Temperature-dependant solid domain constraints #1038

AmishgaAlphonius · 2024-02-22T15:45:30Z

Description of the problem

In phase change simulations, the solid domain was treated as a simple fluid with a higher viscosity (viscous penalization). It worked quite well, but it wasn't the most efficient way to compute fluid dynamics quantities.

Description of the solution

In order to improve calculation time and linear system conditioning, homogeneous constraints are imposed on velocity DOFs in a defined solid domain through a temperature range. Additionally, pressure DOFs in the same solid domain that are not connected to fluid cells are also imposed to homogeneous constraints. The temperature range is specified through a new parameter subsection, namely constrain solid domain.

How Has This Been Tested?

The new feature has been tested on the Melting Cavity example. With solid domain constraints on cells within the temperature range of [0,29] degC, the linear solver solving time of the fluid dynamics physic was reduced by a little more than the half without altering the postprocessed results. This feature will improve greatly computation times of domains with a higher proportion of "solid" phase.
The feature was also tested with the restart feature and no issue occurred. Final results were the same as without the restart.
An application test has been add to test this feature : applications_tests/lethe-fluid/heat_transfer_phase_change_constrain_solid_domain.prm (.output)

Documentation

A new page for the new parameter subsection (constrain solid domain) has been added : doc/source/parameters/cfd/constrain_solid_domain.rst

Future changes

The current implementation of this new feature only works with 1 fluid simulations (an exception is thrown when used with VOF or Cahn Hilliard). However, the implementation with VOF simulations will be added in a near future.

Comments

The use of a dynamic_zero_constraints AffineConstraints object, rather than reconstruction of the zero_constraints one at every time step was determined to be optimal after some profiling tests.
A small mistake found in establish_solid_domain() was corrected.
Parameters are stored in vectors for the upcoming implementation of the feature in multiple fluid simulations.
The Variable enum class was moved from parameters.h to multiphysics.h. Using variables from the Variable enum class, the constrained and constraining fields could be parametrized in the future according to user needs.
Other minor corrections were also made to the documentation.
Missing module imports for the Melting Cavity example postprocessing were added.
Please disregard name of the branch, it depicts a slight change of plans.

hepap

First review, I will go through it again on Monday or Thuesday.

Very nice addition and nicely done! Great job :)

doc/source/parameters/cfd/constrain_solid_domain.rst

include/core/parameters.h

include/solvers/navier_stokes_base.h

include/solvers/simulation_parameters.h

include/solvers/gls_navier_stokes.h

doc/source/theory/fluid_dynamics/navier-stokes.rst

doc/source/parameters/cfd/constrain_solid_domain.rst

include/core/parameters.h

include/solvers/navier_stokes_base.h

source/core/parameters.cc

source/solvers/gls_navier_stokes.cc

source/solvers/navier_stokes_base.cc

include/core/parameters.h

blaisb

A few things there and there. Good job!

blaisb · 2024-02-27T13:51:23Z

examples/multiphysics/melting-cavity/melting-cavity.py

+import pandas as pd

+import os


Why do you need to add these modules?

They are called in the script but they are not imported before.

include/core/parameters.h

include/solvers/navier_stokes_base.h

blaisb · 2024-02-27T14:03:39Z

include/solvers/simulation_parameters.h

+    constrain_solid_domain.declare_parameters(
+      prm, 1); // at the moment, default value of the number of constraints is
+               // set to one since we are only applying to one-fluid simulations


The default value should always be one right? We can use a similar mechanism to the one we use in the boundary conditions to pre-detect the number of contraints. Otherwise, you could have a max constraints to pregenerate them.

Exactly, I was thinking of having a max value variable, but I didn't do it since it was only one. I'll rename the argument just to make it clear.

blaisb · 2024-02-27T14:31:06Z

source/solvers/navier_stokes_base.cc

+NavierStokesBase<dim, VectorType, DofsType>::flag_dofs_in_solid(
+  std::unordered_set<types::global_dof_index> &dofs_are_in_solid,
+  const std::vector<types::global_dof_index>  &local_dof_indices)


Since this function is called within a cell, I would consider inlining it and defining it in Utilities or somewhere else

blaisb · 2024-02-27T14:31:11Z

source/solvers/navier_stokes_base.cc

+template <int dim, typename VectorType, typename DofsType>
+bool
+NavierStokesBase<dim, VectorType, DofsType>::check_cell_is_in_solid(
+  const std::unordered_set<types::global_dof_index> &dofs_are_in_solid,
+  const std::vector<types::global_dof_index>        &local_dof_indices)


