Exploring of the role of knowledge and biases for data-driven methods in the hybrid modelling paradigm

Beatriz Moya Garcia

Chaire Professeur Junior @ENSAM/PIMM

Amphi A

ABSTRACT

In spite of the growing interest in data-driven methods to learn complex correlations, there is still scope for exploration on how to incorporate the knowledge collected of the problem into machine learning algorithms. This talk will showcase methods and practical applications of knowledge-embedded machine learning for dynamical simulation and model order reduction, with a special focus on hybrid twins, by exploiting geometric and physical (PDEs, thermodynamics…) information that will lead to more interpretable (grey box) and explainable (provide useful information) results that lead to efficient decision-making. To finish, new plans and perspectives to be done within the project ITTAI, and ENSAM in general, will be introduced hoping to trigger new collaborations that could be of interest for the members of the laboratory. 

BIOGRAPHY 

After studying a BSc in Mechanical Engineering and an MSc focused on computational sciences, I pursued a PhD in Mechanical Engineering at the University of Zaragoza under the supervision of Profs. Elias Cueto and David Gonzalez. My research focused on developing hybrid and cognitive digital twins (DTs) for real-world physics. To achieve this, I integrated data-driven methods, model order reduction, thermodynamics-informed machine learning, and computer vision to address perception, prediction, and correction within the DT framework.

During my PhD, I also collaborated with Prof. Jan Hesthaven on applying deep learning to model order reduction for unstructured domains, with applications in the study and prediction of bifurcations in fluid dynamics among others. Following my PhD, I was awarded a NextGen postdoctoral scholarship in Spain (Margarita Salas) to explore physics-informed generative design. I then moved to Singapore to work on the DESCARTES project for CNRS@CREATE, focusing on machine learning for structural health monitoring under the guidance of Profs. Paco Chinesta and Eleni Chatzi (ETH Zurich).

Recently I started a new position as Chaire Professeur Junior at ENSAM where, among others, I contribute to the ITTAI project focused on developing smart territories, and where I hope to put into service my expertise in physics-informed machine learning to benefit the department.