Paola Cinnella, Sorbonne University, Paris, France
Title: Physics-Aware Machine Learning and Multi-Fidelity Surrogates for Low-Carbon Energy and Transportation
Abstract:
Accurate numerical models are essential for improving the efficiency of energy and transportation systems, but high-fidelity simulations of complex fluid systems are overly expensive for large-scale design and optimization, and they heavily rely on carbon-intensive high-performance computing. This talk will discuss how physics-aware machine learning and multi-fidelity surrogate modeling can accelerate the development of low-carbon technologies by combining scarce high-fidelity simulations with cheaper lower-fidelity models and experimental data. Examples from aerodynamics and energy systems will illustrate the use of hybrid ML-augmented CFD approaches, where machine learning enhances selected components of the numerical pipeline, such as turbulence closures, while preserving the underlying physical modeling. The presentation will also discuss neural surrogate models, active learning strategies, and emerging directions toward scientific foundation models for engineering physics. More broadly, the talk will highlight how machine learning can help distill and transfer high-fidelity scientific knowledge to accelerate innovation in hard-to-decarbonize industries.
Biography:
Paola is a Full Professor of Fluid Mechanics at Sorbonne University (Institut Jean Le Rond D’Alembert) in Paris, France. She received her Ph.D. in Fluid Mechanics from the École Nationale Supérieure d’Arts et Métiers (ENSAM) in Paris after completing a Master’s Degree in Fluid Mechanics at École Nationale d’Ingénieurs Arts et Métiers ParisTech (ENSAM) and a Mechanical Engineering degree from Politecnico di Bari, Italy.
Her research focuses on Computational Fluid Dynamics (CFD) for the analysis and design of compressible and turbulent flows, with applications in aerospace and energy systems. She has made seminal contributions to high-order numerical methods, uncertainty quantification, robust optimisation, and increasingly to data-driven and machine learning approaches for turbulence modelling and flow prediction.
Currently, she coordinates the European Research Community on Flow, Turbulence and Combustion (ERCOFTAC) Special Interest Group 54 “Machine Learning in Fluid Dynamics”. Her ongoing projects include SciFiturbo https://scifiturbo.eu/, a Horizon-Europe collaborative project lead by DLR, aiming to integrate high-fidelity simulations and data-driven methods to design the next generation of ultra-efficient aeronautic propellers, the chair “Machine Learning for Fluids and Climate” funded by the PostGenAI@Paris AI Cluster, focusing on foundation models for fluid flows, and the Marie Curie doctoral network REALISE https://realise-dn.eu/, aiming at combining AI and igneous petrology data and models. She also collaborates with Safran Tech on machine-learning-assisted turbulence model development.
She serves as Editor-in-Chief of Computers & Fluids, Associate Editor of the International Journal of Heat and Fluid Flow, and Editorial Board Member of Flow, Turbulence and Combustion. She is also Scientific Secretary of the International Conference on Computational Fluid Dynamics (ICCFD).