Progress in Hybrid RANS-LES Modelling

A step change in aircraft performance, the European aircraft industry is convinced this is a necessary objective if it is to stay competitive and to allow continued growth reducing the environmental impact as proposed by the European policymakers / by Innovation Union. New capabilities will be essential in exploring new concepts including alternative configurations, flow control technologies, laminar flow designs, and other new approaches enabling the necessary step change in performance on which the industry is relying. Simulation technologies are considered to ultimately provide thus capabilities that will underpin future aircraft design processes [1].

Aircraft design is a very competitive and demanding field. Highly optimised design with the objective of lower fuel consumption, lighter, quieter, safer, good performance and handling qualities involves a large number of different disciplines (aerodynamics, structure, system, vibration, acoustic, etc.) in the design process. This is a very difficult task which requires large experience together with highly efficient and accurate design/optimisation tools. An advanced toolset acting as a virtual facility, providing full information about design status, is the target of the European Aircraft industry. Automatically predicting flow physics, forces, radiated acoustics, stresses, evolution of the design status, and the optimal shape for any specified constraints. Moreover such tool needs to be extremely accurate and performs in realistic engineering design timescales. Numerical simulation tool is an essential target for every company involved in aeronautics. With this, engineers are set free to design and innovate rather than spending wasted time ‘driving’ their design system. Fast and efficient designs in short timescales, possibility to investigate innovative and challenging solutions with breakthrough technologies, virtual certification with lower costs, and applications involving other disciplines are all outcomes of such a tool. Regrettably, such a tool or toolset does not exist today, however there is recognition that such a capability needs to put in place, if the aerospace industry is to meet future performance and environmental targets. Industrial numerical simulation tools are presently suffering two main drawbacks that prevent their full industrial deployment for massive applications. These are: excessively large computational time for problems of industrial relevance, and the reliability and accuracy of the solutions at flight extremes. These two deficiencies are however linked, and in many cases indistinguishable.

The paper will provide an overview of hybrid RANS-LES methods currently used in industrial flow simulations and will evaluate the models for a variety of flow topologies. Special attention will be devoted to the aspect of global vs. zonal approaches and aspects related to interfaces between RANS and LES zones.

Variable-resolution (VR) turbulence simulations possess ideal attributes for engineering applications as they purport to yield the best accuracy possible for any prescribed level of computational effort. However, at the current time, these accuracy-on-demand approaches are not considered theoretically rigorous. It is argued that pragmatic considerations that motivate the formulation of VR methods automatically preclude a theoretically rigorous approach. In this paper, we argue that VR approaches can be based on strong theoretical underpinnings without sacrificing numerical robustness and practical utility. We demonstrate that the partially-averaged Navier-Stokes (PANS) VR approach is based on strong physical and mathematical foundation and yet is robust enough for complex practical flows. We present important PANS theoretical attributes followed by results from complex flow computations.

Two types of recent results for the simulation of wing flows at relatively low Reynolds numbers are presented. One is a series of the flow simulations over simple wings, which eventually help the wing design of Mars flyer. The other is a series of the similar wing flow simulation but with DBD plasma actuator that reduces flow separation. Simulations are conducted with highly accurate spectral-like compact difference scheme that reduces the number of grid points with keeping same spatial resolution. With this method, iLES is used as a main analytical tool for the simulations. There appears strong Reynolds number effect and small change of the Reynolds number may drastically change the aerodynamic characteristics especially for thick wings. Thin wing has linear lift characteristics similar to thick wings at high Reynolds numbers, but flow structure is totally different from so-called potential flows. Wing flow simulation but with DBD plasma actuator shows that iLES captures flow structure induced by the DBD plasma actuator and transition to turbulent flows may be one of the important factors of the flow control by these devices. It is also shown that both flow separation and flow reattachment are the key factors for the simulation examples presented here, which requires LES type of simulations.

This paper presents the European collaborative project “Advanced Turbulence simulation for Aerodynamic Application Challenges”, i.e. its background, objectives, approach and status. As an example of the outcome of this project, results obtained by some partners for one of the test cases employed are compared and discussed highlighting the status of today’s hybrid RANS/LES approaches.

