Implicit turbulence modeling for large-eddy simulation [Elektronische Ressource] / Stefan Hickel

Informations

• Publié par: technische_universitat_munchen
• Publié le: 01 janvier 2008
• Nombre de lectures: 32
• Langue: English
• Poids de l'ouvrage: 12 Mo

Extrait

Technische Universit¨at Mu¨nchen
Lehrstuhl fu¨r Aerodynamik
Implicit Turbulence Modeling for
Large-Eddy Simulation
Stefan Hickel
Vollst¨andiger Abdruck der von der Fakult¨at fu¨r Maschinenwesen der Technischen
Universit¨at Mu¨nchen zur Erlangung des akademischen Grades eines
Doktor-Ingenieurs
genehmigten Dissertation.
Vorsitzender: Univ.-Prof. Dr.-Ing. Thomas Sattelmayer
Pru¨fer der Dissertation: 1. Univ.-Prof. Dr.-Ing. habil. Nikolaus A. Adams
2. Univ.-Prof. Dr.-Ing. Leonhard Kleiser,
Eidgen¨ossische Technische Hochschule Zu¨rich
Die Dissertation wurde am 14.01.2008 bei der Technischen Universit¨at Mu¨nchen
eingereicht und durch die Fakult¨at fu¨r Maschinenwesen am 16.06.2008 angenommen.Stefan Hickel
Gerichtstrasse 41a
15806 Zossen
Germany
[e-mail]
c Stefan Hickel, 2007
All rights reserved. No part of this publication may be reproduced, modified, re-written, or
distributed in any form or by any means, without the prior written permission
of the author.
Released December 20, 2007
ATypesetting LT XEABSTRACT
The subgrid-scale (SGS)model in alarge-eddy simulation (LES) generally operates ona
range of scales that is marginally resolved by discretization schemes. Consequently, the
discretization scheme’s truncation error and the subgrid-scale model are linked, which
raises the question of how accurate the computational results are. The link between the
SGSmodelandtruncationerrorcanbebeneficiallyexploitedbydevelopingdiscretization
methods for subgrid-scale modeling, or vice versa. Approaches where the SGS model
and the numerical discretization scheme are fully merged are called implicit LES (ILES)
methods.
In order to improve on modeling uncertainties, a systematic framework is proposed
for design, analysis, and optimization of nonlinear discretization schemes for implicit
LES. The resulting adaptive local deconvolution method (ALDM) for implicit LES is a
finite volume method based on a nonlinear deconvolution operator and a numerical flux
function. Free parameters inherent to the discretization allow to control the truncation
error. They are calibrated in such a way that the truncation error acts as a physically
motivated SGS model. An automatic optimization based on an evolutionary algorithm
is employed toobtain a set of parameters that results in an optimum match between the
spectral numerical viscosity and theoretical predictions of the spectral eddy viscosity for
isotropic turbulence. The method is formulated for LES of turbulent flows governed by
the incompressible Navier-Stokes equations and for passive-scalar mixing.
ALDMhasshownthepotentialforprovidingareliable,accurate,andefficientmethodfor
LES. Various applications, such as three-dimensional homogeneous isotropic turbulence,
transitional and turbulent plane channel flow, and turbulent boundary-layer separation,
demonstrate the good performance of the implicit model. Computational results agree
well with theory and experimental data and show that the implicit SGS model performs
at least as well as established explicit models, for most considered applications the
performance is even better. This is possible because physical reasoning is incorporated
into the design ofthe discretization scheme and discretization effects are fully taken into
account within the SGS model formulation.CONTENTS
NOMENCLATURE ix
1 INTRODUCTION 1
1.1 Motivation 1
1.2 Contribution of this Work 5
1.3 Mathematical Formulation 7
1.4 Outline 10
2 DISCRETIZATION-SCHEMEDESIGN 13
2.1 The incompressible Navier-Stokes Equations 13
2.2 Finite-Volume Filter and Differentiation Operator 15
2.3 Solution-Adaptive Local Deconvolution 17
2.4 Numerical Flux Function 22
2.5 Summary 25
3 IMPLICITSGS MODELING 27
3.1 Modified-Differential-Equation Analysis in Real Space 27
3.2 Modified-Differential-Equation Analysis in Spectral Space 30
3.2.1 Mathematical Formulation 31
3.2.2 Numerical Evaluation 33
3.3 Calibration of Model Coefficients 35
3.3.1 Cost Function 35
3.3.2 Evolutionary Optimization 36
