ASSESS
DOCTORAL CANDIDATE 12
Hybrid Digital-Twins Based on Physics-Constrained Graph Neural Networks for Composite Space Structures
EDUCATION
- BSc. in Physics - Karlsruhe Institute of Technology, Germany
MSc. in Physics - Karlsruhe Institute of Technology, Germany
PhD in Robotics, Control, and Intelligent Systems - École Polytechnique Fédérale de Lausanne, Switzerland
ABOUT
Kevin graduated from the Karlsruhe Institute of Technology in 2025, where he studied quantum mechanics in combination with deep learning. In his master's thesis, he explored the use of graph neural networks for data-efficient learning of molecular interactions.
Now, he combines his knowledge of physics and machine learning to develop physics-informed deep graph neural networks for simulation and intelligent maintenance. When not training models, he likes to play chess, swim, and dance Salsa.
02
Objectives
- Develop hybrid digital twins that combine physics-models with physics-informed graph neural networks (GNN).
- Improve predictive accuracy, reduce computational costs, and enhance the design and monitoring of composites.
03
Expected results:
- Develop physics-informed models that can be used for online monitoring of composite health.
04
Planned secondments:
- Beyond Gravity (BEG), Zurich, Switzerland, months 17-22 (6 months): collaborate on digital twin calibration.
- Politecnico di Torino (POLITO), Torino, Italy, months 25-30 (6 months): collaborate on physical constraints in GNN models.
- Cambridge, months - ( months): collaborate on data collection and ML algorithms.