3. Summary CFD Workshop : 3.2 - 3.4

3. Summary of CFD Workshop

3.2. High Lift and Mesh

3.2.1. Case & Geometry

High lift configuration, mostly used for the landing, is hard to be predicted because most high-lift devices are related to the separation, transition, and boundary layer in narrow gaps and discontinued element of airfoil sections. Indeed, it is natural that high lift prediction workshop held long after the drag workshop; CFD techniques is required to have better accuracy than the general configuration of the drag. Compare to the drag workshop, rather than accuracy or convergence, prediction of the tendencies related to the separation and high lift curve are important; cases and configuration of the workshop are shown Fig. 57. The 1st and 2nd WS focused on the baseline configuration of the high-lift while 3rd one intensively figured out difference between filling-gap (separated one is similar to real-world) and addition of the nacelle-pylon. 

Fig. 57. Configuration of the High-lift prediction workshop

3.2.2. Result Comparison

 1) Summary of 1st Workshop

Overall results from the whole participants [21] are summarized in Fig. 58; there are very wide variance of high lift curve while CFD under-predicted and diverged the lift and drag at the certain AoA. When we saw the pitching moment, the magnitude of the moment is lower than the experiment. Despite the result is diverged at the high AoA, maximum lift coefficient is in reasonable range with high CL value from S-A model. It means S-A model shows better performance in high AoA than other models under-predicting lift of the configurations as shown in Fig. 59. 

The grid convergence result is shown at Fig. 60; as described in the drag prediction WS, structured grids shows less diverged and smooth convergence curve than the unstructured ones. Also, it is interesting that dependence of the turbulence model is seemed to be weaken by the unstructured grids. Although, there is a clear trend that the converged lift coefficient is similar to experiment at 28 deg, structured one at 13 deg still underestimate lift coefficient. 

Difference between the flap configuration is shown in Fig. 61. Both CFD and experiment well figured out the changed CL value except few participant’s data. Fig. 62 gives more details about the difference and their CFD codes; all of the codes well described the difference of the configuration however, correlation is not well achieved for more than 20 deg AoA. Discrepancies occurred at the AoA region are unpredictable; flow separation is still tough topic for most of CFD engineers. Fig. 63 and following few figures give more description of the data related to Cp; most CFD result shows good agreement with the experiment. Few points like separation back of the LE flap and edge of the TE flap made discrepancy between the experiment and CFD. Fig. 65 showed more serious difference between CFD and experiment along span-wise length of flap; CFD generally made good prediction however there is a problem near the wing-tip. The WT result shows much stronger suction peak than that of CFD while difference is mitigated by finer grid. 

Fig. 58. Lift curve of the participant’s data

Fig. 59. Maximum lift value of the participant data

Fig. 60. Configuration difference effect on WT and CFD result

Fig. 61. Configuration change for various CFD codes

Fig. 62. Usage of CFD Codes for two configurations. 

Fig. 63. Comparing points for high lift configuration

Fig. 64. Cp comparison for AoA 13 and 28 deg

Fig. 65. Cp comparison for AoA 28 deg along LE of TE flap


Effect of the support brackets for LE and TE flaps are studied as the cases which are expected to decrease the lift, generate improved Cp value form some stations. Fig. 67. Impact of the brackets is consistent at higher AoA, however, amount of the effect is little bit different for individual cases. Few exceptional cases have very different result from others; even the tendency is reversed. At the position of the flap, correlation between CFD and experiment become better; we could think brackets provides vane effect making flow as two dimensional locally. 

Effect of the grid refinement is shown in Fig. 68; the refinement gives positive impact on the variance of the coefficient. And related results are also shown in Fig. 69; more direct convergence of lift, drag, and moment coefficients are described. The converged tendency shows CFD fall to similar value of the CFD, however, the extrapolation of the CFD could not guarantee the expectation become same as that of the experiment. 

In overall, the CFD results under-predict the lift, drag, and moment while scatter of the CFD is larger for high AoA. Local Cp value is hard to predict at wing-tip section, LE, and TE flap. Grid refinement generally gave good tendency without regarding the type of the grid while the turbulence model S-A shows better agreement with the experiment result than the others. 

