Untitled

AIAA 2001-2405

Validation of European CFD Codes for SCT low-speed high-lift Computations U. Herrmann† (DLR), A. Press‡ (BAE SYSTEMS), C. Newbold° (DERA), P. Kaurinkoski¶ (HUT), C. Artilesƒ (INTA), J. v. Muijden♦ (NLR), G. Carrierı (ONERA) † ‡

DLR, Institut für Entwurfsaerodynamik, Lilienthalplatz 7, D-38108 Braunschweig, F. R. Germany

BAE SYSTEMS (Operations) Limited, Advanced Technology Centre - Sowerby, P.O. Box 5, FPC 267, Filton, Bristol, BS34 7QW, United Kingdom

° DERA, Küchemann Building, Farnborough, Ively Road, Farnborough, GU14 OLX, United Kingdom ¶ ƒ

Helsinki University of Technology, Laboratory of Aerodynamics, P.O. Box 4400, 02015 HUT, Finland

INTA, Instituto Nacional de Tecnica Aerospacial, Crta Ajalvir km 4, 28850 Torrejon de Ardoz-Madrid, Spain ♦

NLR, National Aerospace Laboratory, Anthony Fokkerweg 2, 1006 BM Amsterdam, Netherlands ı

ONERA, 29 Avenue de la Division Leclerc, BP 72, F-92322 Chatillon CEDEX, France

Abstract The validation of several European Navier-Stokes solvers for low-speed high-lift flow prediction around a Supersonic Commercial Transport (SCT) is reviewed. Realistic, project based, accuracy requirements are formulated by industrial partners within the EU research project EPISTLE (European Project for Improvement of Supersonic Transport Low Speed Efficiency) and fulfilled by the partners computations. Mesh density and turbulence model effects on flow topology and aerodynamic coefficients are described and compared to experimental results. It is found that increased mesh density compared to the used mesh is needed for accurate separation onset prediction. The algebraic Baldwin-Lomax turbulence model with Degani-Schiff extension (BLDS) was found to be a good engineering tool and able to predict highlift flows well into vortex onset.

about 250 PAX for productivity and the design range should be increased to about 5500 nm so that a reasonable percentage of long-range routes can be served. In addition, compliance with future community noise regulations is one of the most critical design requirement. Rather high L/D values at take-off conditions (in the order of 10) are required so that airport noise can be kept at tolerable levels. Previous research under the EC Framework IV project EUROSUP 1 2 provided and validated design capabilities for cruise conditions over sea (supersonic) and over land (transonic) but it did not resolve the design problem at low speed. For a SCT configuration there is little difficulty in achieving the necessary lift in low-speed flight, because of the relatively large wing area. The major challenge is to reduce the drag at low-speed conditions in order flight with reduced throttle settings and with lower noise levels. Current European SCT designs feature leading and trailing edge flaps in order to adapt the wing for achieving maximum aerodynamic efficiency at supersonic and transonic cruise. Naturally, these devices will also be used to optimize SCT low-speed performance. Current experiences in predicting SCT low-speed performance using Navier-Stokes codes, gained in a SCT low-speed work focused GARTEUR group, show that the governing flow phenomena are captured. This ability offers great potential to integrate numerical Navier-Stokes solvers in the low-speed high-lift system design process. This is the focus of the current EU research project EPISTLE: to overcome the low-speed design deficits detected during the previous EUROSUP project. This will be achieved by developing a design methodology based on numerical tools. The EPISTLE objectives to enable low-noise high-lift system designs by designing for low-drag are threefold:

Introduction Supersonic Commercial Transports (SCT) will be able to fulfill the increasing demand for reduced travel times and increasing transport productivity. The first supersonic commercial transport, Concorde, has demonstrated the feasibility of travel with supersonic speeds. Although Concorde flew scheduled supersonic flights for more than two decades, it did not capture a significant percentage of long range travel. A second generation SCT still targeted at a supersonic cruise speed of Mach 2 needs major improvements of aerodynamic performance data in order to achieve economic viability. The size of the aircraft should be increased from 100 to Copyright © 2000 by U. Herrmann. Published by the American Institute of Aeronautics and Astronautics, Inc. with permission

2 American Institute of Aeronautics and Astronautics

AIAA 2001-2405 •

the numerical tools, the flow solvers, are validated,

•

the low-speed high-lift system design space will be explored using the validated numerical tools,

used by the partners in this study are tabled below:

design rules will be extracted, applied and experimentally verified. The EPISTLE project started in March 2000 and has the following partners (in alphabetical order): EADS Airbus SA, BAE SYSTEMS Airbus UK, CIRA, DERA, DLR, EADS Airbus, HUT, INTA, NLR and ONERA. EPISTLE consists of several tasks, the first being an assessment of the prediction capabilities of European Navier-Stokes codes for aerodynamic effects of high-lift devices and separation onset. Furthermore the different numerical methods are qualified for the design work later on in the project. The partners active in this qualifying and validation task: BAE SYSTEMS, DERA, DLR, HUT, INTA, NLR and ONERA focus here on reporting selected results from this validation process of their codes enabling the forthcoming tasks of the EU-research project.

