Feature ddes #490

EduardoMolina · 2017-12-26T13:55:09Z

Dear SU2 developers,

This PR is a long effort to add Hybrid RANS/LES capabilities, specially the Delayed Detached Eddy Simulation with Grey area mitigation and low dissipation schemes proper to unsteady flows.

Summary of the contribution:

Delayed DES:
- Standard SGS
- Vorticity SGS
- Shear Layer Adapted SGS
Several low dissipation functions:
- DDES FD
- Ducros' shock sensor
- NTS adaptive
Variants of SA turbulence model:
- Compressibility correction
- Edwards and Chandra
- QCR 2000 (Quadratic constitutive relation)
Low dissipation schemes:
- Low dissipation Roe.
- Low Mach Roe.
- Simple low dissipation AUSM (SLAU and SLAU2)

I added only a small regression test for 2D DDES, travis can not handle even the smallest 3D test case that I have. Suggestions are welcome.

Best,

Eduardo

First commit of Hybrid RANS/LES

- Roe Low Dissipation: (1 - f_d) + Ducros Sensor - Turbulence models: Edward’s SA, Compressibility Correction SA, Edward’s + CC. - In SA_DDES, the low Reynolds term was disabled. - Start of Venkatakrishnan Limiter Modification.

# Conflicts: # Common/include/config_structure.hpp # SU2_CFD/include/numerics_structure.hpp # SU2_CFD/src/SU2_CFD.cpp # SU2_CFD/src/driver_structure.cpp # SU2_CFD/src/numerics_direct_mean.cpp

# Resolved conflicts in: # Common/include/config_structure.hpp # SU2_CFD/include/numerics_structure.hpp # SU2_CFD/src/SU2_CFD.cpp # SU2_CFD/src/driver_structure.cpp # SU2_CFD/src/numerics_direct_mean.cpp

- Now ROW_LOW_DISSIPATION flag has the following options: - FD: Numerical blending using the DDES’s Fd function. - NTS: Numerical blending of Travin and Shur. - NTS_DUCROS: NTS + Ducros’ Shock Sensor. All these options must be tested extensively.

- Update Ducros Sensor. - Added a modified version of Venkatakrishnan Limiter.

Preventing the limiter to be greater than one.

It is incomplete right now.

- Add Weiss & Smith Preconditioner + Thornber Low Mach Correction

- Need to fix some conflicts.

…o feature_DDESv5.0

economon

Wow, long effort, indeed. Thanks @EduardoMolina! The community will really enjoy the new DES capabilities, low dissipation schemes, and S-A variants.

I've put a number of comments in the review on items to be cleaned up, but structurally, things look good so far. Can you also please check/remove the couple of Apple time machine files that are referenced in the change log?

economon · 2018-01-04T07:57:55Z

.travis.yml

@@ -12,11 +12,12 @@ compiler:
 notifications:
    email:
        recipients:
-            - [email protected]
+            - [email protected]


Don't forget to revert the .travis.yml file.

economon · 2018-01-04T08:00:14Z

.travis.yml

@@ -80,7 +81,7 @@ install:

 before_script:
    # Get the test cases
-    - git clone -b develop https://github.com/su2code/TestCases.git ./TestData
+    - git clone -b feature_DDES https://github.com/su2code/TestCases.git ./TestData


Please also submit your mesh to the develop branch in the TestCases repo.

economon · 2018-01-04T08:02:05Z

Common/include/config_structure.hpp

+
+  /*!
+   * \brief Get the Kind of Hybrid RANS/LES.
+   * \return Verbosity level for the console output.


