Heat solver fixes for primal and adjoint CHT simulations #1107

pcarruscag · 2020-11-17T19:16:37Z

Proposed Changes

I think one sign was flipped in the Jacobian of the CHT boundary.
I replaced the isothermal BC for a stronger(?) alternative (flying by the seat of my pants here, if it's wrong let me know, but the solver converges better like this).
Fixes issues with deforming heat domains by generalising CDiscAdjMeshSolver::SetSensitivity to accept any solver as "target" (instead of considering always the ADJFLOW_SOL position in the solver container).

Related Work

#951
#1061
(third time lucky I hope...)

PR Checklist

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.

pcarruscag · 2020-11-17T19:19:05Z

SU2_CFD/src/solvers/CHeatSolver.cpp

+          if (implicit) {
+
+            Jacobian_i[0][0] = -thermal_diffusivity*Area;
+            Jacobian.SubtractBlock2Diag(iPoint, Jacobian_i);
+          }


The sign, I also moved this so that it only applies to one of the coupling modes, to be consistent with the implementation of the heat flux wall.

pcarruscag · 2020-11-17T19:19:53Z

SU2_CFD/src/solvers/CHeatSolver.cpp

+  auto iter = System.Solve(Jacobian, LinSysRes, LinSysSol, geometry, config);
+  SetIterLinSolver(iter);
+  SetResLinSolver(System.GetResidual());


Ah yes, it also fixes the output of the linear solver stuff.

…1104.

pcarruscag · 2020-11-19T14:39:49Z

@TobiKattmann since we were talking about it yesterday, the advantage of the "weak" formulation is that it is conservative, heat fluxin equal heat flux out (because the post processed fluxes are consistent with how the boundary condition is applied).
Obviously at the expense of the isothermal temperatures not being imposed exactly.
With the "strong" formulation, conservation can probably only be achieved if the mesh topology of the inlet/outlet boundary is the same, or at grid convergence.
Learning something new everyday, I put it back as it was.

…ver_fixes

…to heat_solver_fixes

WallyMaier

@pcarruscag Thanks for the contribution!

I looked the files changes and it looks good to me!

TobiKattmann

Thanks pedro, I did not test the jacobian influence yet but will so do so soon.
(This Approve excludes my bits in here of course as that is not exactly the idea of a PR-review :D )

TobiKattmann · 2020-11-24T13:40:16Z

SU2_CFD/include/solvers/CDiscAdjFEASolver.hpp

-   * \param[in] solver - The solver container holding all terms of the solution.
   * \param[in] config - Definition of the particular problem.
   */
-  void SetSensitivity(CGeometry *geometry, CSolver **solver, CConfig *config) override;
+  void SetSensitivity(CGeometry *geometry, CConfig *config, CSolver*) override;


took me a second to figure out what happens here with the override function in the child and the default argument in the parent class. Kind of odd that the override function understands that it has to use the default argument of the parent implementation... I hope I understood that correctly (I used Compiler Explorer to make a sample)

You can think of it as the default value being part of the function signature, the signature is defined by the original virtual function which is all a caller needs to know (you include CIteration.hpp to use all the other iterations), the default value is passed when calling, not used when "receiving" the call.

TobiKattmann · 2020-11-24T13:52:36Z

SU2_CFD/src/drivers/CDiscAdjMultizoneDriver.cpp


          // Other surface-related output values
+          // The names are different than in CConfig...


Here you mean that the Output has a different name for an Objective than the map Objective_Map in option_structure provides? thats why you cannot simply use the name of the objective_function that is stored in the config

Yes... those name choices were not very good, it needs to be fixed at some point, but it will have implications for the python stuff.

Oh yes that historyMap.py... maybe automatically create that idk? But that is something for another day/PR :D

SU2_CFD/src/solvers/CDiscAdjMeshSolver.cpp

TobiKattmann · 2020-11-24T14:33:57Z

Common/include/CConfig.hpp

-   * \return Kind of objective function.
+   * \brief Similar to GetKind_ObjFunc but returns the corresponding string.
   */
-  unsigned short GetKind_ObjFunc(unsigned short val_obj) const { return Kind_ObjFunc[val_obj]; }
+  string GetName_ObjFunc(unsigned short val_obj = 0) const;


for completeness can we get the param[in] and return here as well

sorry I just find that kind of description too much fluff, one of my pet peeves with this doxy stuff is that a ton of methods describe the same arguments over and over ad nauseam and they say very little of the actual method. Less is more is what I'm getting at.

