Speaker: Prof. Shahab Mirjalili, KTH Royal Institute of Technology (SE)

Hosted at: SISSA, International School of Advanced Studies, Trieste, Italy, room 133.

Abstract:

Interfacial multiphase flows are central to a wide range of natural and industrial applications. Accurately capturing interface dynamics while maintaining robustness, conservation, and scalability remains a major challenge for predictive simulations. Diffuse interface methods represent interfaces in a smooth manner, providing an attractive alternative to sharp interface approaches owing to their ease of implementation, parallel efficiency, regularity, and favorable cost-to-accuracy tradeoff.
In this talk, I will present recent developments in physically consistent modeling of interfacial multiphase flows, spanning both high-fidelity numerical methods and machine learning approaches. The first part of the talk focuses on diffuse interface formulations based on the conservative Allen–Cahn phase field model. I will discuss how boundedness of the phase field can be ensured, and how the phase field equation must be consistently coupled to momentum transport, resulting in a two-phase flow model that preserves kinetic energy at the discrete level in the absence of viscous and capillary effects. Next, consistent models for scalar transport (heat, species, and surfactants) will be presented, followed by an extension to general N-phase systems.
The second part of the talk explores how machine learning can be integrated into multiphase flow modeling in a physically informed manner. I will present recent work on convolutional autoencoders for reduced-order representations of interfacial flows, as well as data-driven closure models for unresolved interfacial dynamics in large-scale simulations.

Zoom link:
https://sissa-it.zoom.us/j/84847359185?pwd=XIbK32rEkjMFSaopGjlDfana5QOZc5.1