Given a spacecraft that enters the planetary atmosphere at hypersonic speed, due to extreme thermodynamic effects, the surrounding gas and the material of the vehicle’s thermal shield trigger catalytic reactions. In order to model this phenomenon, accurate plasma ground experiments and high-fidelity numerical simulations are required. This thesis presents an advanced numerical model of a high-enthalpy under-expanded jet previously obtained in the von Karman Institute (VKI) Plasmatron facility, using state-of-the-art methodologies for Uncertainty Quantification (UQ). This approach involves, first, the use of the US3D Computational Fluid Dynamics (CFD) solver to produce high-fidelity solutions of the phenomenon of interest, without being able to fully characterize the flow in accordance with the experimental campaign. The use of a nested grid allows to evaluate the numerical error associated with each grid and decide the necessary refinements to train a multi-fidelity surrogate model. The second contribution of this thesis, in fact, concerns an overall UQ framework to employ an efficient and accurate multi-level surrogate model, based on Kriging theory and trained on several CFD simulations, to accelerate both the forward problem and the Bayesian inference. The former consists in the propagation of model and experimental uncertainties to mainly characterize heat flux and pressure at the stagnation point experienced by a re-entry vehicle. Since not all propagated input values provide the experimental observations, a Bayesian calibration is required to discard all those wrong values. In fact, Bayesian analysis employs experimental measurements to infer both total quantities at the nozzle inlet and catalytic efficiencies. Particular attention is given to total temperature and recombination probabilities, for which no direct experimental data are available but a prior knowledge is assumed. However, the nitrogen recombination probability could not be properly defined due to the relatively low temperatures used in the targeted applications.
Dato un veicolo spaziale che entra nell’atmosfera planetaria a velocità ipersonica, a causa di effetti termodinamici estremi, il gas circostante e il materiale dello scudo termico del veicolo innescano reazioni catalitiche. Per modellare questo fenomeno, sono necessari accurati esperimenti al plasma condotti al suolo e simulazioni numeriche ad alta fedeltà. Questa tesi presenta un modello numerico avanzato di un getto ad alta entalpia sotto-espanso precedentemente ottenuto nell’impianto Plasmatron dell’Istituto von Karman, utilizzando metodologie all’avanguardia per la quantificazione di incertezza. Questo approccio prevede, in primo luogo, l’uso del solutore CFD US3D per produrre soluzioni ad alta fedeltà del fenomeno di interesse, senza essere in grado di caratterizzare completamente il flusso in conformità con la campagna sperimentale. L’uso di una griglia nidificata permette di valutare l’errore numerico associato ad ogni griglia e decidere le raffinazioni necessarie per addestrare un modello surrogato multi-fedeltà. Il secondo contributo di questa tesi, infatti, riguarda un quadro generale di quantificazione dell’incertezza per impiegare un modello surrogato multi-livello efficiente e accurato, basato sulla teoria di Kriging e addestrato su diverse simulazioni CFD, per accelerare sia il problema diretto che l’inferenza bayesiana. Il primo consiste nella propagazione di incertezze sperimentali e di modello per caratterizzare principalmente flusso di calore e pressione al punto di ristagno sperimentate da un veicolo di rientro. Poiché non tutti quanti i valori in ingresso propagati forniscono le osservazioni sperimentali, una calibrazione bayesiana è necessaria per scartare tutti quei valori sbagliati. Infatti, l’analisi bayesiana impiega le misure sperimentali per inferire sia le quantità totali all’entrata dell’ugello che le efficienze catalitiche. Particolare attenzione viene data alla temperatura totale e alle probabilità di ricombinazione, per le quali non sono disponibili dati sperimentali diretti ma si ipotizza una conoscenza preliminare. La probabilità di ricombinazione dell’azoto, tuttavia, non ha potuto essere definita correttamente a causa delle temperature relativamente basse usate nelle applicazioni studiate.
Simulations and stochastic inference of supersonic plasma flow experiments
PIRO, VITTORIO
2022/2023
Abstract
Given a spacecraft that enters the planetary atmosphere at hypersonic speed, due to extreme thermodynamic effects, the surrounding gas and the material of the vehicle’s thermal shield trigger catalytic reactions. In order to model this phenomenon, accurate plasma ground experiments and high-fidelity numerical simulations are required. This thesis presents an advanced numerical model of a high-enthalpy under-expanded jet previously obtained in the von Karman Institute (VKI) Plasmatron facility, using state-of-the-art methodologies for Uncertainty Quantification (UQ). This approach involves, first, the use of the US3D Computational Fluid Dynamics (CFD) solver to produce high-fidelity solutions of the phenomenon of interest, without being able to fully characterize the flow in accordance with the experimental campaign. The use of a nested grid allows to evaluate the numerical error associated with each grid and decide the necessary refinements to train a multi-fidelity surrogate model. The second contribution of this thesis, in fact, concerns an overall UQ framework to employ an efficient and accurate multi-level surrogate model, based on Kriging theory and trained on several CFD simulations, to accelerate both the forward problem and the Bayesian inference. The former consists in the propagation of model and experimental uncertainties to mainly characterize heat flux and pressure at the stagnation point experienced by a re-entry vehicle. Since not all propagated input values provide the experimental observations, a Bayesian calibration is required to discard all those wrong values. In fact, Bayesian analysis employs experimental measurements to infer both total quantities at the nozzle inlet and catalytic efficiencies. Particular attention is given to total temperature and recombination probabilities, for which no direct experimental data are available but a prior knowledge is assumed. However, the nitrogen recombination probability could not be properly defined due to the relatively low temperatures used in the targeted applications.File | Dimensione | Formato | |
---|---|---|---|
2024_04_Piro_Executive Summary_02.pdf
accessibile in internet solo dagli utenti autorizzati
Dimensione
1.68 MB
Formato
Adobe PDF
|
1.68 MB | Adobe PDF | Visualizza/Apri |
2024_04_Piro_Tesi_01.pdf
accessibile in internet solo dagli utenti autorizzati
Dimensione
3.94 MB
Formato
Adobe PDF
|
3.94 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/10589/219372