[WIP] NEMO - New Symmetry BC #1168

fmpmorgado · 2021-01-18T12:47:38Z

Proposed Changes

Implementation of a new boundary condition for symmetry plane instead of using Euler Wall.
The implementation works as follows:

Calculates new viscous eigenvalues for nodes in symmetry planes;
~~Calculates new residuals to take into account the symmetry of the mesh;~~ Only corrects momentum residual vector in current version.
~~Uses a modified volume to calculate the solution in symmetry planes;~~ Does not require modified volume in current version.

Related Work

#657

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 the '-Wall -Wextra -Wno-unused-parameter -Wno-empty-body' compiler flags, or simply --warnlevel=2 when using meson).
My contribution is commented and consistent with SU2 style.
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.

Results

For the sake of demonstrating the capability of the new Symmetry Boundary condition, the image shows a comparison of pressure at the surface, using the new implementation and the old symmetry boundary condition, for a flow over a 3D cylinder and considering the lateral and bottom surfaces as symmetry plane:

New Implementation vs Old Implementation for N2 mixture (2nd Order - NEMO_NS solver - AUSM scheme)

I have run this test without changing the viscous eigenvalues, and it converges to the same value. However, for unsteady simulations, changing the eigenvalues provides better results.

New Implementation (on NEMO) vs Current SU2 Implementation for AIR5 mixture (2nd Order - NS solver - AUSM scheme)

CatarinaGarbacz

the implementation seems good to me. some of the comments need to be updated since symmetry is not a boolean. other than that, awesome contribution

SU2_CFD/src/variables/CNEMOEulerVariable.cpp

pcarruscag · 2021-01-18T15:51:36Z

SU2_CFD/src/solvers/CNEMOEulerSolver.cpp

+      if (nodes->GetSymmetry(iPoint))
+         if(nodes->GetSymmetry(iPoint) == 1 && nodes->GetSymmetry(jPoint)) nodes->AddMax_Lambda_Visc(iPoint, 4*Lambda);
+         else if (nodes->GetSymmetry(iPoint) == 2 && nodes->GetSymmetry(jPoint) == 1) nodes->AddMax_Lambda_Visc(iPoint, 8*Lambda);
+         else if (nodes->GetSymmetry(iPoint) == 2 && nodes->GetSymmetry(jPoint) == 2) nodes->AddMax_Lambda_Visc(iPoint, 16*Lambda);
+         else nodes->AddMax_Lambda_Visc(iPoint, 2*Lambda);
+      else nodes->AddMax_Lambda_Visc(iPoint, Lambda);


Maybe you can explain this in the next dev meeting?
This looks so complex that there has to be a simpler way.

@pcarruscag Sure, when will the next dev meeting be ?

The idea behind it is to mimic the eigenvalues given if using a "full" domain. Probably there is a simpler way through manipulation of the Normal values, but not sure how it can be implemented.

Would doubling at the end be enough?

Wednesdays at 4pm CET https://meet.jit.si/SU2_DevMeeting

If the nodes are in a single symmetry plane, yes. A bit more complex when they are in two symmetry planes, but I will look at it.

WallyMaier

@fmpmorgado thanks for doing this work! The results are very promising, I think you should add a couple of pictures here to showcase the improvement.

I made a comment about interp, but i think looks pretty good.

SU2_CFD/src/variables/CNEMOEulerVariable.cpp

jtneedels

Hi Fabio, thanks for implementing this, it looks really good. Yeah I would second what Wally said, maybe posting the images you showed in the slack as a comment, just to show the difference between the current and former implementations.

fmpmorgado · 2021-01-19T16:32:15Z

For the sake of demonstrating the capability of the new Symmetry Boundary condition, the image shows a comparison of pressure at the surface, using the new implementation and the old symmetry boundary condition, for a flow over a 3D cylinder and considering the lateral and bottom surfaces as symmetry plane:

fmpmorgado · 2021-01-21T22:13:25Z

After a meeting with @TobiKattmann, I have decided to change NEMO BC_EULER_WALL to match the one in CEulerSolver.cpp, where the velocity vector and gradients are corrected.

Euler boundary condition is now called inside the symmetry plane boundary condition, before performing the residual correction
This way, I believe I am assuring for all the vector quantities to be parallel to the symmetry plane and guaranteeing the condition of symmetry according to Jiri Blazek's - "Computational fluid dynamics: principles and applications" . However, calling Euler function does not appear to substantially change the the solution substantially in the test I have made so far.

TobiKattmann

@fmpmorgado and I chatted and decided to try out the Version Blazek describes in his book (8.6 in the 3rd edition).

That will also require changing some normal vectors of faces that end on a sym plane (tbd once after mesh reading) and removing components of the gradient for symmetry nodes (tbd each time they get computed). This is not done yet in this PR.

The changes are straight forward once these two mentioned things are done as the sym_plane bc becomes rather short as it just removes the normal residual component (and nothing else). And the Euler wall stays as is so the code needs to be moved over (i.e. again separate implementations for Sym and euler wall). Keeping the viscous stuff has the potential to serve as a slip wall for viscuos flows (if I am not mistaken).

