Latest revision as of 12:26, 12 February 2017

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

Underlying Flow Regime 3-01 © copyright ERCOFTAC 2004

Evaluation

Comparison of CFD calculations with Experiments

Both the EUROVAL and ECARP proceedings [9, 10] include a discussion on the comparison of experimental and CFD results. The figures comparing predictions with measurements, drawn from the ECARP project, are included in Appendix A. More recent results obtained by Godin et al. [14] are included in the discussion, and comparison figures are shown in Appendix B.

Aerodynamic Coefficients

Table 3: Comparison experimental and computed lift and drag coefficients.
Partner	Project	Turbulence Model	2.6% gap α = 13.1°		1.3% gap α = 6°
Partner	Project	Turbulence Model	CL	CD	CL	CD
Experiment	EUROVAL		3.141	0.0445	2.416	0.0229
BAE	EUROVAL	Euler Chien k-ε	3.657 3.165	0.0070 0.0604	2.680 2.457	0.0030 0.0323
SAAB	EUROVAL	Jones Launder k-&epsilon + Wolfshtein Baldwin-Lomax	3.159	0.0658	2.431	0.0349
HUT	EUROVAL	Cebeci-Smith	3.193	0.0613	2.417	0.0343
CERFACS	EUROVAL	Baldwin-Lomax Jones-Launder k-ε	2.949 2.801	0.1060 0.1380	2.300 2.242	0.0360 0.0380
Dornier	EUROVAL	Baldwin Lomax Lam-Bremhorst k-ε	3.315 3.262	0.0911 0.0758	2.359 2.342	0.0381 0.0375
CFD Norway	EUROVAL	Baldwin Lomax Chien k-ε	3.148 3.064	0.0600 0.0680
BAE	ECARP	Chien k-ε +Wolfshtein Kalitzin-Gould	3.270 3.232	0.04346 0.04640	2.482 2.474	0.0231 0.0236
CASA	ECARP	Ganville/Baldwin Lomax	3.23	0.04	2.88	0.02
CFD Norway	ECARP	Baldwin-Lomax Chien k-ε Chien k-ε + length scale correction	3.225 3.235 3.225	0.042 0.051 0.049	2.480 2.465 2.480	0.025 0.026 0.025
Dornier/DASA LM	ECARP	Lam - Bremhorst k-ε	3.225	0.0430	2.470	0.0225
HUT	ECARP	Menter k-ω (+ SST)	3.225	0.0550	2.470	0.0250
KTH	ECARP	Baldwin-Lomax Speziale k-τ	3.136 2.805	0.0756 0.1203	2.416 2.080	0.0178 0.0639
ONERA	ECARP	Le Balleur	3.260	0.0560	2.460	0.0250
SAAB	ECARP	Jones-Launder k-ε + Wolfshtein idem + SST Shih-Lumley-Zhu-Wolfshtein model	3.203 3.146	0.0455 0.0525	2.406 2.429	0.0282 0.0250

Table 3 lists the experimental and computed lift and drag (not all data were available on the ECARP CD-ROM, and missing data were taken from a figure in the book). The first six results in Table 3 concern the EUROVAL project, while the others concern the ECARP project. The conclusions of EUROVAL were that the lift coefficient is reasonably well predicted, while the drag coefficient is largely over predicted. This is attributed to grid inadequacy, and a too high artificial dissipation in the boundary layer. The ECARP results show that the lift coefficient is slightly over predicted. The drag coefficients are computed within 10% of the experimental value, and the systematic over prediction found in the EUROVAL project has disappeared. A discussion on the over prediction of the drag coefficient in the EUROVAL project can be found in [13], and is attributed to the influence of numerical dissipation, and the fact that the far field boundary is too close to the airfoil (10 to 15 chords instead of the required 50 chords). The far field circulation correction mentioned before was developed to correct this last point, and was used by several partners in the ECARP project.

Godin et al. [14] did not provide the aerodynamic coefficients.

Pressure distribution plots

Both the EUROVAL and ECARP results contain Cp plots.