Partner

Flow solver

Turbulence model

BAE SYSTEMS Airbus UK

RANSMB

BLDS

DERA

SAUNAFLO

k-ω

DLR

FLOWer

BLDS / SA / k-ω

HUT

FINFLO

k-ωsst / kωrcsst

INTA

FLUENT

k-ε

NLR

ENSOLV

k-ω

ONERA

CANARI

k-l

•

Table 1: Flow solvers and turbulence models used by partners.

The Flow Solvers The computer codes used in this study solve the 3D unsteady, Reynolds-averaged Navier-Stokes equations in integral form on structured and unstructured meshes. Spatial discretization is done using finite volumes. Convective fluxes are discretized using central differences following the well known Jameson approach for most solvers using structured meshes. The FINFLO solver by HUT uses a formally third order upwind biased discretization. For the viscous fluxes either thin layer approximations or full Navier-Stokes formulations are used that are discretized using central differences. The overall accuracy of the solution procedures are second-order in space. Time integration is done explicitly with a multistage Runge-Kutta scheme for the structured codes. However for the FINFLO code, the steady state solution is obtained by an implicit pseudo-time integration utilizing an approximate factorization. Unsteady time accurate calculations are enabled using a dual time step approach following Jameson in one code. Several acceleration techniques are used like local time stepping, preconditioning, implicit residual averaging and full multigrid. Turbulence models used in this investigation are: Baldwin Lomax with the DeganiSchiff (BLDS) modification3 , the Spalart Allmaras (SA)4 one equation and some two-equation turbulence models, like the k-ε5 , k-l6 and the k-ω7 as well as turbulence model based on the last one8 9 . The different codes and turbulence models being

Validation Test Case and Accuracy Requirements The SCT configuration under investigation during the forerunning EU-research project EUROSUP, the intermediate GARTEUR action group as well as in the actual project EPISTLE is the ESCT configuration defined by the three European industrial partners SA (EADS Airbus SA), BAE (BAE SYSTEMS Airbus UK) and DA (EADS Airbus), see Fig. 1. The wing thickness, camber, twist and the leading edge deflection distribution was designed within EUROSUP for supersonic and transonic, as well as for low-speed conditions in compliance with preliminary design constraints given by European airframe industry. The constraint design target for supersonic and transonic cruise were numerically well met within the EUROSUP project. The objective at low speed was to design the distribution of leading edge deflection to achieve fully attached flow. Only approximate methods (i.e. semi-empirical and inviscid) were used for the low speed design within EUROSUP. Extensive experiments 10 11 confirmed that the EUROSUP design objectives were reached in supersonic and exceeded in transonic flows. At low-speeds attached flow for a large range of angle of attack (AoA) over the main part of the wing was found. Small areas with closed separation bubbles were found for increasing AoA. Furthermore some small vortices were found to develop from the undeflected wing stub by analysing the oilflow pictures. Just below the design AoA the experi-


AIAA 2001-2405 ment revealed a sudden topological change leading to separated flow showing complex vortical surface streamline patterns and significantly increasing drag values with AoA. This observed sudden change in flow topology and associated drag increase is quite challenging to predict with CFD. In a design cycle combinations of parameters leading to configurations with increasing drag (at the design point) need to be prevented. But the designer has to be sure that the numerical tool he is using is able to detect and discard these types of configuration. The validation process for the EPISTLE project is therefore based on this challenging EUROSUP lowspeed design. Three AoA were chosen to be calculated at M=0.25, Re~7*106 by all partners to cover attached, transient and separated flow conditions. Results for attached and transient flow are compared to experiments and are required to fulfill accuracy standards based on European industry input (∆α±0.5°; ∆cD±17dc [1 dc= 1 drag count= 0.0001]; ∆cM±0.0035). Furthermore a recognizable drag increase due to vortical flow is required to be predicted for separated flow. An additional requirement was to detect separated flow by analyzing pressure distributions at three representative spanwise stations for which experimental results exist.

A structured mesh refinement study was performed by BAE. Special attention was paid to find the necessary number of cells in longitudinal and spanwise direction to resolve the measured behavior of the configuration with sufficient certainty. The mesh sizes ranged from 0.98*106 cells to about 4.4*106 cells. Finally NLR generated another basic mesh having about 1.5*106 cells and a fine mesh using about 5.2*106 cells. In all three directions the number of cells was increased by 50%. The incorporated O-O topology resulted in efficient meshing of the configuration. The current mesh sizes, local resolutions and naming conventions are tabled below for the meshes used for the grid density investigation.