The present work deals with the development of an instability-sensitive turbulence model on the Second-Moment Closure level and its application to flow configurations of increasing complexity featured by boundary layer separation. The model scheme adopted, functioning as a ‘sub-scale’ model in the Unsteady RANS framework, represents a differential near-wall Reynolds stress model formulated in conjunction with the scale-supplying equation governing the homogeneous part of the inverse turbulent time scale

ω

h

(

ω

h

 = 

ε

h

/

k

). The latter equation was straightforwardly obtained from the model equation describing the dynamics of the homogeneous part

ε

h

(

ε

h

 = 

ε

 − 0.5

ν

 ∂ 

2

k

/ ( ∂ 

x

j

 ∂ 

x

j

), Jakirlic and Hanjalic, 2002) of the total viscous dissipation rate

ε

by applying the derivation rules to the expression for

ω

h

. The model capability to account for the vortex length and time scales variability was enabled through a selective enhancement of the production of the dissipation rate in line with the SAS proposal (Scale-Adaptive Simulation, Menter and Egorov, 2010) pertinent particularly to the highly unsteady separated shear layer region. The predictive performances of the proposed model (solved in conjunction with the Jakirlic and Hanjalic’s Reynolds stress model equation) were tested by computing the fully-developed channel flow at different Reynolds numbers, backward-facing step flow, periodic flow over a smoothly contoured 2-D hill in a range of Reynolds numbers and flow in a 3D-diffuser.

In this work, we introduce a recently proposed framework for hybrid LES/RANS modeling and its preliminary applications on simple flows. In this framework, the filtered and Reynolds averaged Navier-Stokes (RANS) equations are solved simultaneously in the whole domain. The novelty of the framework is the dual-solution approach and the consistency between the two solutions achieved via additional drift terms. A hybrid LES/RANS solver is developed within this framework and used to simulate flows in a plane channel and flows in a channel with periodic hills. The results demonstrate that the hybrid solver leads to significantly improved results compared to traditional LES on the same grid.

This work presents a scalar flux model in the framework of a hybrid RANS-LES modelling. The model is tested on a heated channel flow at different Prandtl numbers and on a T-junction. Results show a good agreement with both DNS and experimental data.

For high-Reynolds-numbers wall-bounded flows, large-eddy simulation (LES) combined with a wall stress model (WSM) is frequently used. The mean velocity of turbulent boundary layers at high-Reynolds-numbers follows a logarithmic distribution near the wall. However, in LES of high-Reynolds-number wallbounded flows, an overshoot of the mean velocity gradient near the wall is often reported. Many attempts have tried to suppress this overshoot of mean velocity gradient. However, a successful explanation on the relationship between the mean velocity gradient and flow properties, accounting for effects of subgrid-scale model, numerical scheme, and grid set-up has not yet been reported. In the current study, we elaborate a relationship between the mean shear and its budgets for the case of wall-modeled LES.We show that the overshoot of the mean shear is not necessarily caused by over-dissipation, as often reported in literature. Moreover, we proposed a novel hybrid scheme for the Smagorinsky model, where the model coefficient is determined dynamically near the wall, based on the relationship between the desired logarithmic mean shear and the SGS terms composing the bulk of the budgets for the mean shear. The normal Smagorinsky model is then employed far away from the wall. We show that this new model successfully yields the desired logarithmic velocity distribution near the wall.

We present a novel simulation tool-constrained large eddy simulation (CLES), for numerical experiments on the wall-bounded turbulent flows. Different from the traditional large eddy simulation(LES) and the available hybrid RANS/LES approaches, the CLES method computes the whole flow domain by solving the LES equations with a Reynolds-stress-constrained (RSC) subgrid-scale (SGS) stress model in the near-wall region and a traditional SGS stress model in the rest.The CLES approach is validated by simulating the turbulent channel flow and flow around a circular cylinder. With the same grid resolutions, CLES can successfully simulate all these flow regimes as well as DES and other available methods. For the case of attached flows, CLES is able to eliminate the non-physical Log-Layer Mismatch problem in traditional hybrid RANS/LES methods successfully, and to predict mean velocity profile, turbulent stresses and skin friction coefficient more accurately compared with the DES. For the case of detached flows, the performance of CLES is comparable to DES.

