Add advection to the adaptive sparse contacts (ASC) #1113

acdaigneault · 2024-05-03T19:51:16Z

Description of the problem

Periodic nodes were not properly handled with the last implementation: the periodic nodes did not have information on the mobility status on the coincident nodes. This was spotted when there has particles at one side of the periodic cells and the other periodic ones were empty. Those cells where not mobile, but this is one of the criteria for mobile.
Disabling contact of particles was functional in CFD-DEM: cells with low granular temperature deactivated the contacts. However, it is possible for particles within a cell to have a low variance in velocity and a significant average velocity, such as when particles experience bulk motion.

Description of the solution

Periodic nodes:
- Mapping periodic nodes was achieved with the constraints since we can identify the periodic nodes from it (so called identity constraints).
- Some attempted as been done to apply the periodic constraints to the mobility_at_nodes, without any success. The main issue is that constraints force values at PB1 = values at PB2. In our case, we need to check of the prevailing value, so sometime it is PB2 = PB1, with was not compatible with straight constraints operations. "[...] it adds the DoFs of face_1 to the list of constrained DoFs in constraints and adds entries to constrain them to the corresponding values of the DoFs on face_2. " see Deal.II doc here
Disabling contacts with advection:
- New mobility status has been added to the disabling contacts: advected and advected_active
- The new algorithm doesn't detect when particles are advected, this is a user-defined parameter. It could be useful for DEM when particles are moving in bulk by gravity, but this is more practical for CFD-DEM. The first idea was to implement advected particles auto-detection, but this results in more checks, also the average velocities threshold is not easly configurable.
- The way to "apply advection" is that the average velocity and acceleration of the cells are calculated or updated everytime the mobile cells are integrated. The average values are calculated in the same loop than the integration and it allows to always have the last calculated velocity values. For advected cells, all particles of a cell are integrated with an cell-averaged velocity and acceleration * time step. This velocity is also updated with the a*dt values.

How Has This Been Tested?

DEM tests for ASC was been updated
New test in CFD-DEM called adaptative_sparse_contacts.prm

Documentation

Change the "disabling contacts" parameters to "adaptive sparse contacts"
Doc is updated with new parameters and new names

Comments

This algorithm was called Disabling Contacts when it was first introduce in the code for DEM
An example will be added in another PR

blaisb

Few files already done, wil continue tomorrow

applications_tests/lethe-fluid-particles/adaptive_sparse_contacts_files/dem.pvdhandler

blaisb · 2024-05-07T15:49:05Z

doc/source/parameters/dem/model_parameters.rst


-The dynamic disabling controls the disabling contact mechanism for performance enhancement. This feature dynamically searches for cells with low particle motion (granular temperature), disabling the computation of contacts for particles within these cells.
+The ASC controls the disabling contact mechanism for performance enhancement. This feature adaptively searches for cells with low particle motion (granular temperature), disabling the computation of contacts for particles within these cells.


we could put your article on Arxiv and. you could reference it here

I will keep that in mind, and will do a micro-PR to add the reference :)

include/dem/adaptive_sparse_contacts.h

blaisb

Bunch of more files.
Thus far this is going well. this is a well done PR, very good quality work. Great job

include/dem/adaptive_sparse_contacts.h

include/dem/dem.h

include/dem/particle_point_line_broad_search.h

include/dem/velocity_verlet_integrator.h

include/fem-dem/cfd_dem_coupling.h

blaisb

Last batch of comments. Good job @acdaigneault

source/dem/CMakeLists.txt

source/dem/adaptive_sparse_contacts.cc

source/dem/velocity_verlet_integrator.cc

source/fem-dem/cfd_dem_coupling.cc

voferreira

Nice addition and very nice and clear documentation on comments. I will checkpoint here and try to finish it Today. Otherwise, I will wrap it up tomorrow.

include/dem/adaptive_sparse_contacts.h

include/dem/particle_particle_broad_search.h

include/dem/adaptive_sparse_contacts.h

source/dem/particle_particle_broad_search.cc

blaisb

Other than Victor's comment, that I agree with, in my opinion this will be ready for merge after all the reviews

blaisb · 2024-05-09T11:08:50Z

applications_tests/lethe-fluid-particles/adaptive_sparse_contacts.prm

+
+subsection simulation control
+  set method               = bdf1
+  set number mesh adapt    = 0


voferreira

No more comments as well. Good job, really! I will approve it after you have addressed the previous comments.

