Margaux Geuzaine, researcher in Structural and Stochastic Dynamics, is awarded a BAEF fellowship for a postdoctoral stay at the University of Notre Dame



Margaux Geuzaine, a researcher in Structural and Stochastic Dynamics in the Urban and Environmental Engineering Research Unit (School of Engineering) at ULiège, will continue her postdoctoral studies at the University of Notre Dame (Indiana, USA) next year thanks to a BAEF (Belgian American Educational Foundation) grant. Her project consists of developing physics-based machine learning to monitor structures exposed to natural hazards.

W

e end up not noticing them anymore, but civil engineering structures such as bridges and buildings are essential to our daily lives. But they are inevitably subject to possible overloads, and natural hazards like floods, extreme winds, or earthquakes. Unfortunately, these events are expected to become more frequent and more vigorous over the future years because of two major issues of our times: climate change and population growth. In addition, by building ever longer bridges and ever taller buildings, the probability of facing unknown or little understood phenomena is increasing as well. It is therefore crucial to be able to monitor the loading and the response of structures, in order to identify invisible damages that may have weakened them, to understand how they behave in unpredicted situations, and to eventually enhance the safety of the people who use them. 

Given how important this is for society, many researchers are working on the topic and the emphasis is now placed on using artificial intelligence (AI) to process the continuously growing amount of heterogeneous data which can be collected by means of the rapidly evolving sensors. But data-driven models do not always obey the fundamental laws of physics and are most often unable to deal with scenarios from outside the scope of their training set, even though they fit observations very well. 

In Margaux's opinion and in accordance with the current state-of-the-art, integrating physical constraints into learning algorithms would create the perfect ground for developing brand new methods to tackle structural health monitoring problems more effectively, but also to fill some gaps in the understanding of how structures are endangered by certain phenomena, and how it is possible to make them safer. During her postdoctoral studies, she is thus planning to develop such knowledge-enhanced machine learning techniques, especially dedicated to wind-loaded, wave-loaded, and cable-supported structures because she has already acquired some expertise on how they are supposed to behave under given circumstances and proposed further explanations regarding their dynamical and higher-order spectral characteristics in the context of her doctorate.

NatHaz Modeling Laboratory at the University of Notre Dame

In this perspective, Margaux has obtained an invitation to join the research group of Professor Ahsan Kareem, who is the current President of the International Association for Wind Engineering. The mission of his lab is to quantify the effects that loads associated to various natural hazards (NatHaz) have on structures and to introduce efficient strategies to prevent/handle those. Fundamental analytical and original computational methods, as well as laboratory and full-scale experiments, are used there to characterize the dynamic response due to wind, wave, and earthquakes of tall buildings, long-span bridges, offshore systems and other structures. In this regard, the NatHaz Lab focuses on the development of innovative techniques, having recourse for instance to machine learning, cloud-based computing, crowdsourcing, sensing and actuation.

Share this news