Student Projects & Theses
Projects Offers
Please do not hesitate to contact the responsible person (see link to PDF above) with questions or directly Prof. David Kammer at .
We are always happy to meet and discuss the project topic as well as our expectations.
We also welcome spontaneous proposals and personal ideas for projects.
Past Projects
Leveraging avant-garde computational techniques for deciphering crack nucleation
David Arroyave Madrigal (Fall 2025)
This project studies crack nucleation in slender elastic solids, where fracture initiation is governed by confined two-dimensional defects embedded within a plane. Using three-dimensional finite element simulations, it investigates the influence of crack geometry through circular and elliptical nucleation patches. The work is framed within confined-geometry fracture theory and focuses on the relationship between defect shape, elastic energy storage, and stress fields. The study aims to provide a numerical framework for analyzing crack nucleation in confined elastic bodies with non-circular defects.
Accelerating Thermodynamic Modeling with GPUs and Automatic Differentiation
Luca Marei Endell (Spring 2025)
This project explores the use of GPU acceleration and automatic
differentiation to improve the performance of thermodynamic solvers for
modeling chemical equilibria in multiphase systems, such as those
involved in steel corrosion in concrete. The solver is based on the
minimization of Gibbs free energy, formulated as a constrained nonlinear
optimization problem using interior point methods and Newton's method.
It features a modular design, flexible constraint handling, and
integration with the PourPy package for generating Pourbaix diagrams.
Leveraging Automatic Differentiation for Large-Deformation Analysis in Structural Mechanics
Mateo Luzuriaga Merlo (Spring 2025)
This project develops a differentiable finite element framework for the
structural analysis of systems made of Euler–Bernoulli beams, leveraging
automatic differentiation (AD) in structural mechanics. The method is
based on potential energy minimization using Hermite shape functions and
a co-rotational formulation to handle large deformations. Implemented
with the JAX framework, the approach uses AD and just-in-time (JIT)
compilation to compute Jacobians efficiently and accelerate the updated
Lagrangian solution process.