Among various hybrid RANS/LES methodologies, Speziale’s Very Large Eddy Simulation (VLES) is one that was early proposed and is a unified simulation approach that can change seamlessly from RANS to DNS depending on the numerical resolution. The present study proposes a new improved variant of the original VLES model. The advantages are achieved in two ways: (1) RANS simulation can be recovered near the wall which is similar to the Detached Eddy Simulation (DES) concept; (2) An LES subgrid scale model can be reached by the introduction of a third length scale, i.e. integral turbulence length scale. Thus the new model can provide a proper LES mode between the RANS and DNS limits. This new methodology is implemented in the standard

k

 − 

ε

model and Wilcox’s

k

 − 

ω

model. Applications are conducted for the turbulent channel flow at

Re

τ

 = 395 and turbulent flow past a square cylinder at

Re

 = 22000. Results are compared with previous studies. It is demonstrated that the new method is quite effective in resolving the large flow structures, and can give satisfactory predictions on a very coarse mesh.

The difference and connection between RANS and LES is briefly discussed. After reviewing the success of several LES models some necessary qualities for modeled length scale are concluded. A new 3D von Karman length scale which leads to SAS models is thus proposed for real-life flow simulation. Its “LES” function is extensively evaluated via the simulation of benchmark isotropic decaying turbulence. This length scale can fully make use of the local mesh and yield satisfactorily fine results even at coarse mesh resolutions, and moreover the results have some sort of unity.

The present paper aims to provide an efficient and flexible solution to carry out RANS to WMLES transitions, when near wall turbulent flows are involved, in a Zonal Detached Eddy Simulation (ZDES) context. Indeed, WMLES generally suffer from very long transient states. The main purpose is to broaden DES-type method range of applications to cases where the overall flow physics is driven by the near wall turbulence, while being affordable in an industrial context. Among other, shock/boundary layer interactions and shallow recirculation bubles are the main targeted applications.

The solution proposed in this paper consists in resorting to the dynamic forcing method recently proposed by the authors (R. Laraufie, S. Deck, P. Sagaut. A dynamic forcing method for unsteady turbulent inflow conditions.

Journal of Computational Physics

230

(23), 8647-8663 (2011)), combined with a ZDES resolution method and a synthetic turbulence generation approach. The dynamic forcing method, firstly employed with the Synthetic Eddy Method (SEM) achieves dramatic transition distance shortening (~65 % in the present case).

Furthermore, the ability of the dynamic forcing method to regenerate a turbulent boundary layer from one of the most simple turbulence generation method, namely random noise, is also demonstrated. It is then shown that results similar to those obtained when the SEM turbulent inflow is used, can be achieved with a simple white noise at the inlet, when the dynamic forcing method is employed. Such flexibility is expected to make one able to use the dynamic forcing method with whatever synthetic turbulent generation method.

An improvement of the Vortex Method proposed by Sergent [19] and Mathey

et al

. [14] is presented. The method can be used to generate inflow boundary conditions for a LES or instantaneous interfacing conditions in a zonal RANS/LES simulation. The method is tested in a channel flow case at

Re

τ

 = 590 and compared to a bi-periodic LES case. Analysis of the generated field and its evolution in the streamwise direction is provided using the vorticity fluctuations and the velocity-derivative skewness and shows this method as viable and cost-effective.

The present work aims to investigate the usage of embedded regions of turbulent flow simulation near to the point of separation; towards an approach whereby discrete regions of turbulent resolving techniques are used within a domain predominantly solved using the RANS equations. The common Delayed Detached Eddy Simulation (DDES) approach is here used to compute the flow around a 2D hump centred within a ‘full domain’ i.e. also incorporating an upstream section. Subsequently the domain length is reduced and the flow is started at two locations close to the separation point by means of unsteady turbulent inlet conditions. The Divergence Free Synthetic Eddy Method (DF-SEM) and its predecessor are tested for their ability to reproduce the original DDES results from the full domain. In the present case we aim to return a minimal disturbance from the full domain solution and thus herein we do not focus on the predictive accuracy of the selected DDES approach. The motivation for this technique is to provide guidance for the optimal reduction of embedded regions of turbulent simulation in complex applications; i.e. including multiple instances of separated flow. Some comments regarding computational expense of the method are also provided.

