[Wingman]

Hi,

I'm trying to find the correct density of my Airfoil Mesh. I have results of lift, drag and moment of the Airfoil using 20x meshes with Number of Cells between 4.000 and 600.000 with fixed velocity and fixed angle of attack.

Let's assume that the finest mesh gives the 'correct' results but I would like to find the coarsest mesh that gives correct results within some tolerance.

Are there any rule of thumb of tolerance? I've been using Lift Tolerance, Drag Tolerance and Moment Tolerance of 0.01, 0.0001 and 0.01 respectively but I'm not sure if they are correct. According to them my grid isn't converged yet.

Here is an example of the results.

Cells .....Drag Coeff ..... Lift Coeff.......... Moment Coeff
397150 | 4.719471e-02 | 7.436299e-01 | -1.096162e-02
480996 | 4.817767e-02 | 7.527957e-01 | -1.121976e-02
572866 | 4.770149e-02 | 7.576630e-01 | -1.125989e-02


I don't have measured data so I don't know the true values.

[flotus1]

The key to a successful grid sensitivity analysis is controlling all other sources of error and making sure their contribution is much lower than the error you expect from the spatial discretization.
It would be better to use relative tolerances instead of absolute tolerances with different magnitude.
http://www.cfd-online.com/Wiki/Ansys..._inaccurate.3F

[Wingman]

Originally Posted by flotus1

The key to a successful grid sensitivity analysis is controlling all other sources of error and making sure their contribution is much lower than the error you expect from the spatial discretization.
It would be better to use relative tolerances instead of absolute tolerances with different magnitude.
http://www.cfd-online.com/Wiki/Ansys..._inaccurate.3F

Thanks for this, just tried this method using the numbers I provided in first post. The results are for for drag, lift and momentum respectively (vertical) and for {21} and {32} horizontally.


φ_ext =

-0.4755 -0.4165
-7.4158 -6.5702
0.1106 0.0967


e_a =

0.0100 0.0204
0.0064 0.0122
0.0036 0.0230


e_a_ext =

1.1003 1.1157
1.1022 1.1146
1.1018 1.1160


GCI_fine =

0.1366
0.0879
0.0488


So... Am I supposed to use:
0.1366 * 4.770149e-02 = 0.0065 (instead of 0.0001 for drag)
0.0879 * 7.576630e-01 = 0.0666 (instead of 0.01 for lift)
0.0488 * -1.125989e-02 = -0.0005 (instead of 0.01 for momentum)

[flotus1]

Take a look at your drag results: they either show oscillatory convergence or there is an error not related to the cell size. Either way, extrapolating the results is not an option.
To be more clear: have you made sure that all other sources of error are negligible compared to the spatial discretization error? I am especially thinking about iterative errors from running too few iterations.

[Drunken_SheeP]

Have you checked there is really no more change in the values with more iterations? Also, how did you refine the grid? Did you refine all grid parameters equally?