Boundary layer interacting with wakes under adverse pressure gradient - NLR 7301 high lift configuration

Underlying Flow Regime 3-01 © copyright ERCOFTAC 2004

Test Case

Brief description of the study test case

Figure 1 shows a picture of the NLR 7301 configuration which consists of a supercritical wing, followed by a flap at an angle of 20°. The wing upper surface up to 94.36% chord and the wing lower surface up to 60% chord are the same as the NLR 7301 aerofoil profile. The shape of the shroud from 60% chord to the trailing edge was designed such that no flow separation occurs in this region. It should be remarked that the shroud shape does not permit retraction of the flap. The coordinates of the wing and of the flap can be found in [8], and are available on the CD-ROM of ECARP [10]. SAAB military aircraft kindly made multi block structured grids available, which can be downloaded here. In the experiments only one flap deflection angle (δ_F = 20°) was considered. Two flap positions were used for which the overlap (difference of the x-coordinates of the wing trailing edge and flap leading edge) was the same, viz 5.3% chord. For the first geometry the gap width was set at 2.6% chord, while for the second it was set at 1.3% chord.

The free stream Mach number for the experiment was 0.185, and the Reynolds number based on the wing chord was 2.51 10⁶. The chord length was 0.57 m. The first geometry was run at an incidence angle of 13.1° whilst the second was run at an angle of 6°. Transition locations (x/chord) are given in Table 1. It should be noted that the flap lower surface was fully laminar and thus no transition location is specified.

Table 1: Transition locations observed from wind tunnel tests.
Element	gap = 2.6%, α = 13.1°	gap = 1.3%, α = 6°
Wing upper side	0.024 - 0.035	0.027 - 0.040
Wing lower side	0.710 - 0.740	0.635 - 0.675
Flap upper side	1.065 - 1.080	1.063 - 1.081

Figure 1: NLR 7301 wing-flap configuration with pressure and velocity measurement stations (from Brandsma et al, High-Lift Investigations for the NLR7301 Flapped Airfoil, in: Haase et al. (editors), EUROVAL - An European Initiative on Validation of CFD codes. Notes on Numerical Fluid Mechanics, Vol. 42, 1992, Vieweg)

The measured data were reduced to give:

aerodynamic coefficients (C_D, C_L)

boundary layer and near wake velocity profiles at stations indicated in Figure 1 with a number in a circle

Reynolds stresses

wall shear stresses (skin friction coefficient)

Cp values at stations indicated in Figure 1 by a black dot.

Test Case Experiments

Measurements were made in the NLR 3x2 m.wind tunnel in Amsterdam in 1979 [8], and in the new NLR LST 3x2.25m wind tunnel in the North East polder in 1989 [12]. Both facilities are closed, continuously operating facilities at atmospheric conditions. The standard NLR technique for two-dimensional high-lift tests was used [8]. Boundary layer control was applied on turn tables to avoid early flow separation at the model tunnel-wall junctions. Blowing slots were used in front of the wing nose and flap nose, so that it was possible to simulate a 2D flow up to maximum lift.

Measured data consisted of:

Surface pressure data and far wake traverses, from which the aerodynamic forces and the pitching moment were obtained. The positions of the surface pressure holes are indicated in Figure 1 as black dots. The measuring plane was situated at mid span. The drag was obtained from wake traverses at about one chord distance downstream of the model trailing edge. Since it was found that the flow was not completely 2 dimensional at this location, the wake traverses were carried out at a large number of span wise locations so that a mean value could be computed.

Boundary layer and near-wake traverses. Boundary layer probes were used to measure the boundary layer on the wing and flap at the locations indicated by a circle with a number inside. The probes could be traversed with an accuracy of 0.02 mm. The position of the probe relative to the model was obtained by traversing the probe towards the model until electric contact was obtained (under operation conditions). However, some uncertainty exists due to probe vibrations. Two kind of probes were used, a total head tube of 0.3 mm outer diameter and a static-pressure tube of 1.1 mm outer diameter.