A synthetic turbulence generation (STG) method for flows at low and high Reynolds and Mach numbers to provide LES inflow boundary conditions of zonal Reynolds-averaged Navier-Stokes (RANS)- large-eddy simulation (LES) method simulations is presented. The present method separates the LES inflow plane into three sections where a local velocity signal is decomposed from the turbulent flow properties of the upstream RANS solution. Depending on the wall-normal position in the boundary layer the local flow Reynolds and Mach number specific time, length, and velocity scales with different vorticity content are imposed on the LES inflow plane. The STG method is assessed by comparing the resulting skin-friction, velocity, and Reynolds-stress distributions of zonal RANS-LES simulations of subsonic and supersonic flat plate flows with available pure LES, DNS, and experimental data. It is shown that for the presented flow cases a satisfactory agreement within a short RANS-to-LES transition of two boundary-layer thicknesses is obtained.

This paper presents ZDES simulations of jets on industrial configure-tions, taking into account the turbulent flow rate coming from the engine. The strong influence of this turbulent ratio on the jet development has already been demonstrated in previous researches and it has to be taken into account for such realistic simulations. Hence, a turbulent flow generator module for LES-like simulations was first developed and validated. Next, studies on industrial double flux configurations are presented and compared with dedicated experiments. The results show that the ZDES simulations presented are in a very good agreement with measurements and the averaged mean and turbulent results into the jets are properly predicted thanks to the turbulent flow generation technique.

The paper presents a PANS simulation of the flow around a cuboid influenced by crosswind. The results of the PANS prediction are validated against experimental data and results of a LES simulation made using the same numerical conditions as PANS. The PANS shows good agreement with the experimental data. The prediction of PANS was found to be better than that of the reference LES in flow regions where simulations suffered from poor near-wall resolution.

The turbulent flow around a generic configuration of a landing gear (’the tandem cylinder’) is simulated and analysed physically at Re = 1.66x10

5

, by means of hybrid RANS-LES turbulence modelling approaches. In the present study, the Delayed Detached Eddy Simulation (DDES) approach has been employed. The DDES-OES modelling has been considered, especially involving turbulence length scale reconsiderations in the statistical part, by means of the Organised Eddy Simulation, (OES), to take into account non-equilibrium turbulence effects. The DDES-

k

-

ω

SST model is also considered. The results, obtained by means of two different time steps are compared with experiments carried out at the NASA-Langley Research Centre in the context of ATAAC EU-program in which the tandem cylinders is one of the ‘stepping stones’. In the present study, the benefits of these hybrid approaches have been discussed for capturing the vortex dynamics and frequency modes responsible for aerodynamic noise production in the context of landing gear configurations.

Three advanced DES-type models coupled with adaptive dissipation scheme, DDES-2003/2006 and IDDES, are applied to predict the unsteady flow past tandem cylinders. The main differences among them are the shield functions and length scales, which leads to significant differences in mean turbulence kinetic energy, root mean square of pressure, instantaneous vorticity, and so on. The computational results are compared with almost all available measurements. These three models show good capability for the massive separation flows. IDDES performs relatively better than other two SST-DDES models. Furthermore, IDDES matches well with the measurements with trips on the rear cylinder surface.

A shock induced boundary-layer separation (SBLI) occurring in a duct at

M

= 1.4 has been analyzed using hybrid RANS-LES methods. The shock wave interacts with the turbulent wall boundary layers and triggers flow separation in the duct corners. The main purpose of the present work is to highlight the difficulties in modeling SBLI, particularly, when hybrid RANS-LES models are used. Results computed using different turbulence models are presented and discussed in comparison with available experimental data. Based on a number of simulations, some issues are addressed and some critical remarks are provided for potential improvements using turbulence-resolving modeling approaches in future work.

