taken into consideration uncertainties. Those arise from a multitude of sources and their quantification and integration in the inversion…

Abstract: Primary production of metallic iron accounts for a significant share of the industrial green- house gas emissions, and an expected growth in the demand for steel motivates the search for alternative sustainable ways of iron ore reduction. Hydrogen-based direct reduction is a very…

Abstract: In recent years, data-driven approaches have significantly reduced the computational effort required to capture the structure-property linkages in high-contrast microstruc- tures. Convolutional neural networks (CNNs), a special form of artificial neural networks (ANNs), have been shown to…