Common mesh and additional meshes used A common mesh to be used by all partners using structured methods was generated and provided by DLR. The surface mesh of this medium density mesh having ~0.8*106 cells is depicted in Fig. 1. A fine mesh of ~6.4*106 cells is also used by DLR for mesh density investigations. Strong clustering of grid lines in order to resolve the wing-body junction, the wing leading edge, the hinge area of the high-lift device, the trailing edge and the wing tip are visible in Fig. 1. The detail highlights the wing fuselage junction area. Discretization of such SCT wings is quite demanding due to the extremely large range of radii of curvature, typical for thin wings having rounded and deflected leading edges as well as curved knuckle shapes. For different reasons not all partners used the common mesh in the delivered form. HUT generated a new structured volume mesh keeping the wing surface discretization. INTA, using the unstructured commercial FLUENT code, had to generate a new mesh due to their different requirements. First results obtained on a fully unstructured mesh indicated the need for a hybrid mesh having about 20 layers of prismatic cells around the solid walls of the configuration. Grid refinement beyond 1.9*106 cells was not possible due to machine memory limits.

Partner

mesh name

mesh size [in 106 cells]

Y+

BAE

meshD

2.22

~0.5

“

meshE

3.77

~0.5

“

meshF

4.38

~0.5

DLR

c

0.099

~5

“

m

0.79

~2

“

f

6.35

~1

NLR

mO

1.54

~0.7

“

fO

5.2

~0.4

Table 2: Spatial discretization characteristics for the mesh density investigation. It should be mentioned that the spanwise resolution of the NLR mesh is not directly comparable to the other meshes, because the chosen mesh topology requires no grid clustering towards the fuselage to resolve the fuselage boundary-layer in the wing block. General results In Fig. 2 calculated and experimental lift coefficients are given over AoA. Indicated ∆’s and accuracy limits give good indication of what is reached. At the design AoA a slope discontinuity is visible in Fig. 2. The additional lift is generated by a quite rapidly formed vortex. In general the calculated results using different methods and meshes are close to each other and to the experiment for attached, transient and separated flow conditions. The calculated lift is overpredicted by most methods with less than 0.5° AoA. No lift underprediction is observed.


AIAA 2001-2405 The drag polars are compared in Fig. 3. The plotted lift range is identical to Fig. 2. The given drag accuracy limits of ±17dc was found to be challenging to be kept by all flow codes. For attached and transient flow it is found that the drag prediction by all partners is in very good agreement to the measured values and the accuracy limits are kept. As separated flow is calculated a trend is observed that the computations overpredict the drag values. The influence of vortical flow at the low-speed design condition is clearly visible in Fig. 4. The plotted lift range corresponds to the one of Fig. 2 and Fig. 3. As before, for attached and transient flow conditions all computations keep the required accuracy limits. While in the experiment the vortex induced pitching moment reduction is limited to a small range of lift variations, nearly all computations seem to smear this behavior. Pressure distributions for attached flow, M=0.25, α=6° are given in Fig. 5 for three spanwise locations: η=0.29, 0.52 and 0.71. Due to the size of the windtunnel model the pressure holes for each spanwise location had to be put separately for the upper and lower surface of the right and left wing. One pressure hole on both wing halfs lower side is used to indicate flow asymmetries. In addition, it was decided to have at midspan no lower surface pressure measurement holes. This was decided because of model stiffness and monetary reasons. For attached flow all computations are in excellent agreement to the experiments. The different computations are hardly distinguishable. The only marginal exception being the difference predicted by the INTA result at η=0.29 to be discussed later. At M=0.25, α=10°, Fig. 6, separated flow is measured and predicted. For all three sections the predicted pressure levels are in very good agreement on the pressure side. On the suction side, at η=0.29 the predicted suction level differences due to the local vortex suction peak at X/C~0.3 are relatively small although the computed vortex location differ in the partners results. At the midspan and the outboard section large areas of separated flows are visible. Here the differences between the experiment and the computations are clearly visible. Because this was expected, the previous mentioned separated flow indicators are now defined. The first indicator concerning this testcase is the evidence of vortical flow downstream of the hingeline at η=0.29. All but one contribution are predicting a vortex of varying location and strength. The second indicator is the evidence of vortical flow downstream of the hingeline at η=0.52. In the experiment it is hard to distinguish between the leading edge suction, the vortex on the deflected part of the wing and the knuckle induced separation. In the computations different forms of separations are predicted of varying