The numerical analysis of massively separated flow around a circular cylinder at a high Reynolds number is presented in this paper. The simulations are carried out using the hybrid RANS-LES simulations, namely detached-eddy simulation (DES) and its modified version, delayed detached-eddy simulation (DDES). The computed pressure and skin friction coefficients around the surface of the circular cylinder are compared with the available experimental data and other numerical studies with encouraging results. The power spectral density (PSD) comparison of DES and DDES is carried out in the region with high vortex shedding and the energy spectrum depicts the energy cascade in line with the Kolmogorov -5/3 theory for both DES and DDES results. It is found that these hybrid RANS-LES simulation techniques are able to simulate the flow physics of the massively separated flows reasonably well. For such type of flows, the results of the modified scheme DDES are similar to the DES simulation with no adverse effects on the quality of the predicted flow field. The additional computational cost of DDES in comparison with DES is also addressed.

The M219 cavity benchmark problem is used to compare the performance and the applicability of two commercial software, namely Ansys Fluent and Exa Powerflow. The computational domain is small and is concentrated around the cavity to limit the computation time and thus permit to assess the possibility of carrying out such simulations in an industrial context. Despite the small length of the domain, far-field conditions are imposed at the open boundaries which induces some unphysical pressure wave reflections. On the reference cavity case (Ma=0.85, P=62900), the second and third tones are well captured but the background noise and sound pressure level are higher than previously mentioned in the literature. Wave reflection and boundary layer features are thought to be the main causes of these discrepancies. A case at Mach number equal to 0.30 was carried out; Fluent and Powerflow present similar results in terms of acoustics characteristics. The mean flow computation shows slight differences which are certainly due to different boundary layers attacking the cavity leading edge.

Detached-Eddy Simulation (DES) is a promising method for efficient simulation of broadband noise at minimal computational cost. Here, results from a study of broadband noise simulation using state-of-the-art DES methods are presented for a rudimentary landing gear configuration. The DDES and IDDES variants are compared with experiments in incompressible simulations. IDDES shows mild improvement in agreement and some increase in the resolution of high frequencies. An attempt is made to independently verify published results for far-field sound prediction, using a compressible simulation coupled with Ffowcs-Williams/Hawkings (FWH) integration. In contrast to the published results, our results do not provide evidence of unexpectedly strong roles played by the ceiling or by quadrupoles. Our results furthermore predict much lower far-field noise levels than the published results. Good agreement between solid and permeable FWH surfaces is found as long as the permeable surfaces are open downstream.

The direct prediction of acoustic sources requires the numerical integration of the unsteady compressible Navier-Stokes equations in complex geometries. The acoustic sources are extracted from the numerical solution and then injected as a source term in a hybrid methodology. Obtaining the small acoustic scales demands very fine meshes where a statistically converged solution has been obtained. Recently, the efforts of the numerical community have been directed towards the application of Large Eddy Simulation to predict aeroacoustic sources. In this kind of predictions, good compromise with experimental results is obtained, however, it is recognized that the computational cost of this kind of simulations is still excessive, and less demanding solutions become essential. Nevertheless, adequate calibration of 2D or 3D computations with standard turbulence models is expensive, and although in use, it is not clear that their solutions can be as precise as those obtained with LES. In this work, the numerical evaluation of different methodologies and their impact on aeroacoustic prediction is investigated. Two and three dimensional flows and different types of turbulence models are considered in an open cavity flow problem at Mach number of 0.8 and Reynolds number of 8.6 · 10

5

, for which an extensive experimental analysis has been performed by other authors.

Turbulence-resolving simulations have been performed using hybrid RANS-LES approaches for the turbulent flow around a three-element high-lift configuration. The main purpose is to explore the effect of some modeling-related numerical aspects on the simulation of resolved velocity and pressure fluctuations as potent noise-generating sources. Along with a presentation of resolved instantaneous and mean flow features, the impact of the time step and the spanwise extent of the computational domain is investigated. It is shown that the temporal resolution and the spanwise extension of the computational domain impose effects not only on the prediction of mean flow, but more significantly on the correlation of resolved turbulent structures, which may consequently affect the accuracy of flow-generated noise properties.