Fig. 66. Geometry of support brackets

Fig. 67. Effect of support brackets on lift/drag data

Fig. 68. Effect of grid refinement and variance of the data

Fig. 69. Effect of grid refinement and convergence tendency

 2) Summary of 2nd Workshop

The 2nd WS provided additional turbulence model verification and Reynolds number effect to give better insight for the effect of the turbulence models which cannot be discriminated in complex airliner models [22]. Verification case of the models is shown at Fig. 70; DPW5 also used the same case for the same purpose. The result reveals that accurate minimum distance is particularly important for SA model which represent y+ of first near wall cell is sensitive issue for the model. 

Overall result for the three variables, lift, drag, and moment is shown in Fig. 71; it is difficult to tell finer grid makes better result for the prediction especially at high AoA. Added Fig. 72 also represents that very rough converged tendency is revealed while the converged value cannot be exactly same as the WT result. 

Fig. 70. Verification case of turbulence model and important of issue of SA model

Fig. 71. Result summary of lift, drag and moment

Fig. 72. Result summary of lift, drag and moment for their grid convergence at 21deg AoA

Fig. 73. Flow comparison at specific position and grid


Effect of the Reynolds number is shown at Fig. 74, and higher Reynolds number made high CL peak and AoA before stall begins. Rough tendency is also repeated, however, exact value of the WT value is not shown by CFD. Impact of the grid type change is shown for Fig. 75; very random discrepancy occurs for each type of the grid. Whole refinement of the grid via adaptive technique made meaningful improvement however even finer grid cannot represent exact value of the WT. More detailed flow comparison for few positions are shown in Fig. 76. Roughly, CFD could capture the real flow phenomena around the complex flap configuration, however, near the geometry local flow tendency is not repeated by CFD. Even some cases show very different flow configuration than other or WT results. Compare to DPW showing difference is bigger for tip side, difference occurred in the lift workshop seems random for their positions. 

Transition effect with CFD is compared at Fig. 77. For lift and drag, it is hard to tell with-transition option is superior than the no-transition option. Whole summary of the cases is shown in Fig. 78, and authors compare the result to other previous workshops. Finer grid made effective convergence at intermediate AoA however, this convergence is not significant at 16deg of AoA. The intermediate AoA cases having 7deg shows similar variation value compare to other workshop while high AoA region made significant large than that of lowers. 

Fig. 74. Effect of Reynolds number on WT and CFD

Fig. 75. Difference between grid effect (upper), and effect of adaptive grid (lower)

Fig. 76. Flow configuration of the CFD and WT for several position of the wing. 

Fig. 77. Transition effect comparison for lift, drag and pitching moment. 

Fig. 78. Statistical variation summary of workshop

 3) Summary of 3rd Workshop

Again, Rumsey et al [23] provided summary for 3rd Workshop; effect of more detailed geometry difference and attachment are studied, effect of sealing flap gap and nacelle pylon. Actually, in real-world, there is a certain gap between the flaps, however, it can cause numerical instability near the position. As shown in Fig. 80, now the most participant uses SA model because of its robustness and calculation economy. The most of the participant data converged for the verification case which is unclear for high lift cases. 

Effect of the filling gap is clear that it increases the lift for given AoA; more flow is toward to downward. CFD result also show that kind of tendency clearly. At 16deg AoA, induced drag increase is definitely shown however result from 8deg AoA is ambitious. Similar to that of the drag workshop, effect of added nacelle did not make significant impact on the quality of the data as shown in Fig. 82. Fig. 83 shows visualized WT experiment result; it clearly represents how flow is changed by installation of the engine. Cross flow from the engine change the picture of the upper surface flow. Continued Fig. 84 shows Cp value difference of the two cases; result is diverged at the near-wingtip section as described in numerous workshops. Yokokawa et al. [24] shows detailed analysis for why wing-tip section is hard to predict as shown in Fig. 85

Near the wing-tip after certain AoA, flow is separated (even in the root section at the 21.57 deg), then, the separation bubble is observed. Report argue that the separation of the tip is caused by slat support and also FTF part causes local separation around the trailing flap. Although previous part concern that the additional geometry did not make significant impact on the data, at high AoA, small change of the geometry could make large flow separation bubble around the aircraft. Fig. 86 summarized statistical result of the whole cases. 

