[Core] update analysis stage #2135

philbucher · 2018-05-20T13:26:26Z

This PR significantly improves the capabilities of the baseclass of the AnalysisStage. Common functionalities from the already existing AnalysisStages have been moved to the baseclass, to avoid code duplication and make the usage easier.

Solver construction is handled properly now
Now the processes are constructed and handled completely through the baseclass. In case the list of initialization matters (like it does in fluid) the order can be passed as a list
All processes involved in the analysis are now grouped in projectparams.json =>

...
"processes" : {
    initial_processes : [
        { proces_specific_params },
        { proces_specific_params }
    ],
    boundary_processes : [
        { proces_specific_params },
        { proces_specific_params }
    ]
}
...

Along with this come two more goodies:
We can finally do #869 !
Also it will be possible to automatize the computation of the ComputingModelPart (quite error prone atm)

NO existing code breaks!!! I implemented methods in the AnalysisStage of Fluid and Structure that make sure the old format still works
I adapted one test in StructuralMechanics as a proof of concept, the remaining ones I will update in another PR
@jcotela the test in Fluid also works. We should talk next week abt these modifications
The derived AnalysisStagess become much cleaner!

I know this a big step, so please take your time to review.

Requires #2068 to work

@ipouplana this is the PR I was refering to. Please also have a look, I think the PoroAnalysisStage will become much simpler after this

@loumalouomega I also had a look at the one in contact, it should not be much effort adapting it

RiccardoRossi

hi @philbucher this looks pretty fine to me.

what i am not clear is the distinction between "new" and "old" processes. could u explain didactically what is happening?

RiccardoRossi · 2018-05-20T14:02:03Z

applications/FluidDynamicsApplication/python_scripts/fluid_dynamics_analysis.py

+            if len(list_of_processes) == 0:
+                KratosMultiphysics.Logger.PrintInfo("FluidDynamicsAnalysis", "Using the old way to create the processes, this will be removed!")
+                from process_factory import KratosProcessFactory
+                factory = KratosProcessFactory(self.model)


could u explain in short what is happening here?

the KratosProcessFactory is still the same right? or did u change it?

I did not change the ProccesFactory
If I enter this if it means that there are no processes inside a processes dict present in the ProjectParams. This is how the current interface looks like

Then I construct the processes the same way they were constructed previously

philbucher · 2018-05-21T18:10:14Z

@RiccardoRossi I changed the naming, old processes was not really a good name

I meant old interface, i.e. if a block of processes is present outside of the processes dict
=> current interface, e.g. "initial_conditions" or "boundary_conditions" are present in the lowest level of the ProjectParams

In the future they would be grouped under "processes", see the example in the first post

RiccardoRossi · 2018-05-21T18:28:27Z

ok, all of this makes sense to me.

can i ask to document this in the wiki prior to merging the PR?

philbucher · 2018-05-21T20:10:08Z

Sure, will let you know when I am done with it

philbucher · 2018-05-24T10:05:16Z

ping @KratosMultiphysics/technical-committee I added the wiki for the AnalysisStage and the PythonSolver

Please have a look

maceligueta · 2018-05-24T10:29:21Z

Nice workd @philbucher !
In my experience, printing results from a coupled problem can be a complicated task as well. Merging the results in a single file, for example for GiD, requires knowing which variables are going to be printed for just one application or for many. That's also a kind of coupling that I don't really know were would go...

philbucher · 2018-05-24T11:39:56Z

@maceligueta maybe I misunderstand what you mean, but I am not intending to have all the results in one file.
How the output is printed is decided with which output_processes you add

maceligueta · 2018-05-24T12:53:05Z

Of course printing the results of each application in separate files is an option. But I prefer to print together all the meshes from the different coupled applications. It's very convenient, as you can embed your code in GiD (for example) and switch to post-process and take a look at the results, while the computation is still running. No need to do merges or open several post-processors.
So, my previous comment was in this direction: if someone wants to follow this approach, they have to disable the regular output of the original applications (delete the processes somehow, or override some methods with 'pass') and write a new process or method for printing the coupled output. This coupled output process might depend on the specific applications that you are coupling, though I guess it could be coded in a very general way.

philbucher · 2018-05-24T13:34:07Z

@maceligueta I think I understand now:
You have a utility or process with which you can print the results of a coupled simulation in one file to do what you describe, right?