Since this function is called within a cell, I would consider inlining it and defining it in Utilities or somewhere else

blaisb · 2024-02-27T14:31:16Z

source/solvers/navier_stokes_base.cc

+template <int dim, typename VectorType, typename DofsType>
+void
+NavierStokesBase<dim, VectorType, DofsType>::flag_dofs_connected_to_fluid(
+  std::unordered_set<types::global_dof_index> &dofs_are_connected_to_fluid,
+  const std::vector<types::global_dof_index>  &local_dof_indices)


Since this function is called within a cell, I would consider inlining it and defining it in Utilities or somewhere else

blaisb · 2024-02-27T14:31:22Z

source/solvers/navier_stokes_base.cc

+
+template <int dim, typename VectorType, typename DofsType>
+bool
+NavierStokesBase<dim, VectorType, DofsType>::check_cell_is_connected_to_fluid(


Since this function is called within a cell, I would consider inlining it and defining it in Utilities or somewhere else

blaisb · 2024-02-27T14:31:30Z

source/solvers/navier_stokes_base.cc

+NavierStokesBase<dim, VectorType, DofsType>::constrain_pressure(
+  const bool                                 &non_zero_constraints,
+  const std::vector<types::global_dof_index> &local_dof_indices,
+  AffineConstraints<double>                  &zero_constraints)


Since this function is called within a cell, I would consider inlining it and defining it in Utilities or somewhere else

doc/source/parameters/cfd/constrain_solid_domain.rst

include/core/parameters.h

include/solvers/gls_navier_stokes.h

Co-authored-by: hepap <[email protected]>

Co-authored-by: Olivier Guévremont <[email protected]>

Co-authored-by: Olivier Guévremont <[email protected]> Co-authored-by: hepap <[email protected]>

Co-authored-by: Bruno Blais <[email protected]>

Co-authored-by: hepap <[email protected]>

blaisb · 2024-02-27T18:41:56Z

@AmishgaAlphonius can you rebase if the tests don't pass? Then I'll merge after CI

Description of the problem In phase change simulations, the solid domain was treated as a simple fluid with a higher viscosity (viscous penalization). It worked quite well, but it wasn't the most efficient way to compute fluid dynamics quantities. In order to improve calculation time and linear system conditioning, homogeneous constraints are imposed on velocity DOFs in a defined solid domain through a temperature range. Additionally, pressure DOFs in the same solid domain that are not connected to fluid cells are also imposed to homogeneous constraints. The temperature range is specified through a parameter subsection, namely constrain solid domain. Previously, the implementation was done for single-fluid simulations (Temperature-dependant solid domain constraints #1038), but simulations such as laser powder bed fusion (LPBF) meltpool ones require two fluids when modelling. Description of the solution This PR, implements the new temperature-dependent solid domain constraints for two-phase Volume of Fluid (VOF) simulations. To select cells on which the constraints are applied, we check if the filtered phase fraction at quadrature points in the cell are within a range of accepted values. This range of accepted values is defined with the phase fraction tolerance parameter. This parameter defines an absolute tolerance on filtered phase fraction. How Has This Been Tested? The new feature has been tested on the Melting Cavity example by modelling to parts of the domain with different fluids while keeping the same properties in both fluids. An application test has been added to test this feature : applications_tests/lethe-fluid/heat_transfer_vof_phase_change_constrain_solid_domain.prm (.output) Documentation Documentation page was updated: (doc/source/parameters/cfd/constrain_solid_domain.rst). Comments The structure of constrain_solid_domain() changed a bit. More functions were added to avoid code line repetitions in constrain_solid_domain_vof(). Small corrections were made in other files. Co-authored-by: hepap <[email protected]> Co-authored-by: Bruno Blais <[email protected]> Co-authored-by: Olivier Guévremont <[email protected]>

Description This PR adds a small section in the melting cavity example featuring the new constrain stasis feature from (Temperature-dependant solid domain constraints #1038). Documentation Parameter documentation was also updated : doc/source/parameters/cfd/constrain_stasis.rst