EUROVAL: general agreement with experimental data is good, except for the results provided by CERFACS. This is attributed to the Steger-Warming flux splitting scheme which seems to produce an excessive amount of dissipation. Cp distributions of Dornier and CFD Norway on the flap for the 2.6% gap case show some differences with the experimental results, which are attributed to the grid and/or turbulence model. A comparison was made between the full Navier Stokes and Thin layer Navier Stokes formulation, and it was concluded that the full Navier Stokes formulation yielded better results. The thin layer formulation seems to yield more diffusive solutions.

ECARP (see Fig. 5): general agreement with experimental data is good, and it is remarked that the Baldwin-Lomax algebraic model yields good pressure distributions. Some differences are visible on the flap for the 2.6% gap case, where the SST model slightly overpredicts the Cp.

The results of Godin (see Fig. 14) [14] showed a good agreement with experimental data for both the Spalart Allmaras and Menter SST model for the 2.6% gap case. No overprediction of the Cp on the flap was observed for the Menter SST model.

Skin friction distribution plots

EUROVAL: for the 1.3% gap case, the comparison between experimental and computed Cf is reasonable on the wing, except for the results of CERFACS. On the flap, there is only one experimental data point, and none of the computed results comes close to this value. For the 2.6% gap case, only the results of SAAB using the two layer k-ε model, and of CFD Norway using the Baldwin-Lomax model are close to the experimental results on the wing. On the flap large differences are apparent between the computed Cf.

ECARP (see Fig. 6): computed Cf results for the 2.6% gap case are close to the experimental values. Skin friction results obtained with algebraic turbulence models are as good as the results obtained with two-equation models. The Menter SST correction did not yield a significant improvement to the results.

The results of Godin (see Fig. 15) [14] showed a good agreement with experimental data for both the Spalart Allmaras and Menter SST model.

Velocity profiles plots

Velocity profiles were measured at different stations on the flap. Comparison with experimental data is made for stations 8, 12, 13, 14 and 16 (see Figure 1 for the locations). The most interesting phenomena is the mixing of the wake from the wing with the boundary layer on the flap.

EUROVAL: only the results for the 2.6% gap case are presented, and they show substantial differences in measured and computed velocity profiles. Results show that for the upstream stations, the Baldwin Lomax model yields better results than the k-ε model, due to the faster decay of the wing wake predicted by the k-ε model. However, further downstream (stations 14 and 16), the k-ε model produces better results.

ECARP (see Figs. 9 and 11): in general a good agreement between computations and experiments was observed. In particular the BAe Kalitzin-Gould model yielded a very good agreement. Two equation models seem to give a better agreement than algebraic turbulence models, and the results using a coupled viscid/inviscid interaction approach. Using a non-linear approach (SAAB, Shih-Lumley-Zhu-Wolfshtein model) did not improve agreement with experiment.

The results of Godin (see Fig. 16) [14] showed a good prediction between calculations and experiments, with the Spalart Allmaras model performing slightly better than the Menter SST model. Note that the ECARP results of HUT using Menter appear to be slightly better than the results of Godin.

Reynolds stress profiles plots

Reynolds stress profiles are available for the case with 2.6% gap. The plotted values shown in Figure 10 and Fig. 17 are qr/U_∞² with q and r respectively the instantaneous velocity components along the normal to the surface and parallel to the surface.

EUROVAL: only the results by CFD Norway (using both Baldwin Lomax and the Chien k-ε model) are realistic, and they show that the k-ε model predicts the Reynolds stresses slightly better. However, the high Reynolds stress levels at stations 12 and 13 (see Figure 1) could not be predicted.

ECARP (see Fig. 10): the results with the BAe Kalitzin-Gould model yield the best agreement with experimental data, followed by the results from the Jones-Launder, Menter SST, and the Jones-Launder-Wolfshtein models. Results with the Lam-Bremhorst, and the Le Balleur model are worse. The BAe Kalitzin-Gould model is the only model able to predict the high Reynolds stress levels at stations 12 and 13.

The results of Godin [14] showed that the Spalart Allmaras model predicted the Reynolds stresses in closer agreement with experimental data than the Menter SST model. Comparison of the results obtained with the Spalart Allmaras model and the Kalitzin-Gould model showed that the latter results are closer to the experimental values.

Contributors: Jan Vos - CFS Engineering SA