If you want to use this but not follow the usual interface of the processes, then in your derived AnalysisStage you would override e.g. the OutputSolutionStep method to call your special utility

Regarding disabling the regular output of the original applications:
This is very simple, you only have to remove the corresponding process from the json-file, no need to do any modifications in python

Does this answer your question?

maceligueta · 2018-05-24T15:38:27Z

Thanks for the clarification @philbucher . Actually it was just a comment about the wiki: there it says that physics and couplings are bounded to the solver, however, couplings not related to physics can fall outside the solver. They will probably go to the processes or methods of the Analysis. Sorry if I was not very clear. So, even if the explanation at the wiki is simple, the actual coupled codes usually mess things up.

philbucher · 2018-05-28T15:27:23Z

@loumalouomega now I am done with the changes.

Note that the Fluid does not work with this yet, bcs the Fluid solvers do not yet derive from the base-solver and also don't take the model as input.

I spoke with @jcotela and he is going the changes very soon.

Contact is working

loumalouomega · 2018-05-28T15:40:39Z

OK, I was working on the fluid this morning, but I have reverted to avoid conflicts with your changes El lun., 28 may. 2018 17:27, Philipp Bucher <[email protected]> escribió:

…

@loumalouomega <https://github.com/loumalouomega> now I am done with the changes. Note that the Fluid does not work with this yet, bcs the Fluid solvers do not yet derive from the base-solver and also don't take the model as input. I spoke with @jcotela <https://github.com/jcotela> and he is going the changes very soon. Contact is working — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#2135 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ATEMgFugmQPml-3BEm0UbxMqjWy9PpRFks5t3BddgaJpZM4UGExr> .

…s-stage

philbucher · 2018-06-07T12:16:02Z

restart saving is handled through a process, added in #2283

philbucher · 2018-06-07T12:18:01Z

@jcotela can you please have a close look at the updates to the FluidAnalysis in case I missed sth?

philbucher · 2018-06-10T09:13:57Z

ping can we merge this now?

Its blocking further developments

RiccardoRossi · 2018-06-10T11:14:54Z

this looks fine to me, but having seen the troubles in the last weeks, i would like @rubenzorrilla to give his approval

philbucher · 2018-06-10T11:29:24Z

@RiccardoRossi makes sense to me
@rubenzorrilla the changes to the sovlers are minimal in this PR, nevertheless I will look again into the FSIapp to double check

philbucher · 2018-06-10T13:34:12Z

i took a look at the FSI but I cannot get the example from GiD to work.
Neither with Master nor with this branch.

@rubenzorrilla is it up to date?

The tests in FSI work except the one that always fails

philbucher · 2018-06-13T05:06:01Z

ping @rubenzorrilla

rubenzorrilla · 2018-06-13T06:57:25Z

kratos/python_scripts/analysis_stage.py

+
+    def _CreateProcesses(self, parameter_name, initialization_order):
+        """Create a list of processes
+        Format:


This seems too much complex to me. Is it required by any technical reason?

The idea behind grouping the processes is that one can loop them.
this way we can solve #869 and also automatize the computation of the ComputingModelPart (which is quite error-prone)

Without the extra level it would not be possible to specify the order of execution (like it is required e.g. in Fluids)

Users can still add their custom processes, but it should be done under processes or output_processes respectively

Go ahead then!

philbucher · 2018-06-13T09:18:44Z

Thanks

philbucher and others added 13 commits May 18, 2018 19:40

Implementing _GetSolver

453b511

Implemented _GetListOfProcesses

5c24c33

Implemented the OutputProcess logic

bc2841e

Moving prints to baseclass

fb81f7a

Unifying Output

ca29c72

First finished draft

44b3e56

Almost working

49a9d88

cleaning

ec5a88f

adding time-stepping in structural solver

3fe8949

More advances, now most tests are working

b8abae7

Final updates, now tests are working

7c8fae0

Adapting a test to the new structure

e2d5677

minor

41682c2

philbucher added Enhancement Applications labels May 20, 2018

philbucher self-assigned this May 20, 2018

philbucher requested a review from a team May 20, 2018 13:26

RiccardoRossi reviewed May 20, 2018

View reviewed changes

minor improvements

2184c38

philbucher mentioned this pull request May 24, 2018

[Core] Base python solver #2068

Merged

Merge branch 'master' into core/update-analysis-stage

44b999f

final cleanup

278f84a

Merge remote-tracking branch 'origin/master' into core/update-analysi…

f77d222

…s-stage

loumalouomega added a commit that referenced this pull request May 29, 2018

Merge #2135

3fef554

This was referenced May 31, 2018

[ConvectionDiffusionApplication] Modernize application. Adding solver and analysis stage #2224

