Luigi Donnarumma
Data-driven turbulence closure models for RANS and URANS.
Rel. Maria Strazzullo, George Em Karniadakis. Politecnico di Torino, Corso di laurea magistrale in Ingegneria Matematica, 2026
|
Preview |
PDF (Tesi_di_laurea)
- Tesi
Licenza: Creative Commons Attribution Non-commercial No Derivatives. Download (72MB) | Preview |
Abstract
Reynolds-averaged Navier–Stokes (RANS) simulations are still the workhorse of industrial computational fluid dynamics, and their accuracy is limited almost entirely by the turbulence closure. This thesis develops and assesses data-driven closure models for the Reynolds stress tensor in the flow around a two-dimensional circular cylinder, in both the steady and the unsteady RANS (URANS) regime. Reference data are provided by high-fidelity NekRS simulations at four Reynolds numbers between 3900 and 140000. The three lower Reynolds numbers are used for training and the highest one is held out, so that generalization is always assessed out of distribution. Two design axes are explored.
The first axis is the amount of physics embedded in the architecture
Relatori
Anno Accademico
Tipo di pubblicazione
Numero di pagine
Corso di laurea
Classe di laurea
Aziende collaboratrici
URI
![]() |
Modifica (riservato agli operatori) |
