Feature hb #311

arubino · 2016-09-29T15:37:04Z

Hi All,

This is a pull request for the harmonic balance implementation (in the CSpectralDriver). These are the major changes:

Full Harmonic Balance implementation working with any set of input frequencies (as opposed to previous Time Spectral implementation)
Removed Time-Spectral and generalized with the HB implementation
Moved the output from CSpectralDriver to COutput
Moved GetnZone and GetnDim from definition_structure to CConfig

Cheers,
Antonio & Sravya

…rmance

hlkline

I had a couple of comments around the test cases but otherwise this looks like a great addition/update.

hlkline · 2016-09-29T18:14:30Z

TestCases/serial_regression.py

@@ -490,7 +490,7 @@ def main():
    centrifugal_stage.cfg_dir   = "turbomachinery/centrifugal_stage"
    centrifugal_stage.cfg_file  = "centrifugal_stage.cfg"
    centrifugal_stage.test_iter = 100
-    centrifugal_stage.test_vals = [-10.166364, 1.621172, 2.206476e+01, 5.271075e-01] #last 4 columns
+    centrifugal_stage.test_vals = [-10.166717, 1.620590, 30.003500, 0.487959] #last 4 columns


Is there an explanation for why the values changed?

Thanks. Yes, the quantities for the turbomachinery test case are different now. There is a completely new structure for the turbomachinery performance that will be integrated soon when we will merge the turbomachiney features. After discussing about this with Salvo (that prepared that regression test), we decided that it's OK to just change the residuals for the time being. I can, anyway, have a closer look.

As long as someone running the code will be able to get the same results they got previously it should be fine - if the simulation is run to convergence (ie, more than 100 iterations), do the force coefficients (or other relevant turbomachinery output) converge to the same values?

Hi Heather, thanks a lot for this check because there was indeed a small bug related to the rotating frame! This should be fixed now and I also checked the final solution.

hlkline · 2016-09-29T18:15:51Z

TestCases/harmonic_balance/HB.cfg

 %                               POISSON_EQUATION)
 PHYSICAL_PROBLEM= EULER
 %
 % Specify turbulent model (NONE, SA, SA_NEG, SST)
 KIND_TURB_MODEL= NONE
 %
-% Mathematical problem (DIRECT, CONTINUOUS_ADJOINT)
+% Mathematical problem (DIRECT, ADJOINT, LINEARIZED, ONE_SHOT_ADJOINT)


suggest changing to CONTINUOUS_ADJOINT, which is the newer termonology used;

Thanks for the comments Heather. We'll edit the config file and commit the change.

talbring · 2016-09-30T08:24:14Z

SU2_CFD/src/output_structure.cpp

-    if ((SU2_TYPE::Int(iExtIter) >= 100)  && (SU2_TYPE::Int(iExtIter) < 1000))  SPRINTF (buffer, "_00%d.csv",   SU2_TYPE::Int(iExtIter));
-    if ((SU2_TYPE::Int(iExtIter) >= 1000) && (SU2_TYPE::Int(iExtIter) < 10000)) SPRINTF (buffer, "_0%d.csv",    SU2_TYPE::Int(iExtIter));
-    if (SU2_TYPE::Int(iExtIter) >= 10000) SPRINTF (buffer, "_%d.csv", SU2_TYPE::Int(iExtIter));
+  if (config->GetUnsteady_Simulation() == HARMONIC_BALANCE) {


Looks like you removed by accident also the extension for the time accurate unsteady simulation.

Yes, indeed! We should find a way to have Travis to check the .csv files as well. Thanks for checking this Tim!

talbring · 2016-09-30T08:24:56Z

Common/include/config_structure.hpp

+#ifdef HAVE_CGNS
+  #include "cgnslib.h"
+#endif
+


~~This can be removed I think.~~ Sorry, it's correct this way!

Hey Tim, thanks that is needed for the GetnDim now moved in config_structure.cpp

talbring · 2016-09-30T16:01:37Z

TestCases/parallel_regression.py

+    spectral.cfg_file  = "HB.cfg"
+    spectral.test_iter = 25
+    spectral.test_vals = [-1.569573, 3.941896, 0.008780, 0.079775] #last 4 columns
+    spectral.su2_exec  = "SU2_CFD"


Should be "parallel_computation.py -f" instead of SU2_CFD, sorry :)

Yep! Just pushed the correction.

…ure_HB

talbring

Thanks for this nice contribution! I am happy now!

arubino · 2016-10-03T14:58:21Z

Thanks all for the reviews! I would wait for at least another reviewer approval before merging it.

economon

Overall, looking good. Just one important consideration about changing the CSpectraDriver name.

economon · 2016-10-03T23:38:46Z

SU2_CFD/include/driver_structure.hpp

@@ -413,63 +413,60 @@ class CMultiZoneDriver : public CDriver {

 /*!
 * \class CSpectralDriver
- * \brief Class for driving an iteration of a spectral method problem using multiple zones.
+ * \brief Class for driving an iteration of harmonic balance method problem using multiple zones.


Perhaps we should also take this chance to rename the child driver class to CHarmonicBalanceDriver or CHBDriver? This would be more consistent with the replacements of "time spectral" with "harmonic balance" throughout the rest of the code.