Clean up main algo (remove irrelevant checks)

Few test fixes

voferreira

Amazing figure, amazing documentation, amazing feature. Good job!
Very last comment, but the rest is all fine.

P.S.: My bad for the example of lines in the comment, but the cascade thingy is good.

voferreira · 2024-05-10T00:34:04Z

doc/source/parameters/dem/images/mobility_status.png

You set the bar too high, you know?! Amazing, really!

One comment, though: for the dark mode, if you do not set a bg, we won't be able to read the content of the label. Could you add a white bg to your figure?

Ohhh good point, I will do it tonight or tomorrow :)

Damn, thats too much flatteries for me hahaha ❤️

Description of the problem Periodic nodes were not properly handled with the last implementation: the periodic nodes did not have information on the mobility status on the coincident nodes. This was spotted when there has particles at one side of the periodic cells and the other periodic ones were empty. Those cells where not mobile, but this is one of the criteria for mobile. Disabling contact of particles was functional in CFD-DEM: cells with low granular temperature deactivated the contacts. However, it is possible for particles within a cell to have a low variance in velocity and a significant average velocity, such as when particles experience bulk motion. Description of the solution Periodic nodes: Mapping periodic nodes was achieved with the constraints since we can identify the periodic nodes from it (so called identity constraints). Some attempted as been done to apply the periodic constraints to the mobility_at_nodes, without any success. The main issue is that constraints force values at PB1 = values at PB2. In our case, we need to check of the prevailing value, so sometime it is PB2 = PB1, with was not compatible with straight constraints operations. "[...] it adds the DoFs of face_1 to the list of constrained DoFs in constraints and adds entries to constrain them to the corresponding values of the DoFs on face_2. " see Deal.II doc here Disabling contacts with advection: New mobility status has been added to the disabling contacts: advected and advected_active The new algorithm doesn't detect when particles are advected, this is a user-defined parameter. It could be useful for DEM when particles are moving in bulk by gravity, but this is more practical for CFD-DEM. The first idea was to implement advected particles auto-detection, but this results in more checks, also the average velocities threshold is not easly configurable. The way to "apply advection" is that the average velocity and acceleration of the cells are calculated or updated everytime the mobile cells are integrated. The average values are calculated in the same loop than the integration and it allows to always have the last calculated velocity values. For advected cells, all particles of a cell are integrated with an cell-averaged velocity and acceleration * time step. This velocity is also updated with the a*dt values. How Has This Been Tested? DEM tests for ASC was been updated New test in CFD-DEM called adaptative_sparse_contacts.prm Documentation Change the "disabling contacts" parameters to "adaptive sparse contacts" Doc is updated with new parameters and new names Comments This algorithm was called Disabling Contacts when it was first introduce in the code for DEM An example will be added in another PR Former-commit-id: 81fb7fa

acdaigneault added the WIP When a PR is open but not ready for review label May 3, 2024

acdaigneault changed the title ~~Add advection to the adaptive sparse contacts (ASC) (previously called disabling contacts) [wip]~~ Add advection to the adaptive sparse contacts (ASC) [wip] May 3, 2024

blaisb requested review from OGaboriault, voferreira and blaisb May 4, 2024 14:01

acdaigneault force-pushed the cfd-dem_sparse_contacts branch from d7ce118 to 42157e9 Compare May 6, 2024 16:45

