This thesis explores the challenges related to the addition of Large Eddy Simulation (LES) capabilities into an innovative Direct Numerical Simulation (DNS) code that makes use of a novel approach to the Immersed Boundary Method (IBM). The scope is transversal and encompasses the theoretical background, the implementation aspects, the computing performance, the testing and validation in multiple scenarios, and the interpretation of the results. The resulting framework aims to be a fast and efficient tool that allows high-fidelity simulation to be conducted on complicated geometries, while keeping the cost as low as possible. Care is taken to preserve good computing performance characteristics of the solver and to account for possible interactions between the LES modeling and the IBM. A variety of LES models are tested and made available for choosing. Validation is done in a plane channel, benchmarking the performance of different models in progressively more challenging cases. Measurable improvements are found in the prediction of friction, and second moment predictions are in agreement with literature results. Special attention is given to how to properly compare against DNS data, and we confirmed for ourselves the existence of an inseparable link between LES and the underlying discrete numerics. The resulting framework is then tested in a geometrically complex biomechanics case, that emphasizes the importance of the IBM: the flow in the upper human airways. This is a case of paramount importance in otorhinolaryngology, as it has the potential to change how we treat breathing difficulties. The extremely complex anatomy also highlighted strengths and weaknesses of LES modeling. It was shown that it is possible to obtain realistic flow morphology even at extremely reduced computational cost, at the benefit of possible clinical settings deployment. The outcome of this work constitutes an invaluable experience in view of the implementation of a production version of the solver, which will inherit most of the choices discussed in this work.
Questa tesi esplora le sfide relative all’inserimento della capacità di condurre Simulazioni dei Grandi Vortici (LES) all’intorno di un codice innovativo per Simulazione Numerica Diretta (DNS) che utilizza un nuovo approccio al Metodo dei Contorni Immersi (IBM). Lo scopo del lavoro è trasversale e comprende considerazioni teoriche, aspetti implementativi, prestazioni computazionali, prova e validazione in scenari multipli, e l’interpretazione dei risultati. L’architettura risultante mira a diventare uno strumento rapido ed efficiente che consente di condurre simulazioni ad alta fedeltà su geometrie complicate, contenendo il più possibile i costi. Particolare attenzione è stata prestata per preservare le buone caratteristiche elaborative del solutore e a tenere conto delle possibili interazioni tra LES e IBM. Una varietà di modelli LES è stata testata e resa disponibile per uso futuro. La validazione viene eseguita in un canale piano, confrontando le prestazioni di diversi modelli in casi progressivamente più impegnativi. Si riscontrano miglioramenti misurabili nella previsione dell’attrito, e la previsione dei momenti secondi rientra nei valori di letteratura. Particolare attenzione è stata dedicata a come confrontare correttamente con dati DNS ed è stata confermata l’esistenza di un collegamento inscindibile tra LES e i metodi numerici sottostanti. L’architettura così definita viene quindi testata in un complesso caso biomeccanico: il flusso nelle vie aeree superiori. Questo è un caso di fondamentale importanza in otorinolaringoiatria, in quanto potrebbe potenzialmente cambiare il modo in cui trattiamo le difficoltà respiratorie. L’anatomia estremamente complessa ha anche evidenziato i punti forti e deboli della modellazione LES. È stato dimostrato che è possibile ottenere una morfologia del flusso realistica anche a costi computazionali estremamente contenuti, a vantaggio di un possibile impiego in contesto clinico. Il risultato di questo lavoro costituisce un’esperienza inestimabile in vista dell’implementazione di una versione di produzione del solutore, che erediterà la maggior parte delle scelte discusse in questo lavoro.
DNS and LES with the immersed-boundary technique
Esposito, Francesco
2023/2024
Abstract
This thesis explores the challenges related to the addition of Large Eddy Simulation (LES) capabilities into an innovative Direct Numerical Simulation (DNS) code that makes use of a novel approach to the Immersed Boundary Method (IBM). The scope is transversal and encompasses the theoretical background, the implementation aspects, the computing performance, the testing and validation in multiple scenarios, and the interpretation of the results. The resulting framework aims to be a fast and efficient tool that allows high-fidelity simulation to be conducted on complicated geometries, while keeping the cost as low as possible. Care is taken to preserve good computing performance characteristics of the solver and to account for possible interactions between the LES modeling and the IBM. A variety of LES models are tested and made available for choosing. Validation is done in a plane channel, benchmarking the performance of different models in progressively more challenging cases. Measurable improvements are found in the prediction of friction, and second moment predictions are in agreement with literature results. Special attention is given to how to properly compare against DNS data, and we confirmed for ourselves the existence of an inseparable link between LES and the underlying discrete numerics. The resulting framework is then tested in a geometrically complex biomechanics case, that emphasizes the importance of the IBM: the flow in the upper human airways. This is a case of paramount importance in otorhinolaryngology, as it has the potential to change how we treat breathing difficulties. The extremely complex anatomy also highlighted strengths and weaknesses of LES modeling. It was shown that it is possible to obtain realistic flow morphology even at extremely reduced computational cost, at the benefit of possible clinical settings deployment. The outcome of this work constitutes an invaluable experience in view of the implementation of a production version of the solver, which will inherit most of the choices discussed in this work.File | Dimensione | Formato | |
---|---|---|---|
2024_12_Esposito.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: testo tesi
Dimensione
10.38 MB
Formato
Adobe PDF
|
10.38 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/10589/230298