I have recently performed a CFD analysis of a small Tandem wing UAV. My aim is to verify the CFD results with existing literature and check the error prediction margin of SimScale. After running the simulation I found the following results,
% error in CL was 72% w.r.t experimental values
% error in CD was 4% w.r.t experimental values
I am puzzled about how the CD is so well predicted but the CL is so divergent. I was unable to draw any conclusions from it, so any suggestions would be helpful.
Hi @udaych17!
Please make sure to share your project here so that we can have a look at that.
Best,
Jousef
I have shared it with you specifically.
Hey Barry,
I have also shared the project with you.
Hi @udaych17,
A few things which may affect your results that you may need to change.
Your simulation is steady-state and incompressible. Initial conditions for pressure and velocity should be set to 0 so as to prevent forcing of the solution. Similarly, the pressure outlet should also be set to 0 which acts as gauge pressure and again prevents forcing of the solution.
Your outer domain (-x direction) may be too close to the geometry, you may need to extend it to maybe 3 times its current length. This small domain length coupled with your assignment of a no-slip wall condition will affect your results. After adjusting the length of the domain and re-meshing, ensure you either assign a slip-wall condition for the side, top and bottom of the domain or a farfield boundary condition that represents free air.
When using k-w sst as your turbulence model coupled with a wall function, you need to ensure that your y+ (layers) are strictly between 30 & 300. Ensure that your layer calculations adhere to that and while it is very challenging to get layer generation up to 90%, check that your calculation for y+ is adhered to first and we can rectify the layer generation later if needed.
Consider post-processing your data offline in ParaView and deducing CL/CD from ParaView itself rather than relying on the result control. It is much more accurate and controllable. This tutorial may help you on getting CL/CD data out in Paraview. However, keep the result control. It will be able to tell us if the simulation has truly reached steady-state convergence. From your convergence plots and result control plots, your simulation has converged nicely so no problems there.
Under numerics, consider setting your absolute tolerance to 1E-6. This will ensure that if your simulation does converge, then we have very acceptable and ideal numerical residuals. However don’t be worried if it doesn’t converge as reaching 1E-6 can be challenging. On that note, also set your Pressure reference value back to 0.
Lets work from there and see where we stand afterwards.
Cheers.
Regards,
Barry
Thankyou Barry. Any advice on how to go about computing the y+ values analytically? I am not that experienced with it.
1 Like
Thankyou Barry!
Hi,
I have shared the project. Any specific suggestions from your side?
Best,
Uday
Hi Uday,
Hope you’re not in a rush. Might take a bit as I have some other projects that need to be finished. Will have a look at it as soon as possible.
Best,
Jousef
Hi Uday,
before proceeding with the simulation I would definitely make sure to refine the wings a bit more as the layers are generated but the whole wing just looks odd at the moment although we almost have 0 illegal cells. For the boundary conditions I would only assign the wall BC (boundary condition) to the plane itself and not the BC’s on top bottom and the face parallel to the symmetry plane - use a slip BC instead.
@Get_Barried might add his two cents here. And if you allow us to copy the project we can run some mesh testings for you in case you are struggling to make it work.
Cheers!
Jousef
Sure.
Thanks,
Uday