The numerical simulation of mixing using Resonant Acoustic® Mixing (RAM) technology quickly becomes computationally intensive. This issue is even more pronounced when considering 3D configurations. The significant computational time is a substantial drawback for the predictive use of the simulation code in production. The high computational demand is attributed to the simulation's consideration of the smallest spatiotemporal scales over very long integration times.
The objective of the two-year Post-Doc is to develop a model reduction strategy based on constructing small-dimensional reduced bases. The aim is to propose efficient interpolation techniques and build reduced models using Galerkin-type projection on the reduced basis. Finally, a certification step will be implemented to ensure the reliability of solutions derived from the Reduced Order Model (ROM) without resorting to a Full Order Model (FOM) calculation.
We also aim to investigate an AI-inspired models, which incorporate information from physical model. Namely we will consider the Sparse Identification of Nonlinear Dynamics (SINDy) approach to identify the nonlinear dynamical systems from data without assumptions on the form of the governing equations.
This project is supported by the ANR ASTRID and will be achieved in collaboration with the material department of ROXEL group