Fig. 79. Case and Geometry of the 3rd Workshop

Fig. 80. Grid convergence trend for the verification case

Fig. 81. Effect of filling gap between the flap segments

Fig. 82. Effect of filling gap between the flap segments

Fig. 83. Visualization result of WT experiment

Fig. 84. Continued Cp value case w/ pylon

Fig. 85. Visualization result of WT experiment for various AoA

Fig. 86. Statistical variance of the participant data

3.3. Appendix – Mesh workshop

Mesh workshop summarized by Chawner, J. R et al [25] provides more details about the grid used for the previous workshops. Although rule of thumb governing the geometry is same, depending on the generator, spacing of the mesh can be significantly different as shown in Fig. 87. It also can be supported by Fig. 88 which represents deviation of the points from the surface is not negligible. 

Fig. 87. Spacing problem for the grid

Fig. 88. Evaluation of mesh adherence

Fig. 89. Convergence tendency of grid size

3.4. Final Summary of Workshops

 I tried to summarized several CFD workshops related to drag and high lift with high precision; many participants with numerous data and cases are gathered to improve the quality of the CFD results. Due to the effort of the improvement via workshop, compare to 1st DPW, quality of the most data is enhanced for several aspects such as grid convergence, adaptability of complex geometry, turbulence model, grid types, prediction of buffet, and separation bubbles. 

However, still there are several discrepancies between WT and CFD in few points. Exact configuration of separation bubbles, complex flows pictures around the wing-tip, and high AoA performances still need improvements. 

* Reference

[1] Redeker, G., 1994, A Selection of Experimental Test Case for the Validation of CFD Codes, AGARD-AR-303 Vol. II

[2] Wahls, R. A., 2001, AIAA Drag Prediction Workshop Wind-Tunnel Data, AIAA CFD Prediction Workshop

[3] Vassberg, J. C., 2001, Guidelines for Baseline Grids, AIAA CFD Drag Prediction Workshop, 19th Applied Aerodynamic Conference

[4] Levy, D. W., 2001, AIAA CFD Drag Prediction Workshop : Data Summary Comparison

[5] Broderson, O., et al., 2001, Drag Prediction of Engine-Airframe Interference Effects Using Unstructured Navier-Stokes Calculations, AIAA 2001-2414

[6] Laflin, K., 2003, 2nd AIAA CFD Drag Prediction Workshop Data Summary and Comparison

[7] Hemsch, M. J., 2003, Statistical Analysis of CFD Solutions

[8] Tinocco, E. N., 2006, DLR F6/FX2B Summary, 3rd CFD Drag Prediction Workshop

[9] Vassberg, J. C., et al., 2008, Development of a Common Research Model for Applied CFD Validation Studies, AIAA-2008-6919

[10] Sclafani, A. J., et al., 2010, Drag Prediction for the NASA CRM Wing-Body-Tail using CFL3D and Overflow on an Overset Mesh

[11] Rider B. J., 2010, Structured Grid Summary for the 4th Drag Prediction Workshop

[12] Pirzadeh, S., 2010, Baseline Unstructured Grids, AIAA DPW4

[13] Oswald, M., 2010, 4th AIAA CFD Drag Prediction Workshop

[14] Tinoco, E. N., et al., 2009, DPW-IV Summary of Participants Data

[15] Levy, D., et al., 2013, Summary of Data from the Fifth AIAA CFD Drag Prediction Workshop

[16] https://allaboutairplanes.wordpress.com/2012/03/03/the-coffin-corner/

[17] Roy, C., 2017, DPW 6 Summary of Participant Data – Case 1: Code Verification

[18] Tinoco, E. N., 2017, Summary of Data from the Sixth AIAA CFD Drag Prediction Workshop: CRM Cases 2to 5

[19] Derlaga, J., et al., 2017, Preliminary Statistical Analysis of CFD Solutions

[20] Laflin, K., and Orderson, O., 2017, Side-of-Body & Trailing Edge Separations

[21] Rumsey, C. L., et al., Summary of the First AIAA CFD High Lift Prediction Workshop

[22] Rumsey, C. L., HiLiftPW-2 Summary

[23] Rumsey, C. L.., et al., Overview and Summary of the Third AIAA High Lift Prediction Workshop

[24] Yokokawa, Y., et al., 3rd AIAA CFD High Lift Prediction Workshop

[25] Chawner, J. R., et al., 2018, Summary of the 1st AIAA Geometry and Mesh Generation Worshop (GMGW-1) and Future Plans