Speaker: Prof. Luca Magri, Politecnico di Torino and Imperial College (GB)

Hosted at: SISSA, International School of Advanced Studies, Trieste, Italy, room 004.

Abstract:

Fluid mechanics modelling is based both on physical principles, such as conservation laws, which enable extrapolation beyond observed data, and on empirical approaches (AI), which excel at identifying statistical correlations within multiscale and nonlinear datasets. This seminar will show how these complementary paradigms can be integrated for the adaptive modelling and optimization of nonlinear, unsteady, and uncertain PDEs, with emphasis on Navier-Stokes. The focus is on computational strategies for noise filtering, optimal design, and turbulence learning, with methods that leverage both global (e.g., autoencoders, Riemannian geodesic computations, Riemann manifolds) and local (e.g., quantized, cluster-based) representations of turbulent flow. Case studies include flows relevant to aerospace propulsion, e.g., thermoacoustic phenomena, and turbulence. The approaches presented bridge first-principles-intrusive predictions and AI-non-intrusive methodologies for nonlinear reduced-order modelling.

Zoom link:
https://sissa-it.zoom.us/j/84116607869?pwd=kSQzwxkM20lSBdmicRaX3A1WkMGUbQ.1