With the increased application of large eddy simulations and hybrid Reynolds-averaged Navier-Stokes techniques, the generation of realistic turbulence at inflow boundaries is crucial for the accuracy of numerical results. In this dissertation research, two novel turbulence inflow generation methods are derived and validated. The first method, the Triple Hill's Vortex Synthetic Eddy Method, is a new type of synthetic eddy method, where the fundamental eddy is constructed through a superposition of three orthogonal Hill's vortices. The amplitudes of the three vortices that form the fundamental eddy are calculated from known Reynolds stress profiles through a transformation from the physical reference frame to the principal-axis reference frame. In this way, divergenceree anisotropic turbulent velocity fields are obtained that can reproduce a given Reynolds stress tensor. The model was tested on isotropic turbulence decay, turbulent channel flow, and a spatially developing turbulent mixing layer. The Triple Hill's Vortex Synthetic Eddy Method exhibited a quicker recovery of the desired turbulent flow conditions when compared with other current synthetic turbulence methods. The second method is the Control Forced Concurrent Precursor Method which combines an existing concurrent precursor method and a mean flow forcing method with a new extension of the controlled forcing method. Turbulent inflow boundary conditions are imposed through a region of body forces added as source terms to the momentum equations of the main simulation which transfer flow variables from the precursor simulation. Controlled forcing planes imposed in the precursor simulation, allow for specific Reynolds stress tensors and mean velocities to be imposed. A unique feature of the approach is that the proposed fluctuating flow controlled forcing method can be applied to multiple fluctuating velocity components and couple their calculation to amplify the existing fluctuations present in the precursor flow field so that prescribed anisotropic Reynolds stress tensors can be reproduced. The new method was tested on high and low Reynolds number turbulent boundary layer flows, where the proposed fluctuating flow controlled forcing method greatly accelerated the development of the turbulent boundary layers when compared with cases without controlled forcing and with only the original controlled forcing.

It is well known that most fluid flows observed in nature or encountered in engineering applications are turbulent and involve separation. Fluid flows in turbines, diffusers and channels with sudden expansions are among the widely observed areas where separation substantially alters the flow field and gives rise to complex flow dynamics. Such types of flows are referred to as internal flows since they are confined within solid surfaces and predominantly involve the generation or utilization of mechanical power. However, there is also a vast variety of engineering applications where the fluid flows past solid structures, such as the flow of air around an airplane or that of water around a submarine. These are called external flows and as in the former case the downstream evolution of the flow field is crucially influenced by separation. The present doctoral thesis addresses both internal and external separated flows by means of direct numerical simulations of the incompressible Navier-Stokes equations. For internal flows, the wall-driven flow in a onesided expansion channel and the pressure-driven flow in a plane channel with a single thin-plate obstruction have been studied in the fully developed turbulent state. Since such geometrical configurations involve spatially developing turbulent flows, proper inflow conditions are to be employed in order to provide a realistic fully turbulent flow at the input. For this purpose, a newly developed technique has been used in order to mimic an infinitely long channel section upstream of the expansion and the obstruction, respectively. With this approach, we are able to gather accurate mean flow and turbulence statistics throughout each flow domain and to explore in detail the instantaneous flow topology in the separated shear layers, recirculation regions as well as the recovery zones. For external flows, on the other hand, the flow past a prolate spheroid has been studied. Here, a wide range of Reynolds numbers is taken into consideration. Based on the characteristics of the vortical structures in the wake, the flow past a prolate spheroid is classified as laminar (steady or unsteady), transitional or turbulent. In each flow regime, the characteristic features of the flow are investigated by means of detailed frequency analysis, instantaneous vortex topology and three-dimensional flow visualizations.