location and strength. But neither at this section nor further outboard is attached flow predicted. Due to the differences in vortex strength and location prediction at η=0.29 and 0.52, and due to the strong crossflow observed, the pressure distributions differences at the outboard section are amplified. It should be added that the emphasis of this validation task is on the prediction of attached flow and flow with separation onset. Regarding fully separated flow the exact prediction of all flow details is not required but only the detection of the flow status. Flow Topology The experimental data given in Fig. 2-Fig. 4 showed a lift, drag and pitching moment change taking place nearly at the configurations design lift as explained earlier. The corresponding oilflow picture given in Fig. 7 shows the configurations planform and attached flow on the outer wings. The oilflow pattern of the inner wing shows some crossflow, but no evidence of vortical flow was found by the analysis of this picture. Increasing the AoA shows some vortical flow structures on the inner wing, see Fig. 8, but still attached flow on the outer wings. This shows the demanding character of the flow to be predicted: the flow for this configuration is characterized by quite early vortex onset (α~8°) that is neither lift, drag nor pitching moment sensitive and a sudden change in vortex number, location and strength (α~10°) changing the aerodynamic characteristics significantly. Mesh Density Investigation Results In the following figures the variations of the aerodynamic coefficients with mesh density are given. The coefficients are plotted against 1/N2/3, were N denotes the number of mesh cells (see Table 2). Ideally this plot should demonstrate the codes second order (O2) spacial accuracy by showing a linear variation of the aerodynamic coefficients under investigation. This holds if the artificial dissipation is kept constant by neither changing the schemes dissipation nor changing the mesh inherent dissipation through changes of the cell aspect ratios. This is ideally obtained by doubling the cell numbers in all co-ordinate directions for each mesh refinement. To show mesh independence of the results requires that even the coarse meshes being used are able to resolve the dominant physical properties of the flow. The lift variation is given in Fig. 9. The scale is chosen to correpond to the allowed α uncertainty range of ∆α±0.5°. For attached flow, at 6° AoA, the lift is predicted by all finer solutions in excellent agreement to each other. Here it is proven that mesh converged solutions are obtained for the lift.


AIAA 2001-2405 In Fig. 10 the pressure drag is given. At 6° AoA the results are in very good agreement with differences of about only 3 dcs for meshes with more than 1*106 cells (1/N2/3< 0.0001). Coarser meshes show a recognizable pressure drag increase. This is mainly due to a reduced suction force prediction as has been found analysing the sequence of DLR solutions. In Fig. 11 the plotted pressure drag is reduced for the amount of induced drag due to (assumed) elliptical lift distribution. This quantity will be called CDP-CDi (elliptic). The graph shows for attached flow excellent agreement between all computations. For fully separated flow (α>10°) the solutions on the different meshes give estimates in very good agreement to each other. The obtained scatter is below 10dc, except for the DLR k-ω solution on the common mesh. Please note that the net drag prediction around 8°-9° shows the largest differences. This is mainly due to a combination of spatial mesh resolution differences and turbulence model triggered separation behaviour that leads to strong differences regarding the drag prediction during separation onset.

difference is small having in mind the different flow solvers and meshes used. Please note that the finer mesh used has about a factor of 5.5 more cells. At attached flow (α=6°) the overall drag scatter is within the allowed range. At separated flow the computational results are again in reasonable agreement with a bandwidth of about 50 dcs. All results overpredict the experiments. The separation onset behaviour of both the one and two equation TM results show no rapid drag change similar to the BLDS results. A very gradual increase with AoA is obtained instead. The friction drag predicted by the various turbulence models and implementations is given in Fig. 13. The bandwidth of friction drag level is large. Only marginal friction drag variation with AoA is observed. The lowest friction value is calculated by the SA model. This low friction drag is consistent with early separation of the boundary-layer and high drag levels due to vortical flow in the previous picture. A part of the drag difference between both BLDS solutions is visible here to be caused by the friction drag. The pitching moment over AoA is given in Fig. 14. Compared to the experiments the SA results are quite far off, beyond the requested accuracy requirements. Both BLDS results show a trend towards a sudden change in flow type and pitching moment. This trend is overpredicted and happening at a lower AoA than in the experiment as mentioned earlier. The two equation TM results show (not all displayed) quite similar behaviour. A nearly linear pitching moment dependency is observed, except for the NLR result. On this fine mesh the pitching moment reduction is predicted also at a little lower AoA, comparable to the BLDS computations. The reason for this sudden pitching moment and drag change will be described next. Total pressure loss contours in a mesh plane perpendicular at the wing trailing edge (compare Fig. 1) are given in the following two figures. Total pressure contours are used here to depict vortex structures. The results are extracted from DLR computations on the common mesh. Two other results are available (BAE: BLDS and NLR: k-ω) but not shown here. The four pictures compare the vortex development with AoA predicted by the DLR solutions using the BLDS and the k-ω TM. The two pictures in Fig. 15 show results for the BLDS TM, the next graphs in Fig. 16 display k-ω results. In all pictures the mesh in a plane perpendicular to the wing trailing edge is drawn to show the spatial resolution of the vortices obtained on the common mesh. Please note that the lower part of the mesh (perpendicular to the wings lower surface) is not plotted. Results for 9° AoA are given in both top rows,