Wall shear stress measurements. The razor blade technique was used to measure the wall shear stress at surface pressure locations. Other techniques were used too, but were considered less reliable.

Flow visualization techniques. Oil flow visualization was used to indicate the presence of separated flow regions. The sublimation technique was used to detect the position of transition lines.

The experiments made in 1989 [12] concerned the geometry with 2.6% gap and α = 13.1° and consisted of the measurement of the Reynolds stress profiles using the hot-wire technique at stations 8, 12, 13, 14 and 16 (indicated by a circle with a number). However, these data are not publicly available.

Attention was given to ensure a 2 Dimensional flow, but in [8], it is mentioned that the 2- Dimensionality of the flow was not fully satisfied in the wake. The report describing the 1979 experiments includes the corrections used in the data reduction:

a solid blockage correction ΔU_∞/U_∞ = 0.004

a wake blockage correction: ΔU_∞/U_∞ = 0.0475 C_d

corrections due to lift: ΔC_l = -0.0148 (C_l + 2C_m)

ΔC_m = -0.00185 C_l

with C_d the drag coefficient obtained from surface-pressure integration, and C_l and C_m the lift and pitch-up moment coefficients respectively.

Boundary layer and near-wake properties were corrected for compressibility effects. As mentioned, the experiments were repeated in 1989 in the new NLR LST wind tunnel, and these measurements have since been used in conjunction with the 1979 results. Surface pressure and boundary layer measurements made in the two wind tunnels were in good agreement [28].

Information about the quality of the data and the accuracy of the measurements is found in [28], and is summarized as follows:

Flow Quality:

NLR LST 3x2 m: variation of mean velocity across test section < 0.5%

free-stream turbulence level < 0.2%

NLR LST 3x2.25 m variation of mean velocity across test section < 0.2%

free-stream turbulence level < 0.04%

Model Position: Accuracy of geometrical angle of attack of main wing: ± 0.05°

Model deformation: flap gap decreases with 0.2% chord

flap angle with 0.2-0.3°

Pressure data: estimated accuracy of pressure coefficients ± 0.5%

Forces and moments:lift coefficient ± 0.01

pitching moment coefficient ± 0.005

drag coefficient ± 2%

Skin friction: estimated accuracy C_f ± 10%

Mean velocities: accuracy ± 2%

Turbulence quantities: accuracy ± 15%

Since the NLR7301 test case was specifically designed for CFD code validation, it can be assumed that the inflow conditions (Mach number, Angle-of-Attack and Reynolds number) are known with sufficient precision. Transition locations were obtained from wind tunnel measurements, and are given in Table 1.

The experimental data of the NLR 7301 are available on the CD-ROM of the ECARP book [10].

No information is available on consistency of measurements of different quantities and on checks performed for global conservation of physical quantities.

In AGARD AR-303 [28], it is mentioned that “Tunnel wall interference effects are small, and consequently the data can be used for ‘free air’ calculations”.

CFD Methods

The NLR7301 test case was used in 4 different projects a GARTEUR Action Group on inviscid/viscous interaction methods, the EUROVAL project, the ECARP project and the FLOWNET data base project. No details on the GARTEUR Action Group project are known, apart from a reference to the project results in [9].

Table 2 provides a summary of the available CFD calculations from the EUROVAL and ECARP projects. Information on the numerical methods used is available for the results obtained in the EUROVAL project, but unfortunately, this is not the case for ECARP. However, since several partners of EUROVAL continued in the ECARP project, it is probably reasonable to assume that these partners used the same numerical method.

For the EUROVAL project, grids were made by three of the partners, and used by the others. These grids yielded a y+ value of near unity on the surfaces. The far field boundary for these grids was placed at approximately 10 chords from the airfoil. The BAe grid had 28288 cells, the CFD Norway grid 26172 cells, and the Dornier grid (which was flow adapted) had 49392 cells. Several partners used both the BAe and CFD Norway grids, but no firm conclusion on which grid predicted better results could be drawn. The Dornier grid was used only by Dornier, and no comparison with results using another grid was made. The results obtained with the Dornier grid were not fully satisfactorily, in particular in the wake region [9].