In this work the model ASAM is enriched with new eddy viscosity based dynamic Smagorinsky subgrid-scale models. Therefore the model is more physically based to study atmospheric flow configurations at several atmospheric scales with main focus to urban scale flow with building-resolved resolution. The implemented dynamic procedures work well and showed good agreement to literature data. In a convective atmospheric boundary layer (ABL) the dynamic Smagorinsky coefficient reaches maximum values of 0.15 and decreases towards the surface or in stable stratified flow regimes. Vertical profiles of the Smagorinsky coefficient in a diurnal cycle of ABL depict typical behaviour of the dynamic Smagorinsky coefficient in near surface flow, free-stream, or stable stratified flow. Furthermore a modified inflow generation approach is proposed to produce fully turbulent flow fields. To modify a mean flow turbulent fluctuations are generated by superposition of sinusoidal and cosinesoidal modes. Due to the implementation of this inflow method the model ASAM has the ability to reproduce a given wind field with information from its mean wind speed and their fluctuation energy spectrum. The model configuration developed in this work is able to reproduce flow structure in a complex urban geometry. The Mock Urban Setting Test (MUST) experiment represent an urban roughness geometry by placing 120 shipping containers ordinary arranged in an array. The used building-resolved resolution is able to capture dynamic flow structures like specific wake flow, recirculation regions or eddy detachment. The dynamic fluctuating behaviour of the wind velocity components is reproduced by the model with regard to peak magnitudes and their temporal occurrence. Satisfying agreement is found between tracer gas dispersion field measurements and the model results by capturing the fluctuating concentration magnitude and in some extend the mean values.

Lors du développement d’un nouvel avion, l’estimation des échanges d’énergie entre l’air ambiant et les parois est une donnée cruciale pour la conception aérothermique. Cette conception repose de plus en plus sur des simulations numériques mais certains phénomènes d’aérothermique externe, comme le jet débouchant du système de dégivrage des nacelles moteur, montrent les limites des modèles RANS classiques. La simulation des grandes échelles (LES) se révèle bien adaptée à ce type de phénomène mais se heurte à un coût de calcul extrêmement élevé pour ces écoulements pariétaux à très grand nombre de Reynolds. Pour lever cette limitation, cette thèse propose l’étude de deux briques fondamentales : la LES avec loi de paroi (WMLES) conjuguée à l’injection d’une couche limite turbulente à l’entrée du domaine. Pour une meilleure compréhension et une utilisation fiable de l’approche loi de paroi, on se concentre tout d’abord sur les sources d’erreur qui lui sont associées. Après les avoir identifiées, on propose une correction de l’erreur de sous-maille ainsi qu’une loi de paroi adaptée aux écoulements compressibles. Grâce à ces deux éléments, on obtient une estimation correcte du flux de chaleur pariétal sur des simulations WMLES de canal plan supersonique sur parois froides. Puis, pour préparer la transition vers des applications plus industrielles, on introduit un schéma numérique plus dissipatif ce qui nous permet d’étudier l’influence de la méthode numérique sur l’approche loi de paroi. Dans une seconde partie dédiée à l’injection de couche limite pour la WMLES, on sélectionne une méthode basée sur l’injection de perturbations combinée à un terme de contrôle volumique. On montre que des simulations WMLES utilisant cette méthode d’injection permettent d’établir une couche limite turbulente réaliste à une courte distance en aval du plan d’entrée, à la fois sur une plaque plane mais également sur un écoulement de jet débouchant à la géométrie plus complexe, représentative d’un cas avion. / During the design of a new aircraft, the prediction of energy exchanged between the ambient air and the aircraft walls is crucial regarding aerothermal design. Numerical simulations plays a role of increasing importance in this design. However classical RANS models reach their limits on some external aerothermal flows, like the jet-in-cross-flow from the anti-icing system oh the engine nacelles. The large eddy simulation (LES) is well suited to this kind of flow but faces an extremely large computational cost for such high Reynolds number wall-bounded flows. To remove this limitation, we propose two building blocks: the Wall Modeled LES (WMLES) combined with a turbulent inflow generation. For a better understanding and a reliable use of the WMLES, we first focus on the sources of error related to this approach. We propose a correction to the subgrid-scale error as well as a wall model suitable for compressible and anisothermal flows. Thanks to these two elements, we correctly predict the wall heat flux in WMLES computations of a supersonic isothermal-wall channel flow. Then, to allow the computation of more industrial flows, we introduce some numerical dissipation and study its effect on the wall modeling approach. The last part is dedicated to turbulent inflow generation for WMLES. We select a method based on synthetic perturbation combined with a dynamic control term. We validate this method on WMLES computations of a flat plate turbulent boundary layer and a hot jet-in-cross-flow representative of an industrial configuration. In both cases, we show that a realistic turbulent boundary layer is generated at a small distance downstream from the inlet plane.