For example:
* \brief A virtual member.
"No **** Sherlock, I could see that from just the code..." The important is what the method should do.
Anyway, rant over.

SU2_CFD/include/solvers/CSolver.hpp

SU2_CFD/src/iteration/CHeatIteration.cpp

pcarruscag · 2020-12-01T16:54:59Z

Merging this to allow all tests to run so we can create a release tomorrow (there is a fix in this branch for compilation with gcc 5.3.0 so this should be part of the release). Happy to make post mortem changes if needed.

pcarruscag added 3 commits November 16, 2020 23:21

fix CHT boundary jacobian, stronger isothermal wall bc

e800a3c

fixes for disc adj with mesh deformation

afad01e

allow the disc adj mesh solver to set sensitivities on any solver

df1928a

pr-triage bot added the PR: unreviewed label Nov 17, 2020

Merge branch 'develop' into heat_solver_fixes

ff99862

pcarruscag commented Nov 17, 2020

View reviewed changes

Merge branch 'nan_checks_better_defaults' into heat_solver_fixes

b82cf89

pcarruscag added the changelog:fix label Nov 17, 2020

pcarruscag and others added 6 commits November 17, 2020 23:05

fix force and moment coeff names, add bgs outputs for discadj fea

8f040d4

Merge branch 'develop' into heat_solver_fixes

109f37b

Fix Mesh_Quality volume output for incompressible flow introduced in #…

2ebf8f0

…1104.

config method to get the obj fun name (to pass it to the output)

54ee7cc

fix singlezone heat issues, reduce duplication

a004f5d

revert changes to isothermal boundary

0d48aca

TobiKattmann added 2 commits November 19, 2020 17:30

Fix/Enable MESH_QUALITY output for SU2_DEF.

07cc322

Merge remote-tracking branch 'origin/heat_solver_fixes' into heat_sol…

85419fe

…ver_fixes

TobiKattmann mentioned this pull request Nov 20, 2020

MESH_QUALITY output broken for some cases #1109

Closed

pcarruscag added 3 commits November 20, 2020 17:40

update regressions

a16301b

maybe fix the isnan thing

20e012f

Merge branch 'heat_solver_fixes' of https://github.com/su2code/SU2 in…

b6e0398

…to heat_solver_fixes

TobiKattmann mentioned this pull request Nov 20, 2020

Compilation fails with gcc5.3.0 due to isnan() ambiguity #1112

Closed

forgot 1 testcase

6ed646f

WallyMaier approved these changes Nov 23, 2020

View reviewed changes

pr-triage bot added PR: reviewed-approved and removed PR: unreviewed labels Nov 23, 2020

TobiKattmann approved these changes Nov 24, 2020

View reviewed changes

pcarruscag added 2 commits November 25, 2020 10:21

Merge remote-tracking branch 'upstream/develop' into heat_solver_fixes

63f7af0

address some comments

3437f2b

pr-triage bot added PR: unreviewed and removed PR: reviewed-approved labels Nov 25, 2020

pcarruscag merged commit 78363e1 into develop Dec 1, 2020

pcarruscag deleted the heat_solver_fixes branch December 1, 2020 16:55

pr-triage bot added PR: merged and removed PR: unreviewed labels Dec 1, 2020

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Heat solver fixes for primal and adjoint CHT simulations #1107

Heat solver fixes for primal and adjoint CHT simulations #1107

pcarruscag commented Nov 17, 2020

pcarruscag Nov 17, 2020

pcarruscag Nov 17, 2020

pcarruscag commented Nov 19, 2020 •

edited

Loading

WallyMaier left a comment

TobiKattmann left a comment

TobiKattmann Nov 24, 2020

pcarruscag Nov 25, 2020

TobiKattmann Nov 24, 2020

pcarruscag Nov 25, 2020

TobiKattmann Nov 25, 2020

TobiKattmann Nov 24, 2020

pcarruscag Nov 25, 2020

pcarruscag Nov 25, 2020

pcarruscag commented Dec 1, 2020


		// Other surface-related output values
		// The names are different than in CConfig...

Heat solver fixes for primal and adjoint CHT simulations #1107

Heat solver fixes for primal and adjoint CHT simulations #1107

Conversation

pcarruscag commented Nov 17, 2020

Proposed Changes

Related Work

PR Checklist

Choose a reason for hiding this comment

Choose a reason for hiding this comment

pcarruscag commented Nov 19, 2020 • edited Loading

WallyMaier left a comment

Choose a reason for hiding this comment

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

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 Dec 1, 2020

pcarruscag commented Nov 19, 2020 •

edited

Loading