VSPAero Rotor Disks and Convergence

[Gust]

Hello All,

     I have been using the VLM solver in VSPAero to model a distributed propulsion wing. I have been getting extraneous solutions for most of my calculations with negative drag coefficients. Is there a certain number of spanwise nodes I should be using to ensure I get complete convergence for such a problem?

Here are my parameter for the wing:

Num_W:41

Num_U:240(Weighted 60 behind each prop)

Prop Parameters:

Num_U:20

Num_W:21

I appreciate any pointer you can give me.

[Gust]

Here is my VSP file.

[Rob McDonald]

I played a bit with your model and didn't really have any trouble.

First, I deleted all the pods -- having those in such close proximity to the wing in VLM mode will likely not give the kinds of results you want.  If they're well designed, they won't have much influence on the lift distribution, so I'm OK leaving them out for now.  You can try adding them back in to see if there are problems.

Next, you had 100 alpha points planned between -5 and 5 degrees.  That would cause you to run 100 solutions, taking 100 times as long.  Probably not what you wanted.  I only had interest in running one solution, so I ran it at 10 degrees.

The paneling on the actuator disks doesn't matter at all for VSPAERO.  The paneling doesn't actually get transmitted -- the only thing the solver cares about (geometry wise) is the center position, normal vector, and diameter.

Your spanwise resolution is very high, but I left it as-is just to see how it did.

The attached images are from the resulting solution.

There were Cp spikes that lead to an initially unreasonable Cp distribution in the visualization.  I changed the Cp limits to something reasonable [1,-5] or so to create the attached plot.  I would investigate the wingtips to see if I could find a cause of the spikes -- you have very small panels with the prop right there, perhaps there is something you could do.

As far as spanwise resolution, your goal should be to do a decent job resolving the spanwise lift distribution (see attached plot).  I think you've achieved that.  You could certainly reduce the spanwise resolution away from the props (10x), and likely  reduce it near the props (2-3x).

Rob

--
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+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Gust]

 Thanks for the fast reply! 

When I ran the model I kept the pods out of it just like you did. I guess an issue I'm having is that from my understanding, at least for finite element solvers, the higher the number of

panels you use the better convergence you should get on a solution. I've been experimenting with the number of panels for both VLM and the Panel Method and this doesn't seem 

to be the case. Is there a methodology I should use when paneling in VSP?

[Rob McDonald]

What do you mean by 'better convergence'?

VSPAERO has a number of processes that are related to convergence.

First, it uses an iterative GMRES solver to solve the matrix solution
-- this process is tracked in terms of convergence.

Second, it uses time stepping wake filaments that update their shape
in response to the flow. By default, there are five wake iterations,
you bumped it to 10. That number of iterations is user-input, but you
can monitor convergence of force coefficients in the results manager
GUI to see how CL,CDi, etc. converge with wake iteration.

However, I suspect you might mean convergence of the solution as you
change/increase the mesh resolution? In that case, VSPAERO does not
have any built-in convergence criteria -- it is up to the experience
of the user to determine if they have sufficient resolution to
satisfactorily model the problem of interest.


So, if you're increasing the number of panels in an attempt to improve
#1 or #2 above, then that isn't really how things work, you should
describe the convergence problem you're experiencing and we can try to
guide you to a solution.

Or, if you're trying to perform a grid resolution study to determine
if you have sufficient panel resolution, then you need to let us know
what you are trying to get out of the VSPAERO model. Are you
interested in CL? CDi? CM? Load distributions? Pressure
distributions?

In a panel code, it is quite likely that CL and CDi will 'converge'
much earlier than some other characteristics of the solution.

So -- what are you trying to achieve -- and what problem are you experiencing?

Rob

> --
> 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.

[Gust]

I'm trying to get the CL and CDi values for a range of AoA's. The issue I've had is that I will consistently get CDi values that are negative. The CL values also vary widely between  iterations with different mesh sizes, which from your previous reply seems to be part of the issue.

 I've run the wing in the panel code and I've gotten the below results for CDi and convergence. I noticed this convergence issue before but haven't been able to figure out how to get better convergence. There is still something I'm missing.