Description of the problem In phase change simulations, the solid domain was treated as a simple fluid with a higher viscosity (viscous penalization). It worked quite well, but it wasn't the most efficient way to compute fluid dynamics quantities. Description of the solution In order to improve calculation time and linear system conditioning, homogeneous constraints are imposed on velocity DOFs in a defined solid domain through a temperature range. Additionally, pressure DOFs in the same solid domain that are not connected to fluid cells are also imposed to homogeneous constraints. The temperature range is specified through a new parameter subsection, namely constrain solid domain. How Has This Been Tested? The new feature has been tested on the Melting Cavity example. With solid domain constraints on cells within the temperature range of [0,29] degC, the linear solver solving time of the fluid dynamics physic was reduced by a little more than the half without altering the postprocessed results. This feature will improve greatly computation times of domains with a higher proportion of "solid" phase. The feature was also tested with the restart feature and no issue occurred. Final results were the same as without the restart. An application test has been add to test this feature : applications_tests/lethe-fluid/heat_transfer_phase_change_constrain_solid_domain.prm (.output) Documentation A new page for the new parameter subsection (constrain solid domain) has been added : doc/source/parameters/cfd/constrain_solid_domain.rst Future changes The current implementation of this new feature only works with 1 fluid simulations (an exception is thrown when used with VOF or Cahn Hilliard). However, the implementation with VOF simulations will be added in a near future. Comments The use of a dynamic_zero_constraints AffineConstraints object, rather than reconstruction of the zero_constraints one at every time step was determined to be optimal after some profiling tests. A small mistake found in establish_solid_domain() was corrected. Parameters are stored in vectors for the upcoming implementation of the feature in multiple fluid simulations. The Variable enum class was moved from parameters.h to multiphysics.h. Using variables from the Variable enum class, the constrained and constraining fields could be parametrized in the future according to user needs. Other minor corrections were also made to the documentation. Missing module imports for the Melting Cavity example postprocessing were added. Please disregard name of the branch, it depicts a slight change of plans. Co-authored-by: hepap <[email protected]> Co-authored-by: Olivier Guévremont <[email protected]> Co-authored-by: Bruno Blais <[email protected]> Former-commit-id: 332dabe

…-polymtl#1048) Description of the problem In phase change simulations, the solid domain was treated as a simple fluid with a higher viscosity (viscous penalization). It worked quite well, but it wasn't the most efficient way to compute fluid dynamics quantities. In order to improve calculation time and linear system conditioning, homogeneous constraints are imposed on velocity DOFs in a defined solid domain through a temperature range. Additionally, pressure DOFs in the same solid domain that are not connected to fluid cells are also imposed to homogeneous constraints. The temperature range is specified through a parameter subsection, namely constrain solid domain. Previously, the implementation was done for single-fluid simulations (Temperature-dependant solid domain constraints chaos-polymtl#1038), but simulations such as laser powder bed fusion (LPBF) meltpool ones require two fluids when modelling. Description of the solution This PR, implements the new temperature-dependent solid domain constraints for two-phase Volume of Fluid (VOF) simulations. To select cells on which the constraints are applied, we check if the filtered phase fraction at quadrature points in the cell are within a range of accepted values. This range of accepted values is defined with the phase fraction tolerance parameter. This parameter defines an absolute tolerance on filtered phase fraction. How Has This Been Tested? The new feature has been tested on the Melting Cavity example by modelling to parts of the domain with different fluids while keeping the same properties in both fluids. An application test has been added to test this feature : applications_tests/lethe-fluid/heat_transfer_vof_phase_change_constrain_solid_domain.prm (.output) Documentation Documentation page was updated: (doc/source/parameters/cfd/constrain_solid_domain.rst). Comments The structure of constrain_solid_domain() changed a bit. More functions were added to avoid code line repetitions in constrain_solid_domain_vof(). Small corrections were made in other files. Co-authored-by: hepap <[email protected]> Co-authored-by: Bruno Blais <[email protected]> Co-authored-by: Olivier Guévremont <[email protected]> Former-commit-id: 581ad84

Description This PR adds a small section in the melting cavity example featuring the new constrain stasis feature from (Temperature-dependant solid domain constraints chaos-polymtl#1038). Documentation Parameter documentation was also updated : doc/source/parameters/cfd/constrain_stasis.rst Former-commit-id: e843eae

