03-07-2025 - Lecture Hall E

Mini-Symposium: Social Talk

Adaptivity is key!

Jovana Jovanova

Jovana and her research group will present the motivation for their work, their values, and the role of diversity in shaping the group dynamics.
They will share their thoughts on the concept of Work-Life Integrated Design: how their personal journey was shaped by the career choice and how their lives were enriched because of their work experiences.
Personal stories on how they got to TU Delft, alongside what they are currently working on, will show the importance of adaptation is life and in research. The challenges of adapting to a new academic environment can lead to the benefit of adopting different behavior for personal development.
Jovana will also share how her research was inspired by different international experiences, and how they all became puzzle pieces in her own professional trajectory. Her vision for the future centers around Adaptive Structures and her group works to unlock their potential in the Maritime and Transport Tecnology.

26-06-2025 - Lecture Hall E

PostDoc presentation

Inverse Methods: Determine Material Properties with PINNs

Bram van der Heijden

This presentation introduces Physics-Informed Neural Networks (PINNs) as a modern approach to solving inverse problems in solid mechanics. PINNs teach physical laws, such as PDEs, to NNs, enabling machine learning with limited data. While they offer significant speed and flexibility compared to traditional methods, they also come with notable challenges, especially regarding accuracy, stability, and error control.
These benefits and drawbacks are illustrated through a practical example: identifying spatially varying material properties from displacement and load measurements. The talk will explore how these limitations can be addressed by incorporating classical mechanics principles, such as Airy stress potentials and constrained optimization, into the PINN framework. These refinements not only improve solution quality and robustness but also enable error estimation and model validation.
Beyond demonstrating technical improvements, the presentation aims to offer a clear and critical introduction to PINNs. I hope that you will gain some practical insights, a clearer understanding of the method’s capabilities and limitations, and a solid starting point for exploring machine learning in mechanics.