Merged

[ConvectionDiffusionApplication] Modernize application. Adding solver and analysis stage (up-to-date to the master) #2229

Merged

Merge branch 'master' into core/update-analysis-stage

71040d0

jcotela mentioned this pull request Jun 5, 2018

Fluid/feature python solver derivation #2262

Merged

philbucher mentioned this pull request Jun 7, 2018

[Core] restart save process #2283

Merged

philbucher added 3 commits June 7, 2018 13:39

merge master and solve merge conflicts

fdb9a6e

minor in StructuralMechanics

d8d28ed

Updates in Fluid

94969af

philbucher and others added 2 commits June 10, 2018 09:53

Merge branch 'master' into core/update-analysis-stage

9742a66

TEMP disable restart tests, will be fixed in a separate PR!

1938862

philbucher added 2 commits June 10, 2018 11:48

fixing fluid

5279c48

added check

a4e2ae1

rubenzorrilla reviewed Jun 13, 2018

View reviewed changes

rubenzorrilla approved these changes Jun 13, 2018

View reviewed changes

philbucher merged commit 94aae63 into master Jun 13, 2018

philbucher deleted the core/update-analysis-stage branch June 13, 2018 09:19

ipouplana mentioned this pull request Jun 21, 2018

Updating Poromechanics analysis and solver to new standard #2360

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[Core] update analysis stage #2135

[Core] update analysis stage #2135

philbucher commented May 20, 2018

RiccardoRossi left a comment

RiccardoRossi May 20, 2018

philbucher May 21, 2018

philbucher commented May 21, 2018

RiccardoRossi commented May 21, 2018

philbucher commented May 21, 2018

philbucher commented May 24, 2018

maceligueta commented May 24, 2018

philbucher commented May 24, 2018

maceligueta commented May 24, 2018

philbucher commented May 24, 2018

maceligueta commented May 24, 2018 •

edited

Loading

philbucher commented May 28, 2018

loumalouomega commented May 28, 2018 via email

philbucher commented Jun 7, 2018

philbucher commented Jun 7, 2018

philbucher commented Jun 10, 2018

RiccardoRossi commented Jun 10, 2018

philbucher commented Jun 10, 2018

philbucher commented Jun 10, 2018

philbucher commented Jun 13, 2018

rubenzorrilla Jun 13, 2018

philbucher Jun 13, 2018

rubenzorrilla Jun 13, 2018

philbucher commented Jun 13, 2018

[Core] update analysis stage #2135

[Core] update analysis stage #2135

Conversation

philbucher commented May 20, 2018

RiccardoRossi left a comment

Choose a reason for hiding this comment

RiccardoRossi May 20, 2018

Choose a reason for hiding this comment

philbucher May 21, 2018

Choose a reason for hiding this comment

philbucher commented May 21, 2018

RiccardoRossi commented May 21, 2018

philbucher commented May 21, 2018

philbucher commented May 24, 2018

maceligueta commented May 24, 2018

philbucher commented May 24, 2018

maceligueta commented May 24, 2018

philbucher commented May 24, 2018

maceligueta commented May 24, 2018 • edited Loading

philbucher commented May 28, 2018

loumalouomega commented May 28, 2018 via email

philbucher commented Jun 7, 2018

philbucher commented Jun 7, 2018

philbucher commented Jun 10, 2018

RiccardoRossi commented Jun 10, 2018

philbucher commented Jun 10, 2018

philbucher commented Jun 10, 2018

philbucher commented Jun 13, 2018

rubenzorrilla Jun 13, 2018

Choose a reason for hiding this comment

philbucher Jun 13, 2018

Choose a reason for hiding this comment

rubenzorrilla Jun 13, 2018

Choose a reason for hiding this comment

philbucher commented Jun 13, 2018

maceligueta commented May 24, 2018 •

edited

Loading