Computational fluid dynamics

CFD: Viscous flow over a wing profile, turbulence modelling.

Spalart-Allmaras: Wing at 6* angle

The aim of this report was to investigate the behaviour of flow over an isolated symmetric aerofoil (NACA 0012) in free flight. The flow was assumed to be turbulent viscous flow over the wing profile. Testing was carried out using the FLUENT and GAMBIT software. Initially the aerofoil was simulated in GAMBIT and different mesh sizes were used to capture the changes in flow. Meshes were then refined by changing the number of nodes and the ratios. The changes in the flow phenomenon were noted by constantly reducing the mesh size. Then the mesh was imported in fluent and the solutions were analysed for the values of lift and drag coefficient CL and Cd at different angles of attack α. The solution was to be obtained using the RANS model namely the Spalart – Allmaras model. The computational results from this method were compared with the k – ε model to see which one is better.

Comparison of the CL and Cd values of Spalart-Allmaras (S-A) and the K-ε model at 6⁰ angle of attack.

Model	Angle(0)	CL	Cd
S-A	6	5.89E-01	1.39E-02
K-ε	6	6.10E-01	1.20E-02

Spalart-Allmaras: Wing at 17* angle (Onset of separation)

The corresponding CL and Cd values at those angles of attacks were noted and were compared with the given K-ε model.

Model	Angle(0)	CL	Cd
S-A	17	9.89E-01	2.55E-01
K-ε	17	1.64E+00	4.70E-02

Comparison of Experimental Results with the provided K-ε data

Comparision of experimental and theoretical data.

CONCLUSION

After completing the exercise, it was concluded that the Spalart-Allmaras model gave more accurate and better results than the k-ε model. Although there were slight discrepancies in the Splart-Allmaras model, if however the second equation was to be incorporated in this model, it would have better dealt with the separated flows. It is also less expensive then the k-ε model and in fact produced better results. The difference in results was also due to the assumption of 2-dimension. And also if the viscous sub layer was fully modelled, it would have given us a better approximation, however this would have been computationally expensive.