The mechanism and efficiency of noise control in supersonic cavity flows with steady upstream mass blowing are numerically investigated. A slotted jet is placed in the upside of cavity leading edge. The mass blowing is simulated by specifying a vertical velocity ejecting through the slotted jet. The steady upstream mass blowing is an effective approach for the noise suppression in supersonic cavity flows. The strength of the resonant noise and the broadband noise are decreased with a delightful amplitude, that is, approximately 15 dB SPL decrease in the dominant mode and 5 dB SPL decrease in the broadband noise. Two primary mechanisms are addressed for the noise control with steady upstream mass blowing, lifting up of the cavity shear-layer and disruption of shear-layer instability.

Flow control has shown a potential in reducing the drag in vehicle aerodynamics. Its effects are highly related to the flow structures, including small ones, and thus accurate predictions of the flow field are needed for numerical investigation. The numerical study reported in the present paper deals with active flow control for a quasi-2D simplified vehicle model by a synthetic jet. A newly developed PANS turbulence model is used, based on the

ζ

 − 

f

RANS turbulence model. The aim is to validate the performance of this model for the complex flow control problem. Results are compared with previous studies using LES and experimental data, including global flow parameters of Strouhal number, drag coefficients and velocity profiles. The PANS model predicts a drag reduction of 9.8%, which is close to previous LES results. The velocity profiles predicted by the PANS model agree well with experimental data for both natural and controlled cases. It is found that the PANS model is able to predict the flow control problem well, actually in a way not inferior to the LES method, and can thus be used for further flow control studies in vehicle aerodynamics.

Detached eddy simulation (DES) method is applied to investigate the influences of curved hole passage on the turbulent flow characteristics of jet in cross flow (JICF) for film cooling vane. Two types of hole configurations, straight and curved holes, on a flat plate are simulated at blowing ratio 0.5, with Reynolds number

Re

D

=4000 based on the hole diameter and main flow speed. The influences of hole curvature on film cooling are studied from the viewpoints of scalar turbulent advection. Both the longitudinal and transversal two-point auto-correlations of the fluctuating velocity and the corresponding wave number spectra are analyzed. More attention is paid at the differences, caused by hole curvature, in the statistical view of the turbulent flow structures and the transportation of turbulent energy from the energy-containing range to the smaller inertial sub-range in the wake region.

This study deals with the numerical investigation of the flow around a high-lift airfoil with deployed slat and flap, with focus being made on the slat cove region. Two different techniques are used and compared to simulate the unsteady flow around such an airfoil, both being based on a hybrid RANS/LES formulation in order to minimize as much as possible the computational resources. In this paper, a comparison of the results from these two approaches (namely ZDES and NLDE) is presented, both from the statistical and unsteady points of view. The two approaches provide results which are in excellent respective agreement.

In this paper the Spalart-Allmaras based Delayed Detached Eddy Simulation (DDES [1]) and Improved Delayed Detached Eddy Simulation (IDDES [2]) are used to simulate the flow about an industrially relevant airfoil-configuration with deployed high-lift devices. Here, the potential advantage of the computationally very challenging hybrid approaches over pure RANS simulations in the case of incipient separation is investigated.

In the present paper a Delayed Detached-Eddy Simulation (DDES) based on the strain-adaptive linear Spalart-Allmaras (SALSA) model is performed in order to investigate the transonic flow over the OAT15A supercritical airfoil within the buffet regime. The results are compared with 2D and 3D-URANS computations using the SALSA model, as well as with experimental data. This study shows improvements achieved in the prediction of the flow unsteadiness and statistics by means of the DDES.

The flow around the Aérospatiale A-airfoil at the maximum lift condition with a chord Reynolds number of 2.0×10

6

is simulated using Unsteady Reynolds-Averaged Navier-Stocks Simulation (URANS), Detached Eddy Simulation (DES), Delayed DES (DDES) and a new approach. The new approach (named WAD-DES) is based on the Spalart-Allmaras (S-A) model and has a weighted average of destruction terms from the Smagorinksy model and the S-A model. The aim of this study is to investigate the behaviour of the S-A-based model working as a sub-grid scale (SGS) model in simulating mild trailing edge separation. The results show that in the near-wall region WAD-DES is better than DES and comparable to DDES. In the wake region, WAD-DES provides the closest velocity profiles to those from the experimental data, due to a reduced level of modelled turbulent viscosity. It is shown that this new WAD-DES approach inherits the advantages of DDES in simulating shallow separation, and also increases the accuracy of prediction in regions further away from the wall.

