[GeoMechanicsApplication] Integrate the functionality for UPw interfaces by markelov208 · Pull Request #14266 · KratosMultiphysics/Kratos

markelov208 · 2026-03-09T14:06:22Z

📝 Description
A brief description of the PR.

Created integration tests that perform computations for:

column with a horizontal_interface
column with a vertical_interface
column with a horizontal_interface using diff_order_elements

The obtained solutions (displacement and pressure values) are compared with solutions obtained for

column
column using diff order elements.

Tests are Quasi-static and displacement, pressure and fluid flux are compared for all time steps.

Added PUCoupling , Permeability and FluidBodyFlow to UPwInterfaceElementContributions .

This requires allows to get proper values of pressure and fluid fluxes for these tests but it breaks two tests:

building_pit
test_interface_prestress
They have been fixed by adding "IGNORE_UNDRAINED" : true for interface material properties.

A hydraulic discharge flow field is disconnected at the horizontal interface because UPwInterfaceElement does not calculate the discharge now. Therefore a new issue