Please update the "return" comments for these next few getter methods.

economon · 2018-01-04T08:08:42Z

Common/include/option_structure.hpp

 };
 static const map<string, ENUM_LIMITER> Limiter_Map = CCreateMap<string, ENUM_LIMITER>
 ("NONE", NO_LIMITER)
 ("VENKATAKRISHNAN", VENKATAKRISHNAN)
 ("VENKATAKRISHNAN_WANG", VENKATAKRISHNAN_WANG)
+("VENKATAKRISHNAN_2NDLIM", VENKATAKRISHNAN_2NDLIM)


Perhaps this name is a little unclear..

economon · 2018-01-04T08:11:23Z

SU2_IDE/Xcode/SU2_CFD.xcodeproj/project.pbxproj

@@ -161,7 +161,7 @@
 		05E6DBB817EB627400FA1F7E /* output_structure.hpp */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.cpp.h; lineEnding = 0; name = output_structure.hpp; path = ../../SU2_CFD/include/output_structure.hpp; sourceTree = "<group>"; xcLanguageSpecificationIdentifier = xcode.lang.cpp; };
 		05E6DBB917EB627400FA1F7E /* solver_structure.hpp */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.cpp.h; lineEnding = 0; name = solver_structure.hpp; path = ../../SU2_CFD/include/solver_structure.hpp; sourceTree = "<group>"; xcLanguageSpecificationIdentifier = xcode.lang.cpp; };
 		05E6DBBA17EB627400FA1F7E /* SU2_CFD.hpp */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.cpp.h; lineEnding = 0; name = SU2_CFD.hpp; path = ../../SU2_CFD/include/SU2_CFD.hpp; sourceTree = "<group>"; xcLanguageSpecificationIdentifier = xcode.lang.cpp; };
-		05E6DBBB17EB627400FA1F7E /* variable_structure.hpp */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.cpp.h; lineEnding = 0; name = variable_structure.hpp; path = ../../SU2_CFD/include/variable_structure.hpp; sourceTree = "<group>"; xcLanguageSpecificationIdentifier = xcode.lang.cpp; };
+		05E6DBBB17EB627400FA1F7E /* variable_structure.hpp */ = {isa = PBXFileReference; indentWidth = 2; lastKnownFileType = sourcecode.cpp.h; lineEnding = 0; name = variable_structure.hpp; path = ../../SU2_CFD/include/variable_structure.hpp; sourceTree = "<group>"; tabWidth = 2; xcLanguageSpecificationIdentifier = xcode.lang.cpp; };


Looks like the Xcode project had a small change. Try to revert if you can.

economon · 2018-01-04T08:56:51Z

SU2_CFD/src/numerics_direct_mean.cpp

+
+}
+
+void CUpwRoe_Flow::ComputeResidual(su2double *val_residual, su2double **val_Jacobian_i, su2double **val_Jacobian_j, CConfig *config) {


Can you please check the changes in this file for the various upwind schemes? It seems to think that the entire subroutines were changed, when I think that the ordering or only white space changed. I suspect the true number of changed lines should be less.

economon · 2018-01-04T09:00:16Z

SU2_CFD/src/numerics_direct_mean.cpp

+  /*--- L2Roe: a low dissipation version of Roe's approximate Riemann solver for low Mach numbers. IJNMF 2015 ---*/
+
+  zeta = min(1.0,max(Mach_i,Mach_j));
+  //zeta = max(zeta,0.05);


Remove this comment if unnecessary.

economon · 2018-01-04T09:08:43Z

SU2_CFD/src/numerics_direct_turbulent.cpp

+
+    /*--- Evaluate Omega ---*/
+
+    //Omega = sqrt(Vorticity_i[0]*Vorticity_i[0] + Vorticity_i[1]*Vorticity_i[1] + Vorticity_i[2]*Vorticity_i[2]);


Remove commented line if unnecessary.

economon · 2018-01-04T09:10:34Z

SU2_CFD/src/numerics_direct_turbulent.cpp

+        /*--- Production term ---*/;
+
+        //    Original SA model
+        //    Production = cb1*(1.0-ft2)*Shat*TurbVar_i[0]*Volume;


In general, please remove the commented lines from the original S-A model to keep the implementation as clean as possible (especially for new users).

economon · 2018-01-04T09:29:36Z

SU2_CFD/src/solver_direct_mean.cpp

+        AD::SetPreaccIn(Gradient_j, nPrimVarGrad, nDim);
+        AD::SetPreaccIn(Coord_i, nDim); AD::SetPreaccIn(Coord_j, nDim);
+
+        /*--- Simple and Parameter-Free Second Slope Limiter for Unstructured Grid Aerodynamic Simulations---*/


Having a parameter-free option is interesting.. folks still have trouble tuning limiters and avoiding limit cycles/stalled convergence. Have you had a positive experience with this limiter in terms of convergence?

