Fix wall distance computation for points on viscous wall

[rois1995]

Proposed Changes

This PR fixes the problems encountered in #2474, where for some points on a viscous wall the wall distance was higher than EPS, thus the normal distance introduced in #2392 will be wrong for any point connected to the incriminated ones. This will cause a large Omega at the wall when the SST model is considered.

Develop:

Current branch:

Here I initialize the wall distance on all the points belonging to a viscous boundary to 0, thus it won't be overwritten by the wall distance computed by the ADT tree.

Related Work

#2474 #2392

PR Checklist

Put an X by all that apply. You can fill this out after submitting the PR. If you have any questions, don't hesitate to ask! We want to help. These are a guide for you to know what the reviewers will be looking for in your contribution.

• I am submitting my contribution to the develop branch.
• My contribution generates no new compiler warnings (try with --warnlevel=3 when using meson).
• My contribution is commented and consistent with SU2 style (https://su2code.github.io/docs_v7/Style-Guide/).
• I used the pre-commit hook to prevent dirty commits and used pre-commit run --all to format old commits.
• I have added a test case that demonstrates my contribution, if necessary.
• I have updated appropriate documentation (Tutorials, Docs Page, config_template.cpp), if necessary.

[pcarruscag]

This makes sense but I think there is an easier way, change

  for (const auto jPoint : geometry->nodes->GetPoints(iPoint)) {
    const su2double dist = geometry->nodes->GetWall_Distance(jPoint);
    if (dist > EPS) distance = fmin(distance, dist);
  }

to skip points on viscous walls (essentially what you are doing here).

[rois1995]

This makes sense but I think there is an easier way, change

  for (const auto jPoint : geometry->nodes->GetPoints(iPoint)) {
    const su2double dist = geometry->nodes->GetWall_Distance(jPoint);
    if (dist > EPS) distance = fmin(distance, dist);
  }

to skip points on viscous walls (essentially what you are doing here).

I also was thinking about it but how do I know if the iPoint is on a viscous wall?

[pcarruscag]

There is something in geometry->nodes-> I think.

[rois1995]

There is something in geometry->nodes-> I think.

Found it, geometry->nodes->GetViscousBoundary(iPoint) I guess. Something like this?

  for (const auto jPoint : geometry->nodes->GetPoints(iPoint)) {
    const su2double dist = geometry->nodes->GetViscousBoundary(jPoint) ? 0.0 : geometry->nodes->GetWall_Distance(jPoint);
    if (dist > EPS) distance = fmin(distance, dist);
  }

Isn't the proposed fix more general though? As the wall distance on the points on a viscous domain is exactly zero, so we don't have this kind of problems if it is ever needed anywhere else in the same way. Maybe I can do it within the CPhysicalGeometry function, something like

void SetWallDistance(su2double val, bool init=false) override {
    su2double valHere = 0.0;
    for (unsigned long iPoint = 0; iPoint < GetnPoint(); iPoint++) {
      valHere = val;
      if (init && nodes->GetViscousBoundary(iPoint)) valHere = 0;
      nodes->SetWall_Distance(iPoint, valHere);
    }
  }

In either cases I think it is a larger overhead than just cycling on all the viscous points and setting it to 0 after the first uniform initialization.

[pcarruscag]

No I don't think it is because there is nothing wrong with the wall distance computation, the issue is in this normal neighbor distance function that is using an indirect way of detecting boundaries. Fix it there please.

[rois1995]

No I don't think it is because there is nothing wrong with the wall distance computation, the issue is in this normal neighbor distance function that is using an indirect way of detecting boundaries. Fix it there please.

Sure, no problem!

[YairMO]

Hi,

Would this guarantee zero wall distance at the wall? Or in other words, would this cover all possible cases of geomtries?
Why is guaranteeing a zero-wall distance not sufficient? Is it expensive in some way, or will it fail?

[pcarruscag]

The issue is not computing the wall distance at the wall (which is numerically 0); it's computing the distance to the nearest internal node.
In any case no part of the code should be relying on the wall distance of wall nodes, I hope.

[YairMO]

Agreed.
I may have completely misunderstood the issue, but I thought that it was found that a node on the wall has a very small value of wall distance (for some reason). This value "passed" the if statement. As a result, the assigned distance for this node was its own (false) distance. Am I wrong?

Thank you, and sorry for the trouble

[pcarruscag]

The small but nonzero value was interfering with the new way of computing the distance to the nearest node.