3.3.3 Optimized Eddy-Viscosity Model 39
3.4 Summary 40iv Contents
4 VALIDATIONFOR ISOTROPICTURBULENCE 41
4.1 Computational Setup 41
4.2 Decaying Homogeneous Isotropic Turbulence 42
4.3 Forced Homogeneous Isotropic Turbulence 44
4.4 Comte-Bellot - Corrsin Experiment 46
4.5 Transition of the Three-Dimensional Taylor–Green Vortex 49
4.6 Summary 54
5 ADAPTATIONFOR WALL-BOUNDEDTURBULENCE 55
5.1 Application to Turbulent Channel Flow 55
5.1.1 Computational Method 55
5.1.2 Grid and Boundary Conditions 56
5.1.3 Results 57
5.2 Near-Wall Modeling 59
5.2.1 Subgrid Dissipation 59
5.2.2 Near-Wall Effects 62
5.2.3 Wall Correction 63
5.3 Validation 67
5.3.1 Effect of Reynolds Number 67
5.3.2 Grid-Resolution Study for Channel Flow 70
5.3.3 Zero-Pressure-Gradient Turbulent Boundary-Layer Flow 73
5.4 Summary 76
6 EXTENSIONTO PASSIVE-SCALARMIXING 77
6.1 The Passive-Scalar Transport Equation 77
6.2 Discretization-Scheme Design 80
6.3 Implicit SGS Modeling 82
6.4 Validation for Forced Isotropic Turbulence 84
6.5 Application to Turbulent Channel Flow 85
6.5.1 Test case and computational setup 85
6.5.2 Velocity Field 86
6.5.3 Scalar Statistics 87
6.5.4 Grid-Resolution Study 94
6.6 Summary 96Contents v
7 LES OF TURBULENTBOUNDARY-LAYERSEPARATION 97
7.1 Introduction 97
7.2 Computational Setup 99
7.2.1 Numerical Method 99
7.2.2 Computational Grid 100
7.2.3 Statistical Analysis 100
7.2.4 Boundary Conditions 100
7.3 Results and Discussion 103
7.3.1 General Overview 103
7.3.2 Separation Dynamics 106
7.3.3 Comparison of Numerical and Experimental Results 108
7.3.4 Reynolds Stress and Anisotropy Tensor 116
7.3.5 Skewness and Kurtosis 122
7.3.6 Scaling Laws for the Mean Velocity Profile 125
7.4 Summary 129
8 CONCLUSION 131
A BURGERS TURBULENCE 137
A.1 Burgers Equation 137
A.1.1 Discretization design 137
A.1.2 Modified-Differential Equation Analysis 138
A.2 Adaptation for Given Explicit SGS Model 139
A.2.1 Results for Forced Burgers Turbulence 140
A.2.2 Results for Decaying Burgers Turbulence 141
A.2.3 Effect of Time Integration 142
A.3 SGS Modeling by Evolutionary Optimization 142
A.3.1 Results for Forced Burgers Turbulence 147
A.3.2 Results for Decaying Burgers Turbulence 148
A.4 Summary 150
B SIMPLIFIEDADAPTIVELOCAL DECONVOLUTION(SALD) METHOD 151
B.1 Simplified Algorithm 151
B.2 Numerical Results 152vi Contents
B.2.1 Homogeneous Isotropic Turbulence 152
B.2.2 Turbulent Channel Flow 153
B.3 Efficient Implementation 153
B.3.1 Weight Functionals 153
B.3.2 Vectorization 154
B.3.3 Parallelization 156
B.3.4 Performance 156
B.4 Summary 157
C NEAR-WALLSCALINGOF ALDMWEIGHT FUNCTIONS 159
D SUPPLEMENTARYAPG BOUNDARY-LAYERDATA 167
D.1 Instantaneous Wall-Shear Stress 167
D.2 Momentum Balance 171
D.3 Budgets of Turbulence Energy 176
D.4 Passive Scalar Mixing 179
D.5 Tabulated Data 184LIST OF TABLES
2.1 Interpolation directions for 3-D reconstruction. 22
3.1 Result obtained by evolutionary optimization for the discretization parame-
ters of ALDM. 37
3.2 Parameters of the evolutionary optimization algorithm. 38
3.3 Mean values and standard deviation of cost function and parameter values
demonstrate the convergence of the evolutionary optimization algorithm. 38
5.1 Grid for LES of turbulent channel flow at Re = 395. 57τ
5.2 Grid for LES of turbulent channel flow at Re = 180 and at Re = 950,τ τ
respectively 68
5.3 Grids for LES of turbulent channel flow at Re = 590. The cell dimensionsτ
are computed from the nominal Re . 71τ
5.4 Characteristic parameters of the temporal boundary-layer simulations. 76
6.1 Optimized discretization parameters for LES of passive-scalar mixing. 84
7.1 Parameters for the free-stream pressure boundary condition. 102
7.2 Separation state near the wall for present LES in Simpson’s terminology. 106
A.1 Resultforthediscretizationparametersγ tomatchtheexplicitSmagorin-k,r
sky model. 140
+1/2
A.2 Result obtained by evolutionary optimization for the parameters γ , fork,r
the TV form and the WENO form of the smoothness measure β . 146k,r
D.2 Pressure gradient and pressure-gradient parameters. 184
D.1 Grid parameters and averaging time in outer and inner time units. 185
D.3 Boundary-layer thickness and shape parameters. 186
D.4 Reynolds numbers, friction coefficient, and reverse-flow parameter. 187