In this work the model ASAM is enriched with new eddy viscosity based dynamic Smagorinsky subgrid-scale models. Therefore the model is more physically based to study atmospheric flow configurations at several atmospheric scales with main focus to urban scale flow with building-resolved resolution. The implemented dynamic procedures work well and showed good agreement to literature data. In a convective atmospheric boundary layer (ABL) the dynamic Smagorinsky coefficient reaches maximum values of 0.15 and decreases towards the surface or in stable stratified flow regimes. Vertical profiles of the Smagorinsky coefficient in a diurnal cycle of ABL depict typical behaviour of the dynamic Smagorinsky coefficient in near surface flow, free-stream, or stable stratified flow. Furthermore a modified inflow generation approach is proposed to produce fully turbulent flow fields. To modify a mean flow turbulent fluctuations are generated by superposition of sinusoidal and cosinesoidal modes. Due to the implementation of this inflow method the model ASAM has the ability to reproduce a given wind field with information from its mean wind speed and their fluctuation energy spectrum. The model configuration developed in this work is able to reproduce flow structure in a complex urban geometry. The Mock Urban Setting Test (MUST) experiment represent an urban roughness geometry by placing 120 shipping containers ordinary arranged in an array. The used building-resolved resolution is able to capture dynamic flow structures like specific wake flow, recirculation regions or eddy detachment. The dynamic fluctuating behaviour of the wind velocity components is reproduced by the model with regard to peak magnitudes and their temporal occurrence. Satisfying agreement is found between tracer gas dispersion field measurements and the model results by capturing the fluctuating concentration magnitude and in some extend the mean values.:1 Introduction 1 2 Fundamentals of Large-Eddy Simulation in atmospheric boundary layers 7 2.1 The atmospheric boundary layer . . . . . . . . . . . . . . . . . . . . . 7 2.2 Atmospheric turbulence . . . . . . . . . . . . . . . . . . . . . . . . . 8 2.3 Basic equations of LES . . . . . . . . . . . . . . . . . . . . . . . . . . 9 3 Subgrid-scale modelling 15 3.1 Eddy viscosity subgrid-scale models . . . . . . . . . . . . . . . . . . . 15 3.1.1 Smagorinsky subgrid-scale model . . . . . . . . . . . . . . . . 16 3.1.2 Dynamic Smagorinsky subgrid-scale model . . . . . . . . . . . 18 3.1.3 Scale-dependent dynamic Smagorinsky subgrid-scale model . . 23 3.2 Implementation in the All Scale Atmospheric Model (ASAM) . . . . . 26 3.2.1 General description of ASAM . . . . . . . . . . . . . . . . . . 26 3.2.2 Subgrid-scale modelling in ASAM . . . . . . . . . . . . . . . . 27 3.3 Applications to meteorological situations . . . . . . . . . . . . . . . . 37 3.3.1 Stable and unstable stratified atmospheric boundary layers . . 37 3.3.2 Flow over periodic sinusoidal hill . . . . . . . . . . . . . . . . 49 4 Generation of turbulent inflow conditions 51 4.1 The necessity of turbulent inflow . . . . . . . . . . . . . . . . . . . . 51 4.2 Synthetic turbulent inflow generation method . . . . . . . . . . . . . 53 4.3 Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 4.4 2D simulation results . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 5 Mock Urban Setting Test Experiment (MUST) 65 5.1 Micro-scale urban simulation . . . . . . . . . . . . . . . . . . . . . . . 65 5.2 Description of the experiment . . . . . . . . . . . . . . . . . . . . . . 68 5.3 Wind tunnel measurenments of MUST . . . . . . . . . . . . . . . . . 70 5.4 Numerical MUST simulation with ASAM . . . . . . . . . . . . . . . . 72 5.4.1 Choice of initial condition . . . . . . . . . . . . . . . . . . . . 75 5.4.2 Results of simulating case 2682353 . . . . . . . . . . . . . . . 81 5.4.3 Results of simulating case 2681829 . . . . . . . . . . . . . . . 98 5.4.4 Case resume . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 6 Summary and outlook 111 6.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 6.2 Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114 7 Bibliography 117 List of Figures 127 List of Tables 135 Acronyms 137 Nomenclature 139 Acknowledgement 143 List of Publications 145