Description of the problem In phase change simulations, the solid domain was treated as a simple fluid with a higher viscosity (viscous penalization). It worked quite well, but it wasn't the most efficient way to compute fluid dynamics quantities. Description of the solution In order to improve calculation time and linear system conditioning, homogeneous constraints are imposed on velocity DOFs in a defined solid domain through a temperature range. Additionally, pressure DOFs in the same solid domain that are not connected to fluid cells are also imposed to homogeneous constraints. The temperature range is specified through a new parameter subsection, namely constrain solid domain. How Has This Been Tested? The new feature has been tested on the Melting Cavity example. With solid domain constraints on cells within the temperature range of [0,29] degC, the linear solver solving time of the fluid dynamics physic was reduced by a little more than the half without altering the postprocessed results. This feature will improve greatly computation times of domains with a higher proportion of "solid" phase. The feature was also tested with the restart feature and no issue occurred. Final results were the same as without the restart. An application test has been add to test this feature : applications_tests/lethe-fluid/heat_transfer_phase_change_constrain_solid_domain.prm (.output) Documentation A new page for the new parameter subsection (constrain solid domain) has been added : doc/source/parameters/cfd/constrain_solid_domain.rst Future changes The current implementation of this new feature only works with 1 fluid simulations (an exception is thrown when used with VOF or Cahn Hilliard). However, the implementation with VOF simulations will be added in a near future. Comments The use of a dynamic_zero_constraints AffineConstraints object, rather than reconstruction of the zero_constraints one at every time step was determined to be optimal after some profiling tests. A small mistake found in establish_solid_domain() was corrected. Parameters are stored in vectors for the upcoming implementation of the feature in multiple fluid simulations. The Variable enum class was moved from parameters.h to multiphysics.h. Using variables from the Variable enum class, the constrained and constraining fields could be parametrized in the future according to user needs. Other minor corrections were also made to the documentation. Missing module imports for the Melting Cavity example postprocessing were added. Please disregard name of the branch, it depicts a slight change of plans. Co-authored-by: hepap <[email protected]> Co-authored-by: Olivier Guévremont <[email protected]> Co-authored-by: Bruno Blais <[email protected]> Former-commit-id: 332dabe

Description of the problem In phase change simulations, the solid domain was treated as a simple fluid with a higher viscosity (viscous penalization). It worked quite well, but it wasn't the most efficient way to compute fluid dynamics quantities. In order to improve calculation time and linear system conditioning, homogeneous constraints are imposed on velocity DOFs in a defined solid domain through a temperature range. Additionally, pressure DOFs in the same solid domain that are not connected to fluid cells are also imposed to homogeneous constraints. The temperature range is specified through a parameter subsection, namely constrain solid domain. Previously, the implementation was done for single-fluid simulations (Temperature-dependant solid domain constraints #1038), but simulations such as laser powder bed fusion (LPBF) meltpool ones require two fluids when modelling. Description of the solution This PR, implements the new temperature-dependent solid domain constraints for two-phase Volume of Fluid (VOF) simulations. To select cells on which the constraints are applied, we check if the filtered phase fraction at quadrature points in the cell are within a range of accepted values. This range of accepted values is defined with the phase fraction tolerance parameter. This parameter defines an absolute tolerance on filtered phase fraction. How Has This Been Tested? The new feature has been tested on the Melting Cavity example by modelling to parts of the domain with different fluids while keeping the same properties in both fluids. An application test has been added to test this feature : applications_tests/lethe-fluid/heat_transfer_vof_phase_change_constrain_solid_domain.prm (.output) Documentation Documentation page was updated: (doc/source/parameters/cfd/constrain_solid_domain.rst). Comments The structure of constrain_solid_domain() changed a bit. More functions were added to avoid code line repetitions in constrain_solid_domain_vof(). Small corrections were made in other files. Co-authored-by: hepap <[email protected]> Co-authored-by: Bruno Blais <[email protected]> Co-authored-by: Olivier Guévremont <[email protected]> Former-commit-id: 581ad84

Description This PR adds a small section in the melting cavity example featuring the new constrain stasis feature from (Temperature-dependant solid domain constraints #1038). Documentation Parameter documentation was also updated : doc/source/parameters/cfd/constrain_stasis.rst Former-commit-id: e843eae

AmishgaAlphonius added Enhancement New feature or request Ready for review labels Feb 22, 2024

AmishgaAlphonius requested review from blaisb, oguevremont and hepap February 22, 2024 15:45

AmishgaAlphonius self-assigned this Feb 22, 2024

hepap reviewed Feb 23, 2024

View reviewed changes

oguevremont reviewed Feb 23, 2024

View reviewed changes

AmishgaAlphonius force-pushed the vof-solid-velocity-constraint branch from e2a81d9 to 851baf4 Compare February 23, 2024 20:43

