VSPAERO CDi convergence issues

[Diego Elvira]

Version: OpenVSP 3.46.0

Wing Configuration: Trapezoidal wing with 3 sections (4 airfoils, each with a different AF file which includes twist). Model posted below

Problem: I am having some issues with thick model CDi convergence when obtaining results from this wing in VSPAero. Thin model results for CDi, CD0 and CL show good convergence but in the thick model CDi results do not converge, whereas CD0 and CL do. I would appreciate any guidance or suggestions as to what I could be missing.

On a different note, I’m curious why I’m getting CD0​ results in the thin model, since I wouldn’t expect any parasite drag there. 

Thank you in advance!

[Brandon Litherland]

The CD0 term is based on the flat plate drag with some NACA 0012 lift-dependent corrections built in. So the behavior should be very similar between Thin and Thick modes for a simple wing like this.

I noticed that your airfoils all have blunt trailing edges.  I adjusted the Closure, skewing both, to zero thickness to ensure a sharp trailing edge and then ran your example case to 10 wake iterations.  The CDi term converges down to less than E-5 which is very good. Out of curiosity, I tried running the solver without sharpening the edges and, sure enough, the new VSPAERO is able to detect the trailing edge and apply the Kutta condition even with blunt edges.  That's a new development that I must have forgotten about.  The convergence of this blunt TE case was almost exactly the same as with sharp edges.

Can you post an image of the results you're seeing that suggest CDi isn't converging?

[Diego Elvira]

Thank you for your response.

My apologies, I didn't specify it clearly earlier. My question was about mesh convergence. When I increase/decrease the U (spanwise) or W (chordwise) mesh parameters, CDi results vary significantly and do not converge (only in thick model).

Thanks again for your help

--
You received this message because you are subscribed to the Google Groups "OpenVSP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openvsp+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/openvsp/fb9ad100-f0e6-4458-ae47-d49e2cbb4da5n%40googlegroups.com.

[Diego Elvira]

Hello everyone,

I’m including the images of my previous results here to provide more context for the mesh convergence analysis. The figures show the variation of CDi results with different U and W parameters, which might help clarify the issue I mentioned earlier.  

[Brandon Litherland]

I'd like you to try another comparison if you have time.  

First, instead of changing U or W by a fixed amount, try doubling from a relatively low setting.  Start with W of 17 or so, and U of maybe 4.  You may see larger trends in the solution this way that could have been missed by taking incremental steps.

Second, there are other factors to consider such as leading edge, trailing edge, and wingtip clustering.  I recall that your LE clustering was set to something like 0.1 which looks really nice for a thick panel distribution but can very quickly cause thin panels with coarse U spacing. The solver tends to behave a bit better when the panels are more "square" rather than lots of thin panels.  You may want to try finding some ratio of W to U that gives you a nice looking panel distribution.

Finally, you will need to ensure that you are capturing all of the relevant effects well before coarsening back down for efficiency. As you can see in your results above, setting U above 20 doesn't seem to matter much but the solution keeps climbing with W.  Something is off there and I recommend exploring some of the other controls before moving forward.

[Diego Elvira]

Thanks for the advice. I’ll look into these suggestions as soon as I’m able.

To view this discussion visit https://groups.google.com/d/msgid/openvsp/23693a5c-4841-4dbd-b436-f62023645ea9n%40googlegroups.com.