Box wing optimization

[Edward Pickens]

I have been playing with the box wing configuration with the intent to optimize the design and make a small physical test aircraft.  The program idea is simple.  I input a boxwing model.  The program then twists the sections to arrive at a minimum drag configuration for the test condition.  After that I would change the chord.  If the local Cl was below the CL value I would shorten the chord so that every section was operating at near the ideal CL value.  I have had some nice looking results, but I have also found some results that leave reality far behind.

It has been a struggle since my last post.  I think my two main problems are working around VSP's handling of the box wing configuration and understanding how to properly run the analysis.  I don't want to give up and I think if I persevere I can arrive at some good data.

I initially used VLM for the speed, but found the panel method to have fewer problems.  The geometry design has been problematic.  If the winglets overlap the wing then I often find that I get singularity type issues at the point of overlap.  Cl in these areas grows too high.  I am concerned that if I make a gap too great between the wingtip and the vertical winglet I will loose accuracy in my results.  I have three potential solutions that I would like feedback on before I try.

One possibility is to make one wing that extends out then up and back in.  The possible advantage of this solution is a solid body without overlapping nodes that cause singularity type issues.

The other thought, that I have less confidence in, is to use Munk's Stagger Theorem and completely separate the parts longitudinally to avoid nodes overlapping. This should more accurately determine the CDi, but won't work well with moments.

The final possibility is to simply leave a physical gap between the wing and the winglets.  How much though and what will the gap do to the results?

The optimization and the drag calculations have been a struggle.  I initially wrote the code in Matlab, but left that software due to their recent policy changes.  I am now learning / vibe coding python.  I feel dirty.  I have tried a few different techniques to generate a good mesh, but not without issues.  

I have been seeing differences in the drag on the trefftz plain and integration locally across the wing.  I assume this is a meshing problem.  The program initially increased the meshing until the CDi appeared to stabilize.  I would like some advice on what to check to set up an automatic mesh checker in the software.

I have also looked at the local Cl values and set a flag on the software in the event a Cl reached some unreasonable value.  I would assume separation at this point and fail the updated model.  This was a checker I used early on to prevent issues near the wingtip winglet intersection not representing reality.

I can keep going on, but in the interest of brevity I should have ended this already.  What can I do to get this idea to work?  I need to create geometry that wont lead to problems, create a mesh checker to verify I built the mesh correctly, and I need to accurately calculate the CDi to include a checker such as CDi vs CDt?  In the future I will keep better notes and set up each program mode like an experiment in a lab, but at the moment I have exhausted my limited knowledge on this program and need to phone a friend.  If you have made it this far, thank you.

[Rob McDonald]

I would likely work with a single wing wrapped around -- that should take care of most of the possible continuity problems.

The thin surface mode should suffice for this project, give it another try when you try the single lifting surface approach.

Other than that, post a model so people can try to give more specific feedback.

Rob