For the ECARP project, a mandatory (structured multi block) mesh was used to eliminate differences due to the grid. This grid had 36208 cells, with a y+ at the wall of unity, and with 20 to 30 cells spanning the wall adjacent layer extending to y+=100. The far field boundary for this grid was placed at about 10 to 15 chords from the airfoil. With the exception of CASA and ONERA, all ECARP participants used this grid which is available on the ECARP CD-ROM [10], and in the FLOWNET data base. Refined grids (up to 4 times) were made available for grid dependency studies.

Partner

Method

Turbulence Model

BAe (EUROVAL)

Cell Centered Finite Volume method, artificial dissipation, 4 stage Runge Kutta, residual smoothing and multi grid

Chien k-ε [19]

SAAB

(EUROVAL)

Cell Centered Finite Volume method, artificial dissipation, 5 stage Runge Kutta, multi grid

Baldwin-Lomax [15]

Jones-Launder k-ε + Wolfshtein 1 eq model near the wall [17] [26]

HUT

(EUROVAL)

Thin layer Navier Stokes, Cell Centered Finite Volume method, Van Leer splitting, implicit relaxation and multi grid

Cebeci – Smith [23]

Dornier

(EUROVAL)

Cell Centered Finite Volume method, artificial dissipation, 3-stage Runge Kutta, residual smoothing and multi grid

Baldwin-Lomax [15]

Lam - Bremhorst k-ε [20]

CERFACS

(EUROVAL)

Cell Centered Finite Volume method, Steger & Warming flux splitting, Explicit MacCormak scheme

Baldwin-Lomax [15]

Launder-Sharma k-ε [18]

CFD-Norway

(EUROVAL)

Thin layer Navier Stokes, Cell Centered Finite Volume method, artificial dissipation, 3-stage Runge Kutta

Baldwin-Lomax [15]

Chien k-ε [19]

BAe

(ECARP)

Structured

Chien+Wolfshtein [19], [26]

Kalitzin-Gould [24]

CASA

(ECARP)

Unstructured

Granville/Baldwin Lomax [22]

Chien k-ε [15]

CFD Norway

(ECARP)

Structured

Baldwin-Lomax [15]

Launder-Sharma k-ε [18]

Chien k-ε [19]

Chien k-ε + length scale correction

Dornier/DASA LM (ECARP)

Structured

Baldwin-Lomax [15]

Lam - Bremhorst k-ε [20]

HUT

(ECARP)

Structured

Cebeci – Smith [23]

Menter k-ω (+ SST) [21]

KTH

(ECARP)

Structured

Baldwin-Lomax [15]

Speziale k-τ [25]

ONERA

(ECARP)

Viscid-Inviscid method

Le Balleur

SAAB

(ECARP)

Structured

Jones-Launder k-ε + Wolfshtein 1 eq model near the wall [17] [26]

idem + SST

Shih-Lumley-Zhu-Wolfshtein model [27]

Table 2: CFD results EUROVAL and ECARP project

Little is known about the accuracy of the CFD calculations and, apart from grid refinement studies, few sensitivity tests were carried out. Two partners in the ECARP project used the refined grid, and based on the results of one of them, it was concluded that a satisfactory grid convergence was obtained on the mandatory grid [10].

Far field boundary conditions were employed on the outer boundary of the grid. The exact formulation used is, however, unknown. Larsson [13] of SAAB Aerospace investigated the use of a far field velocity correction based on the circulation obtained from the computed lift, and showed that this correction improved significantly the prediction of the drag coefficient.

Besides the above mentioned projects, the NLR 7301 was used in several other studies. An overview is given in the paper by Rumsey et al. [2], and the study by Godin et al. [14] is one example. In this study, a grid of 180,000 grid points was used, and calculations were made with the Spalart Allmaras [16] and Menter SST turbulence models [21]. Results of this study are included in Chapter 6.

Contributors: Jan Vos - CFS Engineering SA