The study presented in this paper is motivated by the increasing need of high fidelity data for aircrafts in off-design conditions. The behaviour of turbulent structures computed using the Zonal Detached Eddy Simulation (ZDES) approach when they come through a chimera interface is investigated. Three different chimera (or overset) are compared to a reference ZDES computation on a structured grid, which is first validated against experiments. The results highlight the improved results of the ZDES compared to RANS and show that the accuracy of the simulation is preserved as long as the ratio of the grid densities of the overlapping meshes remains low.

The flow past a circular cylinder at a subcritical Reynolds number of Re=3900 was simulated by the method of detached-eddy simulation (DES). The objective of this present work is not to investigate the physical phenomena of the flow in detail but to study modeling as well as numerical aspects which influence the quality of DES solutions. Firstly, four typical spanwise lengths (D, 2D,

π

D/2 and

π

D, D the diameter of the cylinder) are chosen and the results are systematically compared. The trend of DES results along the span increment is different from previous large-eddy simulation (LES) investigation. DES method is more sensitive to the spanwise length and a wider spanwise length does not necessary improve the results. Then, the factor of mesh resolution is studied. Three kinds of grids, namely coarse, medium and refined, are adopted and the results show that both too coarse and over refined grids will deteriorate the performance of DES. The reason lies in the construction of DES employing the function of the distance to walls and grid spacing. Finally, different orders of numerical schemes are applied in the inviscid fluxes and the viscous terms. The discrepancies among different schemes are found tiny. However, the instantaneous flow structures produced by 5

th

order WENO with 4

th

order central differencing scheme are much richer than the others. That is, for the time-averaged quantities, the second-order accurate schemes are effective enough, whereas the higher-order accurate methods are needed to resolve the transient characteristics of the flow.

The capability of Detached-Eddy Simulation (DES) to predict the separated flow around a helicopter fuselage is examined and discussed. Results from two European research projects are shown: Completed simulations using a structured solver for a simplified helicopter geometry and initial results with an unstructured solver for a more complex geometry.

DES achieves encouraging agreement with experiments and an improvment over URANS in particular for the wake flow and surface pressure fluctuations. A strong dependency on the RANS model is seen, which is attributed in part to differences in the prediction of the sensitive separation from the smoothly-curved rear fuselage region. The grid design for the complex configuration is discussed and precursor RANS results are shown. Concerns regarding the treatment of multiple separation and reattachment by DES are expressed, which will be investigated further in future work.

The base flow of a generic rocket configuration is investigated numerically with different levels of turbulence modeling. At the nominal flow conditions, the comparison of numerical results with the experiments shows significant deviations in the vertical plane where a side support stands. A simulation of the open test section indicates two necessities of correction. On the one hand, an

C

p

increase of 0.015 is necessary to correlate the measured plenum pressure with the inflow location of the numerical simulation. On the other hand, a − 0.32° angle of attack modification should be accounted for a justified comparison between the DES results and the experiments. A strong sensitivity towards such small angles of attack has been observed later in the experiments but not in respective RANS solutions. The DES results agree well with the experiment based on the above-mentioned corrections.

For highly loaded compressor blades a significant improvement in performance prediction of hybrid RANS-LES methods compared to widely used RANS methods has been observed. The flow in an axial compressor close to stall conditions is simulated by DDES calculations on two different geometrical models to investigate the influence imposing periodicity of different domain sizes on the flow. Time traces of integral forces are compared for both configurations. Time averaged results of the simulations are compared to experiments. It is found that the direct periodicity of the one blade domain damps important features of the separated flow. This is expected to be particularly relevant to the phenomenon of Rotating Instability.

Application of DES method for swirling flows was considered. Swirling flow in a diffuser and vortex breakdown past an abrupt expansion were considered to test the application of detached eddy simulation method for swirling flows. Calculations of unsteady flows in draft tube of Kaplan turbine Holleforsen and in draft tube of high head Francis turbine were provided to investigate vortex rope precession in hydro turbine.

