Current Projects

Institute of Lightweight Engineering, University of the Bunderwehr Munich

Bundeswehr Research Institute for Materials, Fuels and Lubricants (WIWeB)

The InterPlaS project investigates how various damage mechanisms in fiber-reinforced composites - such as plasticity, crack formation, and delamination - influence each other and shape the material behavior up to fracture. To this end, numerical models are being developed on the micro and macro scales that capture both the physical processes in the fibers and the matrix and describe their interaction in the laminate. A particular focus is on predicting delamination under complex loading conditions and quantifying unavoidable uncertainties resulting from material variations or manufacturing tolerances. Modern probabilistic methods are used for this purpose, allowing a more realistic assessment of model quality. The aim of the project is to create a deep understanding of multiscale material behavior and, at the same time, to provide computational methods that enable reliable statements to be made about the safety and service life of components made of fiber-reinforced composites.

Contact is central to many engineering applications and governs force transfer, separation, sliding and friction between interacting bodies. Its nonlinear nature makes accurate simulations computationally demanding, especially for complex three-dimensional models and large deformations.

The PICO project investigates physics-informed machine-learning methods for contact mechanics. Starting from physics-informed neural networks (PINNs), the project plans to investigate energy-based PINNs (EPINNs) and variational formulations that provide closer links to established finite-element methods. The developed approaches will be studied for two- and three-dimensional contact problems, including small and large deformations as well as frictionless and frictional contact.

A central goal is to combine the flexibility of machine learning with the accuracy and reliability of classical numerical methods. Trained physics-informed models will therefore be investigated not only as standalone solvers, but also as components of hybrid solution strategies that combine them with classical finite-element-based solvers.

Czech Academy of Sciences in Prague

University of West Bohemia in Pilsen

This project investigates metamaterial properties of locally periodic structures constituted by conventional and smart materials. Smart materials distinguish themselves from standard passive metamaterials by controllability through electroactive components and time-space alternating geometrical layouts.

Controllable and adaptive structures equipped with embedded actuators and sensors connected to electric circuits provide a higher level of multi-functionality such as the capability of vibration and wave propagation control, shape morphing, or modifying stiffness in space and time.

The research work is embedded into an international consortium and will be conducted in close collaboration with scientific partners at the Czech Academy of Sciences in Prague (Czech Republic) and at the University of West Bohemia in Pilsen (Czech Republic).

Dr.-Ing. Sebastian Brandstäter, Moritz Frey M.Sc.

PD Dr.-Ing. habil. Matthias Mayr, Martin Frank M.Sc.

PD Dr.-Ing. habil. Matthias Mayr, Max Firmbach M.Sc.

The project focuses on the approximate calculation of eigenvalues of various differential operators on areas with cracks using isogeometric analysis. Methods are to be developed that allow the simulated eigenvalues to represent as accurately as possible the eigenfrequencies of test objects with cracks that can be measured in practice, and also to show corresponding convergence results. Subsequently, these natural frequencies shall be used to identify the crack, i.e., to determine the shape of the crack. This inverse problem is to be solved with a neural network, for the training of which we need simulation data of as high quality as possible in order to be able to guarantee a meaningful application in practice.

Prof. Dr. Thomas Apel, Philipp Zilk M.Sc.

In this project, methods for hybrid digital twins of critical infrastructure are investigated and developed. The main focus is the combination of physics-based modeling using Finite Element Methods (FEM) and data-based modeling with machine learning techniques.

The first application case are steel-reinforced concrete bridges. On the physics-based modeling side, a mixed-dimensional FEM model for steel-reinforced concrete components is developed. Complementary, variants of neural networks are investigated. Both are combined to improve the material model with measurement data and obtain a reduced order model of the physical system. 

Still, attention is paid to generality and application independence of the methods, so that the developed techniques can also be used by the project partners. New algorithms are implemented in the form of an in-house software solution. An associated hardware platform for high performance computing is provided at the Data Science & Computing Lab.