There is a trick with this parameter-free second limiter. If you have a high quality 3D grid, this limiter will greatly improve convergence. However for complex geometries (most of the time in aerospace industry), this second limiter will diverge the solution because it will turn off the primary limiter in regions of poor quality cells when M<1.0.
There is a measure of the cell quality (Angle, Skewness, etc) in SU2? If we have, I would like to use here in order to avoid turning off the limiter in this regions.

I see some problems in the current way it is implemented. First of all doing a search involving neighbors shouldn't be done inside an edge loop as you compute the same values twice (just an efficiency thing). Furthermore, in order for the computation to give the same result using different number of processors (which we should aim for) a communication of the variable Mach_Local is necessary.

Unless you think that this limiter greatly improves convergence, I would suggest to wait with this limiter implementation until we have figured out a way of generalizing this.

…o feature_DDESv5.0

talbring · 2018-01-08T08:59:46Z

Hm why are the regression tests failing ? It looks like is has something to do with the changes in the limiter in a recent commit.

talbring · 2018-01-08T13:29:34Z

Ok I think I found the problem, you accidentally removed the Set_MPI_Limiter call. Let's see what travis says.

…iter.

economon

LGTM. Thank you for all of the effort. Time to merge this one...

Feature ddes

Eduardo Molina and others added 30 commits April 11, 2016 21:20

Merge branch 'su2code/master'

a776afd

Merge branch 'su2code/master'

37c4195

Merge branch 'su2code/master'

dcbfd67

DDES

c4b2d99

First commit of Hybrid RANS/LES

Implementations part 1

278611b

- Roe Low Dissipation: (1 - f_d) + Ducros Sensor - Turbulence models: Edward’s SA, Compressibility Correction SA, Edward’s + CC. - In SA_DDES, the low Reynolds term was disabled. - Start of Venkatakrishnan Limiter Modification.

Implementation part 1 - Update

5b44d42

Added Marcello Righi's Delta_max in SA_DDES

f8429e0

Merge branch 'su2code/master' into feature_DDES

1e00fc8

# Conflicts: # Common/include/config_structure.hpp # SU2_CFD/include/numerics_structure.hpp # SU2_CFD/src/SU2_CFD.cpp # SU2_CFD/src/driver_structure.cpp # SU2_CFD/src/numerics_direct_mean.cpp

Update to v4.3

6a543d8

# Resolved conflicts in: # Common/include/config_structure.hpp # SU2_CFD/include/numerics_structure.hpp # SU2_CFD/src/SU2_CFD.cpp # SU2_CFD/src/driver_structure.cpp # SU2_CFD/src/numerics_direct_mean.cpp

Update RoeLowDissipation

8bda76e

- Now ROW_LOW_DISSIPATION flag has the following options: - FD: Numerical blending using the DDES’s Fd function. - NTS: Numerical blending of Travin and Shur. - NTS_DUCROS: NTS + Ducros’ Shock Sensor. All these options must be tested extensively.

Minor changes

69ec03a

Update config_template

ba76cf2

Update Low Dissipation FD_DUCROS

0a31a27

Minor Changes

2b16e1e

- Update Ducros Sensor. - Added a modified version of Venkatakrishnan Limiter.

Initial Preconditioner revision

bf9d403

Initial Implementation of SA-QCR2000

7860619

Venkatakrishnan_Mod minor changes

57507f1

Preventing the limiter to be greater than one.

Initial SA-IDDES implementation

8f6041e

It is incomplete right now.

More IDDES.

1cda79d

Add inputs for IDDES

4edadb0

Minor changes and update config_template.cfg

88a97d3

Merge branch 'feature_QCR' into feature_DDES-Master

7a219c4

Merge branch 'feature_DDES-Master' into feature_LowPrec

c634270

Fix Low Mach Correction.

