Explicit Euler in CScalarSolver #1435

maxaehle · 2021-11-11T17:33:02Z

Proposed Changes

Enable time integration by the explicit Euler scheme for e.g. turbulence solvers, by implementing CScalarSolver<>::ExplicitEuler_Iteration.

Related Work

This fixes #1432 with respect to explicit Euler. I need the feature for further investigation of #1414. CScalarSolver was introduced in #1330.

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.

PrepareExplicitIteration. It's a mixture of the implicit Euler in CScalarSolver and the explicit Euler in CHeatSolver.

maxaehle · 2021-11-11T17:36:19Z

The added lines are a combination of the explicit Euler in CHeatSolver and the implicit Euler in CScalarSolver. Do you consider this approach to be correct?

For the testcase SU2/TestCases/rans/naca0012/turb_NACA0012_sst.cfg, explicit Euler for both flow and turbulence equations and CFL=0.1 converges from the restart solution.

Some more details:

CHeatSolver does something with CVariable::Res_TruncError, which is related to multigrid. Maybe additional adaptations are necessary here.
PrepareExplicitIteration was copied from PrepareImplicitIteration, except for the computation of the matrix of the linear system. We can remove duplicate code here.
CompleteImplicitIteration could be renamed as "add LinSysSol to solution"
.

SU2_CFD/include/solvers/CScalarSolver.inl

pcarruscag

Thanks for taking the hard route to fix this 👍
I think RK4 is going to be harder to implement as it would involve changes to CSingleGridIntegration::SingleGrid_Iteration if you don't go for it please add an error somewhere in config postprocessing.

SU2_CFD/include/solvers/CScalarSolver.inl

SU2_CFD/include/solvers/CScalarSolver.hpp

SU2_CFD/include/solvers/CScalarSolver.inl

pcarruscag · 2021-11-11T19:42:39Z

SU2_CFD/include/solvers/CScalarSolver.inl

+  }
+  END_SU2_OMP_FOR
+
+  CompleteImplicitIteration(geometry, solver_container, config);


This makes a lot of sense given the logic needed to handle SA v SST idiosyncrasies 👍

Co-authored-by: Pedro Gomes <38071223+pcarruscag@users.noreply.github.com>

PrepareExplicitIteration was mainly about computation of residual norms. We moved this code into two functions called from ExplicitEuler_Iteration, and also called from ImplicitEuler_Iteration now (removing duplicate code).

and use it from other locations in the code as well.

Sometimes the residual norm is computed from residuals other than LinSysRes.

adapted from turb_NACA0012_sst_fixedvalues.cfg I will fill in the residuals of the CI pipeline in the next commit. Note that these tests only check that the behavior of the explicit Euler scheme is constant, but not that it is correct (i.e., it is the explicit Euler scheme) at the time of this commit.

With this cfg, serial SU2 converges to rms[Rho]<-12 in iteration 2053191,

for the new testcase introduced in 022b7c0

pcarruscag

Nice work 👍 Good idea with the additional residual reduction cleanup.

Common/src/CConfig.cpp

SU2_CFD/include/solvers/CFVMFlowSolverBase.hpp

SU2_CFD/include/solvers/CSolver.hpp

TestCases/serial_regression.py

TestCases/rans/naca0012/turb_NACA0012_sst_expliciteuler.cfg

TobiKattmann

Thanks Max for this 💐 looks really neat and kudos for the additional generalizations.
My comments are mostly/only nitpicking stylistic things so nothing major. Only for the curly brace init I would like to get an opinion of pedro

Once develop is merged (and you fix the CodeFactor warning 😄 ) I guess this is good to go then

Common/src/CConfig.cpp

SU2_CFD/include/solvers/CScalarSolver.inl

SU2_CFD/include/solvers/CSolver.hpp

SU2_CFD/src/solvers/CFEASolver.cpp

TestCases/rans/naca0012/turb_NACA0012_sst_expliciteuler.cfg

Co-authored-by: TobiKattmann <31306376+TobiKattmann@users.noreply.github.com>

see C.128 in the C++ Core Guidelines suggested by CodeFactor

TobiKattmann · 2021-11-18T16:36:08Z

@maxaehle Can you please also check whether there are problems with periodic BC's? What makes me believe that is CTurbSASolver+167 (and similar for SST)

  /*--- The turbulence models are always solved implicitly, so set the
   implicit flag in case we have periodic BCs. ---*/

  SetImplicitPeriodic(true);

pcarruscag · 2021-11-21T22:43:38Z

@TobiKattmann was right about the periodic comms.

TobiKattmann · 2021-11-22T12:33:13Z

SU2_CFD/src/solvers/CTurbSolver.cpp

+CTurbSolver::CTurbSolver(CGeometry* geometry, CConfig *config, bool conservative)
+  : CScalarSolver<CTurbVariable>(geometry, config, conservative) {
+  /*--- Store if an implicit scheme is used, for use during periodic boundary conditions. ---*/
+  SetImplicitPeriodic(config->GetKind_TimeIntScheme_Turb() == EULER_IMPLICIT);
+}


👍 thanks for checking, then I'll do the same for the SpeciesSolver

Which introduces explicit euler for the scalar solvers. Additional smaller ctor cleanups.