Putting changes to the non-NEMO solvers in this PR is s.th. we could do as the normal-vector and gradient changes should be equal but it would delay the whole PR and a lot of Reg-tests would have changes. AND we have to see whether we have problems in the non-NEMO solvers that would be fixed with this in the first place (correct me if we already know that -> I will test that).

Together with @talbring (thanks!) I briefly looked into the code that Jiri Blazek provides for his book and he implements the boundary he describes (... who would have thought).

TobiKattmann · 2021-01-21T23:17:41Z

SU2_CFD/src/solvers/CNEMOEulerSolver.cpp

+        Jacobian.AddBlock2Diag(iPoint, residual.jacobian_i);
+      }
+
+      if (viscous) {


The viscous part is superflous. (unless you need a slip wall I guess)

Yes, you are right, I have removed the viscous part.

TobiKattmann · 2021-01-21T23:20:10Z

SU2_CFD/src/solvers/CNEMOEulerSolver.cpp

+
+  SU2_OMP_FOR_DYN(OMP_MIN_SIZE)
+  for (iVertex = 0; iVertex < geometry->nVertex[val_marker]; iVertex++) {
+    if (!preprocessed || geometry->bound_is_straight[val_marker] != true) {


If you delete the viscous part then the tangential-vector computation here is not necessary as well and the section can be simplified.
The bound_is_straight container + evaluation could be deleted in that case as well (unless we want to keep for future ideas.)

TobiKattmann · 2021-01-21T23:27:22Z

SU2_CFD/src/solvers/CNEMOEulerSolver.cpp

-  su2double *Normal = nullptr, Area, UnitNormal[3], *NormalArea,
-  **Jacobian_b, **DubDu,
-  rho, cs, P, rhoE, rhoEve, conc, *u, *dPdU;
+  BC_Euler_Wall(geometry, solver_container, conv_numerics, visc_numerics, config, val_marker);


Euler boundary condition is now called inside the symmetry plane boundary condition, before performing the residual correction
This way, I believe I am assuring for all the vector quantities to be parallel to the symmetry plane and guaranteeing the condition of symmetry according to Jiri Blazek's - "Computational fluid dynamics: principles and applications" . However, calling Euler function does not appear to substantially change the the solution substantially in the test I have made so far.

Imo this is not necessary and that part can be deleted (until Jiri's suggested implementation shows problems).

pcarruscag · 2021-01-22T00:23:49Z

Putting changes to the non-NEMO solvers in this PR is s.th. we could do as the normal-vector and gradient changes should be equal but it would delay the whole PR and a lot of Reg-tests would have changes

They are exactly the same type of solver, we should not have two types of symmetry.

…ler call at symmetry BC

pcarruscag · 2021-01-22T15:00:32Z

SU2_CFD/src/output/CNEMOCompOutput.cpp

-  if ((config->GetKind_Solver() == NEMO_NAVIER_STOKES)) {
+  if (config->GetKind_Solver() == NEMO_NAVIER_STOKES && config->GetViscous_Wall(iMarker)) {


I would also remove the check for NEMO_NAVIER_STOKES... This kind of logic causes 90% of the bugs in SU2 every time a new options is added something breaks, the simpler the logic, the easier it is to maintain the code.

pcarruscag · 2021-01-22T15:36:41Z

SU2_CFD/src/solvers/CNEMOEulerSolver.cpp

+  /*--- Correct normal directions of edges ---*/
+  for (unsigned long iMarker = 0; iMarker < geometry->GetnMarker(); iMarker++) {
+    if (config->GetMarker_All_KindBC(iMarker) == SYMMETRY_PLANE){


We should probably move this to draft PR while you guys prototype the new symmetry implementation.
Then we can discuss how and where to implement the different changes without copy-pasting code and making sure the fix is compatible with other features of the code.

stale · 2021-06-02T16:07:46Z

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. If this is still a relevant issue please comment on it to restart the discussion. Thank you for your contributions.

CatarinaGarbacz and others added 8 commits January 12, 2021 22:34

merge with latest developments in other NEMO branches

3e756c3

Changes on SYM_BC. Working with 2D cases

22ea138

pending merge

5652711

2D fully working. Missing 3D still

f7ed768

fix SU2 INTERP stuff after merge

cab2b59

Working in 3D, missing comments

85f1396

frozen mixture option also working with mutation++

9fb212c

Re-organized code and placed some comments

120d6cf

pr-triage bot added the PR: unreviewed label Jan 18, 2021

fmpmorgado requested review from WallyMaier, CatarinaGarbacz and jtneedels January 18, 2021 12:49

Merge branch 'feature_NEMO' into feature_NEMO_Sym_BC

2761f67

CatarinaGarbacz approved these changes Jan 18, 2021

View reviewed changes

SU2_CFD/src/variables/CNEMOEulerVariable.cpp Outdated Show resolved Hide resolved