L'objectif de ce travail est de modéliser les sources et la propagation atmosphérique du bruit généré par les éoliennes afin de mieux comprendre les caractéristiques de ce bruit à grande distance et d'aider les fabricants d'éoliennes et les développeurs de parc à respecter la réglementation. En couplant des modèles physiques de source aéroacoustique et de propagation, nous sommes capables de prédire les spectres de bruit, ainsi que la directivité et les modulations d'amplitude associées, pour différentes conditions atmosphériques. Le bruit aérodynamique large bande, à savoir le bruit d'impact de turbulence,le bruit de bord de fuite et le bruit de décrochage, est généralement dominant pour les éoliennes modernes. Le modèle analytique d'Amiet est choisi pour prédire le bruit d'impact de turbulence et le bruit de bord de fuite, en considérant plusieurs améliorations par rapport à la théorie initial : 1, une correction empirique pour l'épaisseur du bord d'attaque est introduite dans le calcul du bruit d'impact de turbulence ; 2, un modèle spectral des fluctuations de pression pariétale proposé récemment pour un écoulement avec gradient de pression défavorable est utilisé dans le calcul du bruit de bord de fuite. Ces modèles sont validés par comparaison avec des mesures de la littérature en soufflerie avec des profils fixes.Le modèle d'Amiet est ensuite appliqué à une éolienne complète pour prédire le bruit émis en champ proche. L'effet de la rotation des pales et l'effet Doppler sont pris en compte. On utilise d'abord des profils de vent constant sans turbulence, puis l'effet du cisaillement du vent et de la turbulence atmosphérique sont inclus à l'aide de la théorie de la similitude de Monin-Obukhov. De bons accords sont obtenus avec des mesures sur site éolien lorsque l'on considère à la fois les bruits de bord de fuite et d'impact de turbulence. On retrouve à l'aide du modèle les caractéristiques classiques du bruit des éoliennes, comme la directivité et les modulations d'amplitude. Des comparaisons avec un modèle semi-empirique montrent que le bruit de décrochage peut être significatif dans certains conditions.L'étape suivante consiste à coupler la théorie d'Amiet avec des modèles de propagation pour estimer le bruit à un récepteur en champ lointain. On étudie dans un premier temps un modèle analytique de propagation en conditions homogènes au-dessus d'un sol d'impédance finie. On montre que l'effet de sol modifie la forme des spectres de bruit, et augmente les modulations d'amplitude dans certains tiers d'octave. Dans un second temps, une méthode pour coupler le modèle de source à un code d'équation parabolique est proposée et validée pour prendre en compte les effets de réfraction atmosphérique. En fonction de la direction de propagation, les niveaux de bruit varient car l'effet de sol est influencé par les gradients de vent et car une zone d'ombre est présente dans la direction opposée au vent. On discute pour finir l'approximation de source ponctuelle à l'aide des modèles de propagation analytique et numérique. / The purpose of this work is to model wind turbine noise sources and propagation in the atmosphere in order to better understand the characteristics of wind turbine noise at long range and to help wind turbine manufacturers and wind farm developers meet the noise regulations. By coupling physically-based aeroacoustic source and propagation models, we are able to predict wind turbine noise spectra, directivity and amplitude modulation in various atmospheric conditions.Broadband noise generated aerodynamically, namely turbulent inflow noise, trailing edge noise and separation/stall noise, is generally dominant for a modern wind turbine. Amiet's analytical model is chosen to predict turbulent inflow noise and trailing edge noise, considering several improvements to the original theory: 1, an empirical leading edge thickness correction is introduced in the turbulent inflow noise calculation; 2, a wall pressure fluctuation spectrum model proposed recently for adverse pressure gradient flow is used in the trailing edge noise predictions. The two models are validated against several wind tunnel experiments from the literature using fixed airfoils.Amiet's model is then applied on a full-size wind turbine to predict the noise emission level in the near field. Doppler effect and blade rotation are taken into account. Cases with constant wind profiles and no turbulence are used first, then wind shear and atmospheric turbulence effects obtained from Monin-Obukhov similarity theory are included. Good agreements against field measurements are found when both turbulent inflow noise and trailing edge noise are considered. Classical features of wind turbine noise, such as directivity and amplitude modulation, are recovered by the calculations. Comparisons with a semi-empirical model show that separation noise might be significant in some circumstances.Next, Amiet's theory is coupled with propagation models to estimate noise immission level in the far-field. An analytical model for the propagation over an impedance ground in homogeneous conditions is studied first. The ground effect is shown to modify the shape of the noise spectra, and to enhance the amplitude modulation in some third octave bands. A method to couple the source model to a parabolic equation code is also proposed and validated to take into account atmospheric refraction effects. Depending on the propagation direction, noise levels vary because the ground effect is influenced by wind shear and a shadow zone is present upwind. Finally, the point source assumption is reviewed considering both the analytical and numerical propagation models.