TobiKattmann · 2021-11-23T11:21:23Z

@maxaehle as far as I see there are no further comments from pedro or me, so if you merge develop + pass the reg test feel free to merge this. (of course if you also have no further plans with this PR)
Thanks again for this contribution!

…ler_scalarsolver

Add CScalarSolver<>::ExplicitEuler_Iteration,

d7f73f7

PrepareExplicitIteration. It's a mixture of the implicit Euler in CScalarSolver and the explicit Euler in CHeatSolver.

remove unused variable

61392cc

pcarruscag reviewed Nov 11, 2021

View reviewed changes

SU2_CFD/include/solvers/CScalarSolver.inl Outdated Show resolved Hide resolved

pcarruscag reviewed Nov 11, 2021

View reviewed changes

pcarruscag added the changelog:fix label Nov 11, 2021

maxaehle and others added 6 commits November 12, 2021 07:34

Apply suggestions from code review

976222d

Co-authored-by: Pedro Gomes <38071223+pcarruscag@users.noreply.github.com>

Merge PrepareExplicitIteration into ExplicitEuler_Iteration

dd69a15

PrepareExplicitIteration was mainly about computation of residual norms. We moved this code into two functions called from ExplicitEuler_Iteration, and also called from ImplicitEuler_Iteration now (removing duplicate code).

Move residual norm calculation helper to CSolver

aab916f

and use it from other locations in the code as well.

Fix error introduced in aab916f

b9eb7bb

Sometimes the residual norm is computed from residuals other than LinSysRes.

Error if selected TIME_DISCRE_FLOW not impl/expl Euler

35193ce

pcarruscag linked an issue Nov 12, 2021 that may be closed by this pull request

Explicit Euler for Turbulence Solver #1432

Closed

maxaehle added 2 commits November 12, 2021 13:20

CFL=0.1 instead of 0.5 in explicit Euler testcase

c5039ef

With this cfg, serial SU2 converges to rms[Rho]<-12 in iteration 2053191,

Set "test_vals (stored)" to current "sim_vals (computed)"

fcccf16

for the new testcase introduced in 022b7c0

maxaehle changed the title ~~[WIP] Explicit Euler in CScalarSolver~~ Explicit Euler in CScalarSolver Nov 12, 2021

pr-triage bot added the PR: unreviewed label Nov 12, 2021

maxaehle added 2 commits November 13, 2021 08:15

make ResidualReductions_PerThread static

27ba505

update config_template.cfg

aa04688

pcarruscag approved these changes Nov 13, 2021

View reviewed changes

pr-triage bot added PR: reviewed-approved and removed PR: unreviewed labels Nov 13, 2021

Minor spacing fixes

96090bc

pr-triage bot added PR: unreviewed and removed PR: reviewed-approved labels Nov 13, 2021

maxaehle added 5 commits November 17, 2021 08:54

Move ResidualReductions_* to protected part

d20b2c5

Make residual modifications in CSolver protected, too

296e757

remove descriptions of options in turb_NACA0012_sst_expliciteuler.cfg

46542d9

remove/correct options in turb_NACA0012_sst_expliciteuler.cfg

b6c7620

remove descriptions of options in turb_NACA0012_sst_fixedvalues.cfg

30af929

TobiKattmann approved these changes Nov 18, 2021

View reviewed changes

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

Apply suggestions from code review

5360ece

Co-authored-by: TobiKattmann <31306376+TobiKattmann@users.noreply.github.com>

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

maxaehle added 2 commits November 18, 2021 10:05

clang-format

bb09595

don't mark functions as both virtual and override

217db02

see C.128 in the C++ Core Guidelines suggested by CodeFactor

pcarruscag and others added 2 commits November 21, 2021 22:32

Merge branch 'develop' into feature_expliciteuler_scalarsolver

5a184a7

do not assume implicit for turbulence

f7cfaeb

TobiKattmann reviewed Nov 22, 2021

View reviewed changes

TobiKattmann added a commit that referenced this pull request Nov 22, 2021

Dont assume implicit solving for periodic BC's. See PR #1435.

d455298

Which introduces explicit euler for the scalar solvers. Additional smaller ctor cleanups.

Merge remote-tracking branch 'origin/develop' into feature_expliciteu…

27a7687

…ler_scalarsolver

maxaehle merged commit e3fa7d9 into develop Nov 23, 2021

maxaehle deleted the feature_expliciteuler_scalarsolver branch November 23, 2021 12:32

pr-triage bot added PR: merged and removed PR: unreviewed labels Nov 23, 2021

maxaehle mentioned this pull request May 2, 2022

Divergence when using fluid interface #1414

Open

Explicit Euler in CScalarSolver #1435

Explicit Euler in CScalarSolver #1435

Uh oh!

Conversation

maxaehle commented Nov 11, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Proposed Changes

Related Work

PR Checklist

Uh oh!

maxaehle commented Nov 11, 2021

Uh oh!

Uh oh!

pcarruscag left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

pcarruscag Nov 11, 2021

Choose a reason for hiding this comment

Uh oh!

pcarruscag left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

TobiKattmann left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

TobiKattmann commented Nov 18, 2021

Uh oh!

pcarruscag commented Nov 21, 2021

Uh oh!

TobiKattmann Nov 22, 2021

Choose a reason for hiding this comment

Uh oh!

TobiKattmann commented Nov 23, 2021

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

maxaehle commented Nov 11, 2021 •

edited

Loading