Alessandro Nigra
Application of machine learning methods for aerodynamic design.
Rel. Gioacchino Cafiero. Politecnico di Torino, Corso di laurea magistrale in Ingegneria Aerospaziale, 2026
|
|
PDF (Tesi_di_laurea)
- Tesi
Accesso limitato a: Solo utenti staff fino al 29 Settembre 2029 (data di embargo). Licenza: Creative Commons Attribution Non-commercial No Derivatives. Download (5MB) |
Abstract
This thesis addresses the prediction of transonic flow fields around aero-engine nacelles over a parametric geometric design space, by means of coordinate-based Implicit Neural Representations. An INR framework previously developed for two-dimensional axisymmetric nacelles has been extended along four directions: the refinement of the near-wall discretisation of the reference data, the inclusion of the freestream Mach number in the conditioning vector, the extension to three-dimensional non-axisymmetric configurations, and the prediction of the surface static pressure in place of the volumetric Mach number. Each direction has been instantiated in a case study trained on Reynolds-Averaged Navier–Stokes datasets of parametrised nacelle geometries. The network predicts the value of the target field at a given point from its spatial coordinates, taking as additional inputs the four design variables that define the nacelle geometry — twelve for the three-dimensional configurations — and, where applicable, the freestream Mach number.
Accuracy has been assessed on held-out test sets through field-level, shock-based and surface-based metrics
Relatori
Anno Accademico
Tipo di pubblicazione
Numero di pagine
Corso di laurea
Classe di laurea
Ente in cotutela
Aziende collaboratrici
URI
![]() |
Modifica (riservato agli operatori) |
