Polynomial adaptivity in a discontinuous Galerkin framework is presented as a way to enhance the efficiency of Large Eddy Simulation (LES) of compressible turbulent flows. The LES simulations are still computationally expensive, and since for non trivial geometries they rely on implicit filtering of the equations through the grid, the choice of the grid resolution plays an important role both in the resolution and filtering of the equations. However in most cases the choice of resolution is still subject of many assumptions. With the aim of increasing the automation in the resolution choice adaptive resolution techniques were introduced in the LES context. Discontinuous Galerkin discretization was employed for its high order accuracy combined with flexibility and parallel efficiency, and polynomial adaptivity was chosen for its simplicity in such context. Since in LES spatial resolution changes affect both the filtering and equations resolution is particularly complex to employ error estimates as refinement indicator as done in other context, the focus was set on physically based indicators that could give insight on the turbulence physics. In particular an indicator based on the structure function was developed and employed successfully. First static adaptivity was validated and employed on different test cases, leading to accurate results with up to 60% saving in computational effort with respect to the uniform polynomial resolution. Subsequently dynamic adaptivity was employed to efficiently represent transient phenomena, namely the advection of vortexes, both in a uniform flow and interacting with obstacles in turbulent flows. A particular attention was devoted to the direct computation of aeroacoustics in turbulent flows, since it is one of the main fields of application of high fidelity turbulence simulations such as LES. The best practices to employ high order discontinuous Galerkin LES for acoustics computations were investigated, and the effects of dynamic adaptivity on acoustics computations were briefly investigated and discussed.
L'adattività polinomiale per un metodo Galerkin discontinuo è presentata come soluzione per migliorare l'efficienza della Large Eddy Simulation (LES) di correnti turbolente comprimibili. Le simulazioni LES sono ancora computazionalmente costose, e dato che per geometrie non semplici utilizzano il filtraggio implicito delle equazioni attraverso la griglia, la scelta della risoluzione della griglia gioca un ruolo essenziale sia nella risoluzione che nel filtraggio delle equazioni. Tuttavia nella maggior parte dei casi la scelta della risoluzione è tutt'ora soggetta a molte supposizioni. Con lo scopo di aumentare l'automazione nella scelta della risoluzione tecniche di adattività spaziale sono state introdotte nel contesto della LES. La discretizzazione con un metodo di Galerkin discontinuo è stata utilizzata per le sue caratteristiche di accuratezza di alto ordine unita a flessibilità e efficienza nel calcolo parallelo, e l'adattività polinomiale è stata scelta per la sua semplicità in tale contesto. Dato che nella LES modifiche della risoluzione spaziale influenzano sia il filtraggio che la risoluzione delle equazioni stesse è particolarmente complesso utilizzare stime dell'errore come indicatore di raffinamento come fatto in altri contesti, l'attenzione è stata posta su indicatori basati su quantità fisiche legate alla turbolenza. In particolare un indicatore basato sulla funzione di struttura è stato sviluppato e impiegato con successo. Per prima cosa l'adattività statica è stata validata ed utilizzata su diversi casi test, producendo risultati accurati con un risparmio fino al 60% del costo computazionale rispetto alla risoluzione polinomiale uniforme. In seguito l'adattività dinamica è stata utilizzata per rappresentare efficacemente fenomeni transitori, in particolare l'advezione di vortici, sia in correnti uniformi che durante l'interazione con ostacoli in correnti turbolente. Un'attenzione particolare è stata riservata al calcolo diretto dell'aeroacustica in correnti turbolente, dato che è uno dei principali campi di applicazione di simulazioni turbolente ad alta fedeltà come la LES. Le migliori metodologie per l'impiego di LES basata su Galerkin discontinuo ad alto ordine sono state analizzate, e gli effetti dell'adattività dinamica sul calcolo dell'aeroacustica sono stati brevemente investigati e discussi.
Polynomial adaptivity for Large Eddy Simulation of compressible turbulent flows
TUGNOLI, MATTEO
Abstract
Polynomial adaptivity in a discontinuous Galerkin framework is presented as a way to enhance the efficiency of Large Eddy Simulation (LES) of compressible turbulent flows. The LES simulations are still computationally expensive, and since for non trivial geometries they rely on implicit filtering of the equations through the grid, the choice of the grid resolution plays an important role both in the resolution and filtering of the equations. However in most cases the choice of resolution is still subject of many assumptions. With the aim of increasing the automation in the resolution choice adaptive resolution techniques were introduced in the LES context. Discontinuous Galerkin discretization was employed for its high order accuracy combined with flexibility and parallel efficiency, and polynomial adaptivity was chosen for its simplicity in such context. Since in LES spatial resolution changes affect both the filtering and equations resolution is particularly complex to employ error estimates as refinement indicator as done in other context, the focus was set on physically based indicators that could give insight on the turbulence physics. In particular an indicator based on the structure function was developed and employed successfully. First static adaptivity was validated and employed on different test cases, leading to accurate results with up to 60% saving in computational effort with respect to the uniform polynomial resolution. Subsequently dynamic adaptivity was employed to efficiently represent transient phenomena, namely the advection of vortexes, both in a uniform flow and interacting with obstacles in turbulent flows. A particular attention was devoted to the direct computation of aeroacoustics in turbulent flows, since it is one of the main fields of application of high fidelity turbulence simulations such as LES. The best practices to employ high order discontinuous Galerkin LES for acoustics computations were investigated, and the effects of dynamic adaptivity on acoustics computations were briefly investigated and discussed.File | Dimensione | Formato | |
---|---|---|---|
2018_01_PhD_Tugnoli.pdf
accessibile in internet per tutti
Descrizione: Thesis
Dimensione
13.13 MB
Formato
Adobe PDF
|
13.13 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/10589/137304