Metacomp Technologies' Contribution to 2^nd AIAA Drag Prediction Workshop

Orlando, Florida

June 21-22, 2003

The commercial CFD solver CFD++, from Metacomp Technologies, Inc., was used to compute the flow-field around the DLR-F6 Wing/Body/Pylon/Nacelle (WBPN) geometry. The Wing/Body (WB) configuration by itself was also computed with the aim of predicting the incremental total vehicle drag due to the presence of the pylon and nacelle. Experimental data from ONERA (AGARD-AR-303 and subsequent publications) were used to assess the CFD predictions.

CFD++ solver highlights

• Basic algorithm: finite volume cell-based mixed-element unstructured

• Spatial discretization: multi-dimensional TVD (inviscid), non-decoupling non-limited face polynomials (viscous)

• Time integration: point implicit with multi-grid relaxation (steady state)

• Turbulence closure: several topography-parameter-free models are available. The realizable k-e model was employed in the current effort (Goldberg et al, ASME JFE 120, Sept. 1998, 457-462).

Grids employed

Hexahedral grids, produced with the ICEM mesh generator, were used. In all grids was maintained at the wall-adjacent cells, with a growth rate of 1.23-1.28, hence wall functions were not used. About 20 cells were maintained within boundary layers. The following table shows the various grid sizes employed for mesh sensitivity studies, with the medium-size grids used for the bulk of the calculations.

Mesh	Coarse	Medium	Fine
W+B+P+N	4.8 M	8.5 M	12.8 M
W+B	5.5 M	7.4 M	9.6 M

Table 1: Grid sizes used for the two geometries.

Solution information

• Computer platform: PIV Xeon 2.4 GHz
• Number of processors: 12
• Run time wall clock: 12-13 Hrs. (6-8 Hrs. for restarts)
• Force convergence: < 400 time steps

Turbulence management

1. Turbulence intensity: 0.002 (based on AGARD-AR-303)
2. Turbulence length-scale: 0.6 mm (assumed)
3. Transition: no transition imposed; flow was allowed to transition naturally over the wing and fuselage.

Flow conditions:

• M = 0.75
• Re = 3 x 10⁶ (based on mean aerodynamic cord)
• T
• Angles-of-attack:

Forces and moments

The following figures show forces and moments predictions for the two configurations (WB and WBPN)

Fig. 2: Drag polars.

Geometry	Max. Cd deviation (drag counts)
WB	5
WBPN	7

Table 2: Maximum Cd deviation in drag counts.

Fig. 3: Effect of grid refinement on solution quality.

The above figure shows improved predictive quality with mesh refinement.

The following set of figures compares predictions with data for Cp profiles at several wing. Roof-tops, shock locations and suction peaks are observed to be well predicted.

Fig. 4: Cp profiles at eight wing sections.

Fig. 5: Lift coefficient vs. angle-of-attack.

Fig. 6: Moment coefficient vs. angle-of-attack.

Fig. 7: Drag coefficient vs. angle-of-attack.

Lower figure shows inviscid and viscous contributions to total drag for the WBPN configuration

Fig. 8: Details of trailing edge separation.

Fig. 9: Mach contours showing shock over wing and flow through nacelle. Boundary and shear layers are observed.

Fig. 10: k contours corresponding to Fig. 9. Note absence of k in stagnation regions due to realizability constraints.

Fig. 11: Streamlines showing wing leading-edge saddle point and vortex, wing/fuselage flow separation and pylon/nacelle separation.

Fig. 12: Typical force convergence history. Less than 400 iterations are needed for convergence.

Mesh	Coarse (4.8 M)	Medium (8.5 M)	Fine (12.8 M)	Exp.
Delta Cd total	0.0056	0.0049	0.0046	0.0043

Table 3: Effect of grid refinement on incremental drag prediction.

The above table shows that the fine mesh enabled Delta Cd total prediction within 3 drag counts of the experimental measurement.

Summary and conclusions

Fully turbulent flow computations were performed, allowing the realizable k-e model to choose its own natural transition over the wing and fuselage.

• Excellent polar predictions were obtained for both configurations.
• Maximum Cd deviation measured from the polars: 7 counts for WBPN, 5 counts for WB.
• Grid refinement led to improved predictions with CFD++.
• CFD++ took less than 400 iterations to converge forces and moments.