Problem using CFD Mesh

[Steve]

Hi Rob,

I'm working on making a store model with VSP and exporting that to Cart3d.  However, on one side of the fin, the nodes don't match such that there is a free edge (it's not watertight) on one side of the fin.  The free edge is shown in pink in the picture and occurs on all 4 fins.  It looks like it should closed given the gap size.  I assume I'm doing something wrong in terms of my mesh sizing.  I looked at the meshing strategy guide, but it's not obvious.  Do you have any ideas on how to work around this?

I've attached the model that creates this issue.  Also, it'd be nice to be able to see free edges in VSP directly as lines.

The CFD meshing strategy guide (for those who haven't seen it).  http://www.openvsp.org/wiki/lib/exe/fetch.php?media=file_type_meshing_2013.pdf

Thanks,

Steve Doyle

[Rob McDonald]

Well, there are a handful of things I would do differently to model
this shape, but as it turns out, none of them seemed to affect the
problem you were having. In the end, you've hit on a bug we have....

First, you can visualize bad edges -- they are normally included in
the visualization, but they can be hard to see with the entire mesh
shown and the triangles filled in. So, under the 'Display' tab in
CFDMesh, you can turn off 'Mesh' and 'Color Tags'. That will usually
make bad edges/tris easy to see. If not, keep turning stuff off --
they'll be what is left in Red.

Rather than use XY and YZ symmetry for your fins, I would use X radial
symmetry. And, instead of locating them relative to an offset axis, I
would use the 'Attach UW' capability under the XForm tab to attach
them to the body of the store. With radial symmetry, it is really
easy to change the number of fins.

Next, I saw that your body skinning parameters were a bit odd. I
noticed this because CFDMesh wasn't sharply resolving some of the
sharp corners on the back of the store. Looking more closely, you
were setting some tangents to be C1, 45 deg, but with zero strength.
What you really wanted was a C0 corner -- and if they are straight
lines, you can just choose to not specify the angle and it will take
care of itself. I changed the back-half of the store's skinning to be
a bit more simple. There are some possible simplifications on the
front end, but I didn't want to change your shape.

I tweaked some other things -- to no avail. I tried to switch them
all back, but I may have missed one or two. So, please be careful
with the attached model, it may have some unintended changes.

In any case, I'm sure you're most interested in the bug and a workaround...

CFDMesh has a number of tolerances hard coded in it. Some of them are
in model dimensions, others are relative to some parameter or another.
In any case, they can have strange interactions. We certainly need to
improve this situation to make the mesher as robust as it can be.

In this case, I recognized that your store is 203 units long (probably
inches). That is a relatively big number -- that sometimes gives some
of the tolerances trouble. So, I used the 'Scale' property to scale
each component of your model down to 1/10th size. I then manually
scaled things like mesh size parameters and the sources you had
specified.

Once I scaled the model to 1/10th original size, I coudn't get a mesh
to mess up. So, most of the above (while good practice) probably
didn't matter.

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.

[Steve]

Rob,

The model looks much better.  I was running with an up scaled model because on another model I had run into issues had issues with fins collapsing at the 20% chord location, but were fine at the 10% chord location.  In regards to the implicit tolerances, is there a recommended model length?  I had assumed 100 was more or less the standard length.

Also, thank you for fixing the boat tail.  I thought that error at the base of the CFD mesh was just an issue of the max gap since the geometry looked superficially correct.  Ironically, it's easier to work with complex (but smooth) geometry rather than simple shapes.  I definitely need to learn how to control the strengths better.

Thanks again for the help.

Steve

[Rob McDonald]

Unfortunately, I don't know that I have a good answer for 'best' model
size -- things roughly the size of the default components basically
get tested the most...

When OpenVSP draws the feature lines, it tries to detect the
discontinuities -- those then get passed to the CFDMesher as key lines
to get picked up. In that case, I bet it didn't show as a feature
line, so that would be a visual giveaway that the surface wasn't as
sharp as you thought it was.

Rob