Turbulence Model Influence The influence of turbulence models (TM) on the calculated flow topology and corresponding aerodynamic coefficients in general is known to be quite large. As long as we analyze attached flow these differences predominantly effect the friction drag. Here we concentrate on the prediction of separation onset and are therefore dealing with the prediction of the boundary-layer in 3D Navier-Stokes computations, their tendency to separate, to form vortices and to generate associated vortex drag. Determining the TM effects for these complicated flow is not trivial nevertheless three major trends are found. First the global aerodynamic coefficients will be compared. Because the lift is less sensitive to numerical and physical influences, the presentation will be for the pitching moment and drag. Fig. 12 shows the drag over AoA obtained on the finest meshes used with the TMs given in Table 1. Results for an algebraic (BLDS), the one-equation Spalart Allmaras (SA) and two variants of the widely used k-ω turbulence model are given. Closest agreement to the experiments is obtained with the BLDS TM. Especially the flow type change from attached to separated flow is predicted by the BLDS TM in good agreement to the experiment. The vortex induced drag increment is found to be overpredicted by the BLDS TM, at a slightly lower AoA. A nearly constant drag difference of about 20 dcs for α>6° between both BLDS results exists until separation onset and the associated drag increase. This


AIAA 2001-2405 the bottom rows show results for 10° AoA. At α=9° AoA the BLDS TM (Fig. 15) predicts only one wing tip vortex. At 10° AoA two additional vortices of greater strength in addition to the tip vortex are calculated. In the BAE BLDS results (not shown) weak vortices originating from the wing apex are calculated even at α=6° AoA. Results for the k-ω solution are depicted in Fig. 16. Here a pronounced double vortex structure is visible at 9° AoA. At 10° AoA the vortex strength and location has altered but the overall structure is not changed. The analysis of the NLR results on the fine mesh (not shown) reveals that similar to the k-ω result in Fig. 16 a double vortex structure exists at the lower AoA. Increasing the AoA in the NLR k-ω results increases the vortex strength more than in the depicted DLR solution. Please note that the inner vortex (z~0.18) has similar strength in the computations of both TMs, whereas the other big outboard vortex changes more in strength with AoA in the BLDS computation than in the k-ω result. It becomes clear from these figures that the BLDS TM predicts a sudden vortex onset between 9° and 10° AoA which is significantly different to the k-ω prediction. For the BLDS results this explains the aerodynamic coefficient changes obtained similar to the experiment in Fig. 14. The vortex strength in the BAE solutions (not shown) increases also rapid with AoA. The rapid vortex strength increase or vortex onset is the reason for the drag and pitching moment trend found in Fig. 12 and Fig. 14 to be similar between the BLDS computations. In case of the k-ω results it is believed that the spatial resolution of a developing vortex has a more pronounced influence than for the algebraic TM. This could be the reason that the sudden change of the aerodynamic coefficients is suppressed on the common mesh and visible with the fine NLR mesh. Furthermore these figures shows that the spatial resolution and mesh clustering near the wing is adjusted to capture the vortices with both TMs. The previous mentioned trends regarding the TMs summarize as follows: the large differences in friction drag is a severe source of uncertainty. For engineering purposes the algebraic BLDS seems to be quite well usable until vortex onset. Decreasing prediction accuracy is observed with vortex onset. For more accurate predictions higher order TMs requiring finer meshes seem to give more physical settled answers especially regarding the flow topology and the local flow physics also during vortex onset. Unfortunately this did not lead to significantly better aerodynamic coefficient prediction.

Structured / Unstructured Mesh Influence INTA performed calculations with the commercial CFD tool FLUENT. Several meshes had to be generated to derive sufficient spatial resolution of all geometric features of this configuration. Finally a hybrid mesh with a structured prismatic layer of twenty cells to enable an efficient boundary layer resolution was used having 1.9*106 cells. The turbulence model employed is the two-equations k-ε turbulence model. The first mesh off-wall distance resulted in Y+-values of around 30 and could not be reduced further for technical reasons. The global and local mesh resolution of the hybrid (left hand side) and the common mesh is compared in Fig. 17. In both pictures the same part of the deflected leading edge flap near the fuselage is displayed. In the common mesh the streamwise clustering to resolve the knuckle and the leading edge is much more pronounced than in the unstructured mesh. On the other hand side the spanwise resolution is better in the hybrid mesh. The local geometric detail of the leading edge resolution gets lost to some degree in the unstructured mesh. It is worth mentioning that the local pressure peaks at the LE and at the flap knuckle are lower in the unstructured solution than in the structured ones. So the local mesh resolution deficits are reflected in the pressure distribution. A comparison to the experiment and other fine mesh computations at 9° AoA is given in Fig. 18. The INTA solution predicts very destinct vortices at all three wing sections. Similar vortices exist also in the DLR solution, less strong and spread a little more, but in both cases the drag increasing effect is similar (not shown). The reason for higher pressure drag in the INTA solution (not shown) maybe extracted from Fig. 5 and Fig. 13. Even for attached flow the unstructured solution predicts a vortex at η=0.29 in Fig. 5. Having in mind the higher Y+-values (compare Table 2) the lower friction drag in Fig. 13 may reflect the trend of the boundarylayer for early separation. This is supported by Fig. 18 and Fig. 6. Both figures show very pronounced vortex induced suction peaks in the unstructured solution. Another reason for quite high pressure drag levels is the coarse leading edge resolution in the unstructured mesh preventing the calculation of a realistic leading edge suction level. Two main findings from this can be formulated regarding unstructured meshes: First a good near wall boundary-layer resolution is necessary if correct flow features are to be obtained, especially if flow separation is expected. Due to the nature of unstructured meshes this can only be obtained by using a hybrid mesh with a reasonable number of cells in


