The objective of this project is to develop novel methodologies based on physics-informed graph neural networks (PI-GNNs) to understand and model the impact of operational loads on system degradation at the compenent level in complex engineering systems, with a particular focus on wind turbines.
The research will focus on explicitly integrating physical laws, load dynamics, and degradation mechanisms into graph-based models, enabling a principled understanding of how operating conditions drive the evolution of system health over time. Particular emphasis will be placed on spatiotemporal modeling of interacting subsystems, where degradation emerges from coupled physical processes across components.
The project will explore how graph-based representations can capture:
- the propagation of loads and stresses across interconnected components,
- the accumulation of fatigue and damage under variable loading conditions, and
- the interaction between structural dynamics and degradation processes.
A central aspect of the research is the incorporation of physics-based inductive biases into learning architectures. This will enable the development of models that are physically consistent, interpretable, and robust under varying operating conditions, going beyond purely data-driven approaches.
Applications will include complex industrial and energy systems, with a particular focus on wind turbines, where load conditions directly influence the degradation of critical components such as blades, gearboxes, and bearings. The developed methods will contribute to improving lifetime modeling, reliability assessment, and physics-informed predictive maintenance.
This PhD position is part of an ERC Consolidator Grant, supporting cutting-edge research on physics-informed AI, intelligent maintenance, and the modeling of degradation processes in complex systems.