Thanks Tom. Yes, we actually discussed about this together with Sravya. I changed to CHBDriver now. I used CHBDriver instead of CHarmonicBalanceDriver thinking at the next implementation that would be something like CMultiZoneHB. Let me know what you think about it.

economon · 2016-10-03T23:46:15Z

TestCases/parallel_regression.py

+    ### Harmonic Balance               ###                                                                                  
+    ######################################                                                                                    
+
+    spectral           = TestCase('spectral')


Change the test name to "harmonic_balance" rather than "spectral" for consistency

Yes, I also changed the old SPECTRAL_METHOD option in the test case config files and in the extended config template with the last commit.

…nge test cases options

hlkline

LGTM

economon

Looks Good!

arubino · 2016-10-05T07:08:47Z

Thanks All! Deleting the branch now ...

sravya91 and others added 30 commits August 7, 2015 17:09

Changes to SU2 made

da38a7e

Working SU2_TS

619a8df

Merged into develop

6c53041

Added Harmonic Balance Method

eea9805

Fixed line endings

a580809

Updated to Version 4.1

a6e8461

Must provide Time period for all spectral method computation

cc00631

Added HB,TS Spectral Method type options to config file

463eed2

resolve conflicts with develop

925f9ff

Resolve conflicts

7f86b56

Non-dimensionalization of spectral methods

2d49194

Merge branch 'develop' into feature_HB

c96a1f6

Change double to su2double

6fda7b0

Added regression test for HB flow solution, SetSpectralMethod(iZone)

19f504f

Fix conflicts with new develop

c437f7f

Small change indentation of travis config

84297e9

Some changes for new HB names

279ec29

Change residuals for spectral method regression test

0655ecb

Change residuals for turbomachinery due to a different definition

9606ee0

Move TS_file to COutput

e065419

Clean output and rename in view of turbomachinery merge

bf301b5

Change the way the spectral operator is computed to improve the perfo…

22d13ea

…rmance

Some comments

a830ba4

Clean CSpectralDriver::Run

1c76f7a

Update residuals in regression test for spectral testcase

2c44f54

Remove wrong residula list

ebe24dc

Remove SetTimeSpectral_Velocities()

8f96a1d

Implement the spectral operator using complex numbers

ae77f07

Clean old structure

ea86ed5

Remove print

2ac7ddc

arubino added 10 commits September 27, 2016 10:59

Change output format for spectral methods and multizone

2622722

Add GetnZone in config and modify the SU2 accordignly

11c522f

Move GetnZone and and GetnDim into CConfig

c7cea03

Merge branch 'develop' of https://github.com/su2code/SU2 into feature_HB

e75b55d

Rename spectral method for harmonic balance only and change .cfg file

9543e59

Rename for HB

d3042d2

Change regression test names and folder

7ce3563

Add new config file for HB

1cea431

Change travis config for pull request

f33d988

Delete old files

4a68b65

hlkline reviewed Sep 29, 2016

View reviewed changes

sravya91 added 2 commits September 29, 2016 11:52

Terminology edits to HB config file

392d1f0

Fixed minor bug in the MultiZone Dynamic mesh update

d4bffb9

talbring reviewed Sep 30, 2016

View reviewed changes

arubino and others added 3 commits September 30, 2016 15:43

Fix unsteady .csv output issue

3d19b61

Revert to old centrifugal_stage residuals

e7d9229

Merge branch 'develop' into feature_HB

a5fe9db

talbring reviewed Sep 30, 2016

View reviewed changes

arubino added 2 commits September 30, 2016 18:07

Change parallel regrassion test .su2_exec to parallel_computation.py

e56b39f

Merge branch 'feature_HB' of https://github.com/su2code/SU2 into feat…

38f52cc

…ure_HB

talbring approved these changes Sep 30, 2016

View reviewed changes

economon reviewed Oct 3, 2016

View reviewed changes

Change CSpectralDriver into CHBDriver, rename for consistency and cha…

91c7dd0

…nge test cases options

hlkline approved these changes Oct 4, 2016

View reviewed changes

economon approved these changes Oct 5, 2016

View reviewed changes

economon merged commit 6eb7aeb into develop Oct 5, 2016

arubino deleted the feature_HB branch October 5, 2016 07:09

arubino mentioned this pull request Jan 17, 2017

Some old options in config for HB #363

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Feature hb #311

Feature hb #311

arubino commented Sep 29, 2016

hlkline left a comment

hlkline Sep 29, 2016

arubino Sep 29, 2016

hlkline Sep 29, 2016

arubino Sep 30, 2016

hlkline Sep 29, 2016

sravya91 Sep 29, 2016

talbring Sep 30, 2016

arubino Sep 30, 2016

talbring Sep 30, 2016 •

edited

Loading

arubino Sep 30, 2016

talbring Sep 30, 2016

arubino Sep 30, 2016

talbring left a comment

arubino commented Oct 3, 2016

economon left a comment

economon Oct 3, 2016

arubino Oct 4, 2016

economon Oct 3, 2016

arubino Oct 4, 2016 •

edited

Loading

hlkline left a comment

economon left a comment

arubino commented Oct 5, 2016

Feature hb #311

Feature hb #311

Conversation

arubino commented Sep 29, 2016

hlkline 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

talbring Sep 30, 2016 • edited Loading

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 left a comment

Choose a reason for hiding this comment

arubino commented Oct 3, 2016

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

arubino Oct 4, 2016 • edited Loading

Choose a reason for hiding this comment

hlkline left a comment

Choose a reason for hiding this comment

economon left a comment

Choose a reason for hiding this comment

arubino commented Oct 5, 2016

talbring Sep 30, 2016 •

edited

Loading

arubino Oct 4, 2016 •

edited

Loading