AIAA 2001-2405 the structured layer and with appropiate low Y+-values. Secondly the global aerodynamic coefficients do not seem to be very sensitive to local resolution differences. This holds for all solutions presented here in general.

It is found that increased mesh density compared to the common mesh is needed for accurate separation onset prediction. The algebraic BLDS turbulence model was found to be a good engineering tool and able to predict high-lift flows well into vortex onset. However at high AoA the BLDS appears to overpredict the vortex effect with respect to the experiment. A trend towards more physical settled answeres is obtained with higher order turbulence models. Two equation turbulence models seem to give more accurate flow topology predictions as long as appropriate higher mesh densities (compared to BLDS computations) are used. Significantly better aerodynamic coefficient prediction are not obtained with two equation turbulence models. Results using various turbulence models showed a severe large bandwidth of friction drag prediction.

High-Lift without a High-Lift System Here the computed low-speed flow around the supersonic wing (Swing) without leading edge deflections will be given and compared to the experiment. Flow solutions are obtained on a mesh having similar resolution to the common mesh. The k-ω turbulence model is used. The computed lift overpredicts the measured lift slightly, by about ∆CL=0.007. This corresponds to a small AoA inaccuracy of about 0.1°. The computed drag is in very good agreement to the experiments. The predictions are about eight dcs lower than the experiment. The pitching moment shows a small systematic offset between computation and experiment. The measured and computed pressure distributions are compared at three locations (η=29, 52 and 71%) in Fig. 19. The general agreement is good. Some deviation is found concerning the vortex location. Nevertheless, the computed vortex induced suction peak levels are predicted in quite good agreement, indicating that the main physical features of this flow are numerically captured. In the pressure distribution at η=52% in Fig. 19, two different vortex induced suction peeks are detectable in the CFD solution. The experiment shows a nearly plateau-like pressure distribution on a different pressure level. The pressure distribution differences are the reason for the observed pitching moment differences. Mainly the pressures on the outer wing seems underpredicted leading to higher CM-values than in the experiment (not shown here). The aerodynamic efficiency CL/CD is given over AoA in Fig. 20. The numerical performance estimation of the Swing is in very good agreement to the experiment. Some inaccuracies regarding the vortex location prediction for this fully separated flow are obtained. This influences mainly the predicted pressure distributions and the resulting pitching moments. But the numerically predicted drag in very good agreement to the experiment.

Outlook The capability to predict low-speed high-lift flow fields around SCT configurations having deflected leading edges with industry relevant drag accuracy levels even for near off-design cases is shown. This result encourages us to continue the promising EPISTLE research work and will enable the development a design methodology based on modern European Navier-Stokes solvers. References 1. Lovell, D. A., Aerodynamic Research to Support a Second Generation Supersonic Transport Aircraft the EUROSUP Project, Eccomas 98, 1998. 2. Lovell, D.A., European research of wave and lift dependant drag for supersonic transport aircraft, AIAA Paper No. 99-3100, 1999. 3. Degani, D., Schiff, L., Computations of Turbulent Supersonic Flows around Pointed Bodies having Crossflow Separation, Journal of Computational Physics 66, pp. 173-196, 1986. 4. Spalart, P. R., Allmaras, S. R., One-Equation Turbulence Model for Aerodynamic Flows, AIAA Paper No. 92-439, 1992.

Conclusion The validation of the involved viscous European CFD codes to predict low-speed flow fields around SCT configurations with high-lift systems to industry relevant drag accuracy levels is shown. The ability to predict flows to these accuracy levels is shown not to be restricted to the usage of one mesh or structured methods and was reached using a variety of turbulence models.

5. Chien, K.-Y., Prediction of Channel and BoundaryLayer Flows with a Low-Reynolds-Number Turbulence Model, AIAA-Journal, Vol. 20, 1982, pp. 3338. 6. Smith, B. R., The k-l Turbulence Model and Wall Layer Model for Compressible Flows, AIAA Paper No. 90-1259, 1990. 8

American Institute of Aeronautics and Astronautics

AIAA 2001-2405 7. Wilcox, D.C., Reassessment of the Scale-Determining Equation for Advanced Turbulence Models, AIAA-Journal, Vol. 26, 1988, pp. 1299-1310.