acdaigneault changed the title ~~Add advection to the adaptive sparse contacts (ASC) [wip]~~ Add advection to the adaptive sparse contacts (ASC) May 6, 2024

acdaigneault added Ready for review Rebased and removed WIP When a PR is open but not ready for review labels May 6, 2024

blaisb requested changes May 7, 2024

View reviewed changes

blaisb requested changes May 8, 2024

View reviewed changes

voferreira requested changes May 8, 2024

View reviewed changes

blaisb approved these changes May 9, 2024

View reviewed changes

voferreira requested changes May 9, 2024

View reviewed changes

acdaigneault added 8 commits May 9, 2024 19:59

Add advective particles to disabling contacts

5b47246

Clean up main algo (remove irrelevant checks)

Review comments

b69651b

Few test fixes

Change "disable contacts" to "sparse contacts"

b9bc6d8

Fix contact search with PBC and test

bca441e

Change parameter names and update documentation

8c34152

Remove unused parameter in bunny drill example

e386de5

Apply changes request of Bruno's review

4249c80

Applied Victor's comments from review

b3538f2

acdaigneault force-pushed the cfd-dem_sparse_contacts branch from 85f1610 to b3538f2 Compare May 10, 2024 00:01

acdaigneault added Enhancement New feature or request Reviewed and ready to merge and removed Ready for review labels May 10, 2024

Change log

84af220

acdaigneault requested a review from voferreira May 10, 2024 00:06

voferreira approved these changes May 10, 2024

View reviewed changes

blaisb merged commit 81fb7fa into master May 10, 2024
8 checks passed

acdaigneault mentioned this pull request Jul 24, 2024

Remove forcing particle displacement in PBC with CFD-DEM at last DEM iteration #1204

Merged

11 tasks

blaisb deleted the cfd-dem_sparse_contacts branch August 5, 2024 13:12

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add advection to the adaptive sparse contacts (ASC) #1113

Add advection to the adaptive sparse contacts (ASC) #1113

acdaigneault commented May 3, 2024 •

edited

Loading

blaisb left a comment

blaisb May 7, 2024

acdaigneault May 9, 2024

blaisb left a comment

blaisb left a comment

voferreira left a comment

blaisb left a comment

blaisb May 9, 2024

voferreira left a comment

voferreira left a comment

voferreira May 10, 2024

voferreira May 10, 2024

acdaigneault May 10, 2024

acdaigneault May 10, 2024


		The dynamic disabling controls the disabling contact mechanism for performance enhancement. This feature dynamically searches for cells with low particle motion (granular temperature), disabling the computation of contacts for particles within these cells.
		The ASC controls the disabling contact mechanism for performance enhancement. This feature adaptively searches for cells with low particle motion (granular temperature), disabling the computation of contacts for particles within these cells.

Add advection to the adaptive sparse contacts (ASC) #1113

Add advection to the adaptive sparse contacts (ASC) #1113

Conversation

acdaigneault commented May 3, 2024 • edited Loading

Description of the problem

Description of the solution

How Has This Been Tested?

Documentation

Comments

blaisb left a comment

Choose a reason for hiding this comment

blaisb May 7, 2024

Choose a reason for hiding this comment

acdaigneault May 9, 2024

Choose a reason for hiding this comment

blaisb left a comment

Choose a reason for hiding this comment

blaisb left a comment

Choose a reason for hiding this comment

voferreira left a comment

Choose a reason for hiding this comment

blaisb left a comment

Choose a reason for hiding this comment

blaisb May 9, 2024

Choose a reason for hiding this comment

voferreira left a comment

Choose a reason for hiding this comment

voferreira left a comment

Choose a reason for hiding this comment

voferreira May 10, 2024

Choose a reason for hiding this comment

voferreira May 10, 2024

Choose a reason for hiding this comment

acdaigneault May 10, 2024

Choose a reason for hiding this comment

acdaigneault May 10, 2024

Choose a reason for hiding this comment

acdaigneault commented May 3, 2024 •

edited

Loading