VSPAERO solver and convergence

[Jacob]

I have two questions sort of related:

1) Can you explain how to interpret the solver's outputs for each iteration? As in what do the values for "Reduction" and "Maximum" mean?

2) My design is very simple, just a wing and horizontal tail, yet I can't seem to get the solver to converge. In the .history file, the CL jumps around varying by about +/- 0.2, and the induced drag coefficient is sometimes negative. I have tried increasing wake iterations and it still jumps around. Is there a way to tell the solver to not stop until it converges on a solution? Or is there anything that you've encountered that might lead a simple two-wing setup to not converge in <20 wake iterations?

Thanks

[Rob McDonald]

In my experience, convergence problems are usually caused by the wake going a bit nuts.  That in turn is usually caused by an unfavorable interaction between components.

Does the mean surface of your wing come close to coinciding with the mean surface of your body?  This can often cause problems.

For many VLM type studies, you're just as well off to drop the body entirely.  Does that help you make any progress?

Also, how many spanwise/chordwise panels are you trying to use?

Rob

[Jacob]

I got it to converge finally, thanks for mentioning the interference. I didn't have a fuselage in there, but there was a tiny extra section at the root of the tail that had a near-zero chord length for some reason, which I didn't notice. I guess the solver didn't like that.

I'm still curious though how to interpret "Reduction" and "Maximum" in the solver console.

[Rob McDonald]

The GMRES solver is an approximate iterative matrix solver. These are
measures of the error of the matrix solution a that point in the
iteration.

rho0 is the residual on the first iteration.

Reduction is log10(rho/rho0) and maximum is log10(rho).

Convergence is set to rho<1e-3, so log10(rho) is -3.

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.
> For more options, visit https://groups.google.com/d/optout.

[Dustin Otten]

Is there any way to change the convergence criteria so it so that convergence could be set to either a value of reduction or a different value of Maximum?

Is there a way to set the max number of GMRES iterations?

[Rob McDonald]

Dustin,

If you are willing/able to re-compile the solver, yes.  However, these parameters have not yet been exposed for the user to adjust.

Rob