Cremonesi Massimiliano

Associate Professor

Contacts

+39 02 2399 6230
Department of Civil and Environmental Engineering Politecnico di Milano

© 2024 - All rights reserved - Privacy Policy

Machine-Learning methods for fluid and fluid-structure interaction problems

Numerical simulation of fluid and fluid-structure interaction plays an essential role in modelling many physical phenomena. However, the computational cost limit the use of the numerical solvers in particular when small space and time features are required. Machine learning (ML) techniques can be used to improve approximations inside computational frameworks to enhance their predictivity capability. In this work, different ML techniques will be coupled with existing solvers to test the best solution in terms of accuracy and efficiency.

Model order reduction for free-surface fluid flows

In this work, reduced-order models will be used to enhance the computational efficiency of a free-surface fluid flow solver. In particular, the computational convenience of these methods is related to finding a proper expansion of the numerical solution over a representative reduced basis, enabling to drastically reduce the size of the original problem. Innovative solution should be proposed to solve problems in which a significant movement of the mesh makes the use of standard methodologies impossible (e.g. the propagation phase of a landslide).

Numerical simulation of phase change phenomena

Solid-liquid phase changes occur in many natural processes (landslide triggering, magma formation, iceberg evolution) and industrial applications (metal casting, material melting, 3D printing). Solidification and liquefaction phenomena involve changes in the microstructure of the material, while the consequences of these changes are of importance at the macroscale as in the case, e.g., of the inception of a landslide. The accurate and computationally effective numerical simulation of these highly nonlinear multiscale and multiphysics transient phenomena is still a challenge. In this project, a numerical approach for the simulation of macroscopic solid-liquid-solid phase change phenomena will be developed.

Numerical simulation of fluid-structure interaction problems

The goal of the master thesis is the application of a finite element approach to the simulation of fluid-structure interaction problems. Starting from an approach already developed, different improvements can be done during the master thesis work: (i) test new coupling techniques; (ii) increase the numerical efficiency of the tool; (iii) verify with new benchmarks and (iv) apply to real scale engineering problems.