Solved conflict issues and misleading comments

eb5b981

pcarruscag reviewed Jan 18, 2021

View reviewed changes

WallyMaier reviewed Jan 18, 2021

View reviewed changes

SU2_CFD/src/variables/CNEMOEulerVariable.cpp Outdated Show resolved Hide resolved

jtneedels approved these changes Jan 18, 2021

View reviewed changes

pr-triage bot added PR: reviewed-approved and removed PR: unreviewed labels Jan 18, 2021

Simplification of if conditions to calculate the used eigenvalue

b86663c

pr-triage bot added PR: unreviewed and removed PR: reviewed-approved labels Jan 19, 2021

Simplifying conditions in time-step calculation

84da6e2

recovering some forgotten stuff

201a869

WallyMaier approved these changes Jan 19, 2021

View reviewed changes

pr-triage bot added PR: reviewed-approved and removed PR: unreviewed PR: reviewed-approved labels Jan 19, 2021

pr-triage bot added PR: reviewed-changes-requested and removed PR: unreviewed labels Jan 21, 2021

Modified Euler_wall_bc to be the same as in CEulerSolver

de47148

pr-triage bot added PR: unreviewed and removed PR: reviewed-changes-requested labels Jan 21, 2021

Fixing CIntegration

bd13010

TobiKattmann requested changes Jan 21, 2021

View reviewed changes

pr-triage bot added PR: reviewed-changes-requested and removed PR: unreviewed labels Jan 21, 2021

Created routine to fix Edge Normals at Symmetry Planes and removed Eu…

3d96408

…ler call at symmetry BC

pr-triage bot added PR: unreviewed and removed PR: reviewed-changes-requested labels Jan 22, 2021

Addressing PR requested changes and Remove viscous part in Euler wall BC

f8426e5

pcarruscag reviewed Jan 22, 2021

View reviewed changes

fmpmorgado marked this pull request as draft January 23, 2021 17:39

pcarruscag changed the title ~~NEMO - New Symmetry BC~~ [WIP] NEMO - New Symmetry BC Feb 8, 2021

pr-triage bot added PR: draft and removed PR: unreviewed labels Feb 8, 2021

Merge branch 'develop' into feature_NEMO_Sym_BC

acfec29

pcarruscag mentioned this pull request May 7, 2021

Different results between 3D and 2D problems when using JST #1280

Closed

stale bot added the stale label Jun 2, 2021

stale bot closed this Jun 9, 2021

This was referenced May 10, 2022

NEMO symmetry boundary condition problem #1625

Closed

[WIP] Fix symmetry plane issues in SU2, change SU2-NEMO sym plane to match flow solver #1635

Closed

ShiheJia mentioned this pull request Nov 22, 2023

Set the symmetry BC mandatorily, only for velocity. #2174

Closed

6 tasks

bigfooted mentioned this pull request Jan 3, 2024

Correction of symmetry plane implementation #2194

Merged

6 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[WIP] NEMO - New Symmetry BC #1168

[WIP] NEMO - New Symmetry BC #1168

fmpmorgado commented Jan 18, 2021 •

edited

Loading

CatarinaGarbacz left a comment

pcarruscag Jan 18, 2021

fmpmorgado Jan 18, 2021

pcarruscag Jan 18, 2021

fmpmorgado Jan 18, 2021

WallyMaier left a comment

jtneedels left a comment

fmpmorgado commented Jan 19, 2021 •

edited

Loading

fmpmorgado commented Jan 21, 2021 •

edited

Loading

TobiKattmann left a comment

TobiKattmann Jan 21, 2021

fmpmorgado Jan 22, 2021

TobiKattmann Jan 21, 2021

TobiKattmann Jan 21, 2021

pcarruscag commented Jan 22, 2021

pcarruscag Jan 22, 2021

pcarruscag Jan 22, 2021

stale bot commented Jun 2, 2021

		if ((config->GetKind_Solver() == NEMO_NAVIER_STOKES)) {
		if (config->GetKind_Solver() == NEMO_NAVIER_STOKES && config->GetViscous_Wall(iMarker)) {

[WIP] NEMO - New Symmetry BC #1168

[WIP] NEMO - New Symmetry BC #1168

Conversation

fmpmorgado commented Jan 18, 2021 • edited Loading

Proposed Changes

Related Work

PR Checklist

Results

New Implementation vs Old Implementation for N2 mixture (2nd Order - NEMO_NS solver - AUSM scheme)

New Implementation (on NEMO) vs Current SU2 Implementation for AIR5 mixture (2nd Order - NS solver - AUSM scheme)

CatarinaGarbacz left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

WallyMaier left a comment

Choose a reason for hiding this comment

jtneedels left a comment

Choose a reason for hiding this comment

fmpmorgado commented Jan 19, 2021 • edited Loading

fmpmorgado commented Jan 21, 2021 • edited Loading

TobiKattmann left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

pcarruscag commented Jan 22, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

stale bot commented Jun 2, 2021

fmpmorgado commented Jan 18, 2021 •

edited

Loading

fmpmorgado commented Jan 19, 2021 •

edited

Loading

fmpmorgado commented Jan 21, 2021 •

edited

Loading