AmishgaAlphonius requested review from hepap and oguevremont February 23, 2024 20:44

oguevremont approved these changes Feb 23, 2024

View reviewed changes

include/core/parameters.h Show resolved Hide resolved

blaisb requested changes Feb 27, 2024

View reviewed changes

hepap approved these changes Feb 27, 2024

View reviewed changes

doc/source/parameters/cfd/constrain_solid_domain.rst Outdated Show resolved Hide resolved

include/core/parameters.h Outdated Show resolved Hide resolved

include/solvers/gls_navier_stokes.h Outdated Show resolved Hide resolved

AmishgaAlphonius requested a review from blaisb February 27, 2024 16:22

AmishgaAlphonius and others added 15 commits February 27, 2024 13:32

Implement temperature-dependant zero constraint

76b34ed

Add documentation

df9dc13

Add exception for multiple fluid simulations

5bbd25b

Add dynamic_zero_constraints

6c48347

Small correction to NS theory guide

1a40d66

Add application test

b2a5976

Update comments

e4094c1

Remove temperature statistics in test

a359a0f

Correct typos

113712e

Indent

a11f10d

Remove warning

6b2217a

Remove unused variables

5292af5

Correct small mistake

c15ee27

Correct other small mistakes

2b1839d

Apply suggestions from code review

997ce29

Co-authored-by: hepap <[email protected]>

AmishgaAlphonius and others added 7 commits February 27, 2024 13:32

Apply suggestions from code review

c0c0a4d

Co-authored-by: hepap <[email protected]>

Apply suggestions from code review

457293c

Co-authored-by: Olivier Guévremont <[email protected]>

Apply suggestions from code review

4129dfe

Co-authored-by: Olivier Guévremont <[email protected]> Co-authored-by: hepap <[email protected]>

Update source/solvers/gls_navier_stokes.cc

1775c8c

Co-authored-by: Bruno Blais <[email protected]>

Apply suggestions from code review.

e41eaf4

Co-authored-by: Bruno Blais <[email protected]>

Update include/core/parameters.h

10362a6

Co-authored-by: hepap <[email protected]>

Apply suggestions from code review

dce627f

Co-authored-by: hepap <[email protected]>

AmishgaAlphonius force-pushed the vof-solid-velocity-constraint branch from cd9380f to dce627f Compare February 27, 2024 18:33

blaisb approved these changes Feb 27, 2024

View reviewed changes

blaisb added Reviewed and ready to merge and removed Ready for review labels Feb 27, 2024

blaisb merged commit 332dabe into master Feb 27, 2024
8 checks passed

blaisb deleted the vof-solid-velocity-constraint branch February 27, 2024 23:48

AmishgaAlphonius mentioned this pull request Feb 29, 2024

Implement temperature-dependant solid domain contraints in VOF #1048

Merged

AmishgaAlphonius mentioned this pull request Mar 11, 2024

Add stasis constraint in melting cavity example #1061

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Temperature-dependant solid domain constraints #1038

Temperature-dependant solid domain constraints #1038

AmishgaAlphonius commented Feb 22, 2024 •

edited

Loading

hepap left a comment

blaisb left a comment

blaisb Feb 27, 2024

AmishgaAlphonius Feb 27, 2024

blaisb Feb 27, 2024

AmishgaAlphonius Feb 27, 2024 •

edited

Loading

blaisb Feb 27, 2024

blaisb Feb 27, 2024

blaisb Feb 27, 2024

blaisb Feb 27, 2024

blaisb Feb 27, 2024

blaisb commented Feb 27, 2024

		import pandas as pd

		import os

Temperature-dependant solid domain constraints #1038

Temperature-dependant solid domain constraints #1038

Conversation

AmishgaAlphonius commented Feb 22, 2024 • edited Loading

Description of the problem

Description of the solution

How Has This Been Tested?

Documentation

Future changes

Comments

hepap left a comment

Choose a reason for hiding this comment

blaisb 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

AmishgaAlphonius Feb 27, 2024 • 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

Choose a reason for hiding this comment

Choose a reason for hiding this comment

blaisb commented Feb 27, 2024

AmishgaAlphonius commented Feb 22, 2024 •

edited

Loading

AmishgaAlphonius Feb 27, 2024 •

edited

Loading