Large eddy simulation (LES) of turbulent flows in an experimental reciprocating internal combustion engine was carried out. The engine had a rectangular shaped combustor geometry and rectangular channel intake and exhaust manifolds with a large optical window to allow for detailed two-dimensional velocity field measurement in the entire combustion chamber. The objectives of this work were to study the structures of the tumble flow and turbulence in the combustion chamber and to examine different approaches for characterizing the incylinder turbulent flows. LES were performed for two different engine configurations, one with the intake channel included in the simulation and one with intake flow modeled as a simple plug flow at the exit of the intake channel, to investigate the effect of intake flow on the tumble flow and turbulence. The convergence of cycle-averaged statistics was investigated. It was found that for the ensemble averaged mean flow field 10 cycles LES could give reasonably converged mean velocity; however, more than 60 cycles were needed to generate converged rms of velocity fluctuation. A global turbulence intensity defined based on single cycle LES or PIV data was analyzed. This quantity was shown to characterize the overall turbulence intensity in the cylinder reasonably well.

A transonic cavity flow with a 5:1:1 aspect ratio is studied in the present work using a 2

nd

generation URANS modeling technique adapted to an EARSM model. An unstructured mesh made from hexahedral cells is used to perform these computations. The first part of the paper reports on recent improvement brought to the code. Results obtained on the M219 cavity are then presented and include the prediction of the mean flow and the tonal modes. The study also briefly looked at the influence of the time-step on the the prediction of the flow features.

A multiscale finite element method is applied to the Spalart-Allmaras turbulence model based detached-eddy simulation (DES). The multiscale method arises from a decomposition of the scalar field into coarse (resolved) and fine (unresolved) scales. It corrects the lack of stability of the standard Galerkin formulation by modeling the unresolved scales that cannot be captured by a given spatial discretization. The stabilization terms appear naturally and the resulting formulation provides effective stabilization in turbulent computations, where reaction-dominated effects strongly influence the boundary layer prediction. The validation of the multiscale-based DES is carried out on a backward-facing step. The time-averaged skin friction coefficient and pressure coefficient distributions are compared with the experimental and direct numerical simulation (DNS) results. Furthermore, the potential use of multiscale DES in computational wind engineering (CWE) is investigated. High-Reynolds flow over the Commonwealth Advisory Aeronautical Council (CAARC) standard tall building model is simulated by DES with both uniform and turbulent inflow. Time-averaged pressure coefficients on the exterior walls are compared with experiments. It is demonstrated that DES is able to resolve the turbulent features of the flow and accurately predict the surface pressure distributions under atmospheric boundary layer flows.

This study focuses on the application of detached eddy simulation (DES) for siting of wind turbines in complex terrain. The DES method uses a SST

k-ω

model as the Reynolds-averaged Navier-Stokes (RANS) model in the near-wall regions. The model switches to large eddy simulation (LES) mode if the dynamic length scale is greater than the local length scale. Therefore, in flow separation zones where the turbulent kinetic energy is large, the flow field is simulated with LES mode. This method is a standard practice in DES and many commercially available computational fluid dynamics (CFD) codes use it to determine the model behaviours. In contrast to traditional RANS studies, a significant advantage of DES is its capability of resolving a time-dependent flow field. One can observe the transient flow behaviours instead of a stationary mean value. This is useful if we want to understand the scale of fluctuating wind and the unsteadiness of the wind across the rotor area of a wind turbine. Applying this DES method can distinguish flow separation zones in complex terrain and this has been helpful to identify wind problems which may cause difficulties in operation of wind turbines.

The turbulent flow and the pollutant dispersion in organized cubic structures simulating an idealized building arrangement in a city is numerically investigated. The basic building arrangement under investigation is the experimental arrangement reported by Uehara et al (2000); it consists of an in line arrangement of cubic blocks having upwind a number of low rise square blocks which act as roughness elements for creating turbulent flow. The purpose of the investigation is to gain physical knowledge on the structure of the three dimensional flow and the pollutant dispersion mechanism, to compare with published two dimensional simulations and mainly to investigate the unsteady character of the flow field and its time scale characteristics.