500eb39

- Corrected variable declaration in Low_Mach_Preconditioner.

3392577

- Add Weiss & Smith Preconditioner + Thornber Low Mach Correction

Added Low_Mach_Prec flag.

59bae52

Added IZDES - Improved DDES with vorticity based SGS.

8f302c5

Merge branch 'master' into feature_DDES-v5.0

0c129c5

- Need to fix some conflicts.

Delete incompressible variables in solver_direct_mean.cpp

48464e7

Adjusted .travis.yml to prepare for DDES testing with this branch.

68a041f

talbring and others added 13 commits October 4, 2017 13:49

Fixed regression test vals for DDES case (due to bug fixes in ILU)

9f0cd6b

Reduced time-steps for regression test.

a614ae6

Removed allocation of CAdjTurbSolver when running disc.adj. solver

b11d15b

Fixed initialization of variables

696ae5f

Fixed reg. test values

1119581

Merge remote-tracking branch 'origin/develop' into feature_DDESv5.0

7499b55

Update DDES comments

b22f550

Removed flow average output for now.

9fcd51e

Moved Vortex tilting computation to SA solver

3f18d89

Merge branch 'feature_DDESv5.0' of https://github.com/su2code/SU2 int…

6c0ef03

…o feature_DDESv5.0

Merge remote-tracking branch 'su2code/develop' into feature_DDESv5.0

5f0f32c

Update ddes test case

abda05a

Merge branch 'develop' into feature_DDESv5.0

7d202d7

economon reviewed Jan 4, 2018

View reviewed changes

EduardoMolina added 4 commits January 4, 2018 15:54

Merge remote-tracking branch 'su2code/develop' into feature_DDESv5.0

5c24029

Update Venka_Wang + 2nd limiter

662e74b

Removed comments and unnecessary methods.

a3b4619

Merge branch 'feature_DDESv5.0' of https://github.com/su2code/SU2 int…

6270774

…o feature_DDESv5.0

Fixed missing Set_MPI_Limiter call

5ee61c9

talbring and others added 4 commits January 9, 2018 13:08

Renamed Sensor computation routines

7512542

Merge remote-tracking branch 'su2code/develop' into feature_DDESv5.0

ef58488

Reverted Xcode project, removed unnecessary files and Wenka + 2ND lim…

b20dfe4

…iter.

Reverting .travis.yml.

546525c

economon approved these changes Jan 12, 2018

View reviewed changes

economon merged commit 61e7b90 into develop Jan 12, 2018

economon deleted the feature_DDESv5.0 branch January 12, 2018 08:03

clarkpede mentioned this pull request May 9, 2018

Correct NTS central/upwind blending function #532

Merged

4 tasks

CatarinaGarbacz pushed a commit that referenced this pull request Aug 26, 2019

Merge pull request #490 from su2code/feature_DDESv5.0

9f1300b

Feature ddes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Feature ddes #490

Feature ddes #490

EduardoMolina commented Dec 26, 2017

economon left a comment

economon Jan 4, 2018

economon Jan 4, 2018

economon Jan 4, 2018

economon Jan 4, 2018

economon Jan 4, 2018

economon Jan 4, 2018

economon Jan 4, 2018

economon Jan 4, 2018

economon Jan 4, 2018

economon Jan 4, 2018

EduardoMolina Jan 4, 2018

talbring Jan 8, 2018

talbring commented Jan 8, 2018

talbring commented Jan 8, 2018

economon left a comment


		}

		void CUpwRoe_Flow::ComputeResidual(su2double val_residual, su2double val_Jacobian_i, su2double val_Jacobian_j, CConfig config) {


		/--- Evaluate Omega ---/

		//Omega = sqrt(Vorticity_i[0]Vorticity_i[0] + Vorticity_i[1]Vorticity_i[1] + Vorticity_i[2]*Vorticity_i[2]);

Feature ddes #490

Feature ddes #490

Conversation

EduardoMolina commented Dec 26, 2017

economon 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

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

talbring commented Jan 8, 2018

talbring commented Jan 8, 2018

economon left a comment

Choose a reason for hiding this comment