CL

8. Menter, F. R., Two-Equation Eddy Viscosity Turbulence Modes for Engineering Applications, AIAA Journal, Vol. 32, 1994, pp. 1598-1605.

∆CL=0.1

9. Hellsten, A., Some Improvements in Menter’s k - ω SST Turbulence Model, AIAA-Paper No. 98-2554, 1998.

Accuracy limits

∆α=2°

10. Elsenaar, A., Windtunnel test of the EUROSUP ‘SCT’ configuration, EUROSUP/NLR/T017/1, 1998.

M=0.25; Exp.; Lwing; St3 BAe; meshD; BLDS; Lwing DLR; f; k-ω; Lwing HUT; m*; k-ωrcsst; Lwing INTA; H1.9; k-ε; Lwing NLR; m; k-ω; Lwing ONERA; m; k-l; Lwing; LAD

α[°]

11. Rohne, P. B., Hoolhorst, A., Goossens, J. D., Data report on low speed measurements in the HST for the EPISTLE project (test number 0008, model 1:80 EUROSUP ‘SCT’), EPISTLE-12-NLR-002, 2000.

Fig. 2

Calculated and experimental lift over angle of attack for the ESCT configuration at M=0.25.

CL

Figures ∆CL=0.1 M=0.25; Exp.; Lwing; St3 BAe; meshD; BLDS; Lwing DLR; f; k-ω; Lwing HUT; m*; k-ωrcsst; Lwing INTA; H1.9; k-ε; Lwing NLR; m; k-ω; Lwing ONERA; m; k-l; Lwing; LAD

Accuracy limits

∆CD=0.01

CD Fig. 3

Fig. 1

Calculated and experimental drag polar for the ESCT configuration at M=0.25.

CM

SCT geometry under investigation having deflected leading edges; common medium mesh.

M=0.25; Exp.; Lwing; St3 BAe; meshD; BLDS; Lwing DLR; f; k-ω; Lwing HUT; m*; k-ωrcsst; Lwing INTA; H1.9; k-ε; Lwing NLR; m; k-ω; Lwing ONERA; m; k-l; Lwing; LAD

Accuracy limits

∆CM=0.01 ∆CL=0.1

CL Fig. 4

Calculated and experimental pitching moments for the ESCT configuration at M=0.25.


AIAA 2001-2405 Exp.; η=0.29 BAe; meshD; BLDS; η=0.29 DERA; m-; η=0.29 DLR; m; k-ω; η=0.29 * HUT; m ; k-ωsst; η=0.29 INTA; H1.9; k-ε; η=0.29 NLR; m; k-ω; η=0.29

-1.2

ONERA; m; k-l; η=0.29

CP -1.2

DLR; m; k-ω; η=0.52 * HUT; m ; k-ωsst; η=0.52 INTA; H1.9; k-ε; η=0.52 NLR; m; k-ω; η=0.52 ONERA; m; k-l; η=0.52

CP -1.2

-0.9

-0.9

-0.9

-0.6

-0.6

-0.6

-0.3

-0.3

-0.3

0.0

0.0

0.0

0.3

0.3

0.3

0.0

0.5

Fig. 5

X/C 1.0

0.0

-1.2

X/C 1.0

0.0


CP -1.2

DLR; m; k-ω; η=0.52 * HUT; m ; k-ωsst; η=0.52 INTA; H1.9; k-ε; η=0.52 NLR; m; k-ω; η=0.52

CP -1.2

ONERA; m; k-l; η=0.52

-0.9

-0.9

-0.6

-0.6

-0.6

-0.3

-0.3

-0.3

0.0

0.0

0.0

0.3

0.3

0.3

Fig. 6

X/C 1.0

0.5

X/C 1.0

0.0

0.5

X/C 1.0

0.0

Oil flow pattern at M=0.25, α=8°.

Fig. 8

Exp.; η=0.71 BAe; meshD; η=0.71 DERA; m-; k-ω; η=0.71


-0.9

0.0

0.5

CP distribution at M=0.25; α=6° at η=0.29, 0.52 and 0.71.


CP

0.5


CL


BAe; α=6°; BLDS DLR; α=6°; k-ω NLR; α=6°; k-ω

∆CL=0.01

CP

Exp.; η=0.71 BAe; meshD; BLDS; η= DERA; m-; η=0.71

Exp.; η=0.52 BAe; meshD; BLDS; η=0.52 DERA; m-; η=0.52

0.5

X/C 1.0

0

CP distribution for M=0.25; α=10° at η=0.29, η=0.52, η=0.71.

Fig. 9

0.0002

1/N2/3

0.0004

Variation of the lift with mesh density for different solutions; M=0.25; α=6°.

