[milind543]

Hi,
I was thinking to apply the text book problem of minimizing the induced drag for the elliptic wing loading or the shape, hence i created a rectangular wing n thought of using the same cfg file as of onera case in shape optimization (tutorial problem) with some changes and i though instead of surface movement in z direction in my case i can give movement in x direction in order to get a chord distribution for the optimized wing, well i realized that after some optimization cycle( 9) the flow solver blows up and the drag after actually increasing in 2nd cycle drops down to same as the initial rectangular wing after 9th cyle, and the shape also remains the same. Hence i was wondering where am i going wrong. Allthough i must agree i m using a very very course grid(order of 25000 cells) hence i dont expect a good solution but i would just expect some bench solution for me to know the input cfg is correct. Please help me. I have attached my cfg file.
Thank you

[fpalacios]

Hi,

The problem that you have proposed is very interesting, but very expensive in computational terms. To give you an idea, to properly capture the induced drag you will need a structured grid with ~10M cells, otherwise the induced drag and the spurious drag are of the same order and a gradient based optimization won't work. A good reference about this problem is http://oddjob.utias.utoronto.ca/dwz/...nduceddrag.pdf

Once you have the proper grid, my recommendation is that you should fix the trailing edge and move the leading edge to change the platform of your geometry. You should keep in mind that in sharp edges (like trailing edges) the gradient computation is much complex.

Thanks for using SU2.

Best,
Francisco

Originally Posted by milind543

Hi,
I was thinking to apply the text book problem of minimizing the induced drag for the elliptic wing loading or the shape, hence i created a rectangular wing n thought of using the same cfg file as of onera case in shape optimization (tutorial problem) with some changes and i though instead of surface movement in z direction in my case i can give movement in x direction in order to get a chord distribution for the optimized wing, well i realized that after some optimization cycle( 9) the flow solver blows up and the drag after actually increasing in 2nd cycle drops down to same as the initial rectangular wing after 9th cyle, and the shape also remains the same. Hence i was wondering where am i going wrong. Allthough i must agree i m using a very very course grid(order of 25000 cells) hence i dont expect a good solution but i would just expect some bench solution for me to know the input cfg is correct. Please help me. I have attached my cfg file.
Thank you

[milind543]

Well yaa i realise the importance of grid independence, but i was wondering if I can just have around 2M mesh n then if I try i would get something, for that i wanted to know if my cfg input file is correct hence i thought i should probably start with a very small grid and as i increase the grid size i should see how the shape becomes clearer and clearer, Is it possible that i optimize the rectangular wing with the spurious drag and as i increase grid size my solutions become better.

Thanks