Formula One is the pinnacle of Motorsport: every year each Team that participates to the competition builds and develops its own car, trying to push to the limit its performance on track. A lot of effort is dedicated to the aerodynamic development: Computational Fluid Dynamics (CFD) is one of the most used tool to develop the car. Each development is tested in wind tunnel and track to check for correlation of the expected performances. Very often Particle Image Velocimetry (PIV) measurements are captured in wind tunnel sessions to get two dimensional snapshots of the velocity field around the model. Despite it is very useful to be able to visualise the velocity filed in the key areas of the car with a certain accuracy, no conclusions can be done on the pressure field. In literature they are available many numerical methods that allow to get an approximation of the pressure given the velocity components, mainly solving the momentum equations of the Navier-Stokes equations. Unfortunately, most of the times the PIV snapshots capture highly three dimensional phenomena and it is not possible to reconstruct the pressure due to the lack of information about the velocity component normal to the PIV plane, for which it is not possible to compute the spatial derivative. The present thesis will explore the capability of Convolutional Neural Networks (CNNs), a class of Supervised Neural Networks, to learn predicting the pressure flow field from the velocity one in a two dimensional slice. Due to the huge amount of CFD simulations that every week is performed, the training and the development of the network will use two dimensional slices of the velocity and pressure field around the car from these data, trying to replicate the PIV measurement snapshots obtained in wind tunnel. For what it concerns the neural networks, they have been tested a class of architectures called Autoencoders, developing each part of the network to improve its learning capability. The approach is a standard Supervised Learning where two dimensional velocity slices are input to the network and prediction for the pressure field at the same slices is obtained.
La Formula Uno è la massima competizione motoristica: tutti gli anni ogni squadra che partecipa alla competizione costruisce e sviluppa la sua macchina, spingendo al limite le prestazioni in pista. Molte energie sono dedicate allo sviluppo aerodinamico: Computational Fluid Dynamics (CFD) è uno degli strumenti piu utilizzati per sviluppare la macchina. Ogni sviluppo è testato in galleria del vento e in pista per verificare la correlazione dei prestazioni previste. Molto spesso misurazioni Particle Image Velocimetry (PIV) sono eseguite in galleria del vento per ottenere foto bidimensionali della velocitá attorno al modello. Nonostante sia molto utile visualizzare con discreta accuratezza il campo di velocitá nelle parti piu importanti dell’auto, non è possibile fare alcuna discussione sul campo di pressione. Il letteratura sono disponibili molti metodi numerici che permettono di ottenere un’approssimazione della pressione date le componenti di velocitá, principalmente risolvendo le equazioni di conservazione del momento (equazioni di Navier-Stokes). Sfortunatamente, in molti casi le immagini PIV acquisite trattano fenomeni altamenti tridimensionali e non è possibile ricostruire la pressione a causa della mancanza di informazioni sulla componente di velocitá normale al piano della PIV, per il quale non è possibile calcolare tutte le derivate spaziali. Questa tesi vuole esplorare le capacitá delle Convolutional Neural Networks (CNNs), una classe di Supervides Neural Networks, ad imparare a predire il campo di pressione a partire dal campo di velocitá in immagini due dimensionali. A causa del grande numero di simulazioni CFD che ogni settimana sono risolte, la rete sará allenata e sviluppata usando immagini due dimensionali della velocitá e della pressione attorno alla macchina ottenuti da questi dati, cercando di replicare le immagini PIV ottenute in galleria del vento. Per quanto rigurda la rete neurale, è stata sperimentata l’applicazione di una classe di architetture chiamata Autoencoders, sviluppando ogni parte della rete al fine di migliorare le capacitá di apprendimento. L’approccio usato è uno standard Supervised Learning dove immagini due dimensionali sono inserite nella rete per predire il campo di pressione nelle stesse immagini.
Application of Machine Learning techniques for flowfield features reconstruction
PETRUCCI, LUDOVICO
2022/2023
Abstract
Formula One is the pinnacle of Motorsport: every year each Team that participates to the competition builds and develops its own car, trying to push to the limit its performance on track. A lot of effort is dedicated to the aerodynamic development: Computational Fluid Dynamics (CFD) is one of the most used tool to develop the car. Each development is tested in wind tunnel and track to check for correlation of the expected performances. Very often Particle Image Velocimetry (PIV) measurements are captured in wind tunnel sessions to get two dimensional snapshots of the velocity field around the model. Despite it is very useful to be able to visualise the velocity filed in the key areas of the car with a certain accuracy, no conclusions can be done on the pressure field. In literature they are available many numerical methods that allow to get an approximation of the pressure given the velocity components, mainly solving the momentum equations of the Navier-Stokes equations. Unfortunately, most of the times the PIV snapshots capture highly three dimensional phenomena and it is not possible to reconstruct the pressure due to the lack of information about the velocity component normal to the PIV plane, for which it is not possible to compute the spatial derivative. The present thesis will explore the capability of Convolutional Neural Networks (CNNs), a class of Supervised Neural Networks, to learn predicting the pressure flow field from the velocity one in a two dimensional slice. Due to the huge amount of CFD simulations that every week is performed, the training and the development of the network will use two dimensional slices of the velocity and pressure field around the car from these data, trying to replicate the PIV measurement snapshots obtained in wind tunnel. For what it concerns the neural networks, they have been tested a class of architectures called Autoencoders, developing each part of the network to improve its learning capability. The approach is a standard Supervised Learning where two dimensional velocity slices are input to the network and prediction for the pressure field at the same slices is obtained.| File | Dimensione | Formato | |
|---|---|---|---|
|
Master_thesis_Ludovico_Petrucci.pdf
non accessibile
Dimensione
7.23 MB
Formato
Adobe PDF
|
7.23 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/214148