∆CDP=0.001

CDP

BAe; α=6°; BLDS DLR; α=6°; k-ω NLR; α=6°; k-ω

0

Fig. 7

Oil flow pattern at M=0.25, α=6°.

0.0002

1/N2/3

0.0004

Fig. 10 Variation of the pressure drag with mesh density for different solutions; M=0.25; α=6°. 10

American Institute of Aeronautics and Astronautics

CDp-CDi(elliptic)

AIAA 2001-2405

CM

∆CM=0.005

BAe; meshD; BLDS BAe; meshE; BLDS BAe; meshF; BLDS DLR; m; k-ω DLR; f; k-ω NLR; mO; k-ω NLR; fO; k-ω

20 dc

M=0.25; Exp.; St3 BAe; meshF; BLDS DLR; f; k-ω DLR; m; SA DLR; m; BL-DS NLR; fO; k-ω

6

8

10

α [°]

6

12

8

9

α [°]

10

Fig. 14 Pitching moment over AoA for various turbulence models and meshes.

Fig. 11 Variation of the net drag with AoA and mesh density for different solutions.

CD

7

0.05

BLDS; α=10°

0.15

M=0.25; Exp.; St3 BAe; meshF; BLDS DLR; f; k-ω DLR; m; SA DLR; m; BL-DS NLR; fO; k-ω

0.038

0

0.1 0.05

0.1

0.15

y 50 dc

0.2

0.25

z

0.05

0.044

0.059 0.032

0

0.05

6

7

8

9

α [°]

0.1

k-ω α=10

0.05 BAe; meshF; BLDS DLR; f; k-ω DLR; m; SA DLR; m; BLDS HUT; m*; k-ωsst INTA; H1.9; k-ε ONERA; m; k-l; LAD NLR; fO; k-ω

0.15

0.25

0.049

0.041 0.035

0

0.1 0.05

0.1

0.15

0.2

0.25

z

y

10 dc

0.2

Fig. 15 Total pressure loss contours at the wing trailing edge (mesh underlaying) at M=0.25; α=9° (top) and α=10° (bottom), common mesh; BLDS.

Fig. 12 Configurations drag over AoA; different turbulence models and meshes at M=0.25.

CDf

0.15

z

10

0.05

0.046 0.028

0.043 0

6

8

10

α [°]

0.05

12

0.1

0.15

0.2

0.25

z

Fig. 16 Total pressure loss contours at the wing trailing edge (mesh underlaying) at M=0.25; α=9° (top) and α=10° (bottom), common mesh; k-ω.

Fig. 13 Friction drag over AoA for various turbulence models and meshes.


AIAA 2001-2405

∆CL/CD=2

CL/CD

M=0.25; Exp.; Swing; St1 DLR; m; k-ω; Swing

4

Fig. 17 3D view of the unstructured surface mesh (deflected flap) used by INTA compared to the common mesh. Exp.; η=0.29 BAe; gridF; BLDS; η=0.29 DLR; f; k-ω; η=0.29

CP

INTA; H1.9; k-ε; η=0.29

CP

INTA; H1.9; k-ε; η=0.52

CP

-1.2

-1.2

-1.2

-0.9

-0.9

-0.9

-0.6

-0.6

-0.6

-0.3

-0.3

-0.3

0.0

0.0

0.0

0.3

0.3

0.3

0.0

0.5

X/C 1.0

0.0

0.5

X/C 1.0

0.0

INTA; H1.9; k-ε; η=0.71

0.5

X/C 1.0

Fig. 18 Pressure distributions on structured (BAE, DLR) and unstructured (INTA) meshes at M=0.25; α=9°. M=0.25; Exp. Swing η=0.29

M=0.25; Exp. Swing η=0.52

DLR; Swing; α=9°; η=0.29

DLR; Swing; α=9°; η=0.52

M=0.25; Exp. Swing η=0.7 DLR; Swing; α=9°; η=0.71

CP

CP

CP

-1.2

-1.2

-1.2

-0.9

-0.9

-0.9

-0.6

-0.6

-0.6

-0.3

-0.3

-0.3

0.0

0.0

0.0

0.3

0.3

0.3

0.0

0.5

X/C 1.0

0.0

0.5

X/C 1.0

0.0

0.5

8

10

α[°]

Fig. 20 Aerodynamic efficiency of the ESCT configuration (Swing), compared to numerical results at M=0.25.

Exp.; η=0.71 BAe; gridF; BLDS; η=0.7 DLR; f; k-ω; η=0.71

Exp.; η=0.52 BAe; gridF; BLDS; η=0.52 DLR; f; k-ω; η=0.52

6

X/C 1.0

Fig. 19 Calculated and experimental pressure distributions for the ESCT configuration (Swing) at M=0.25; α=9°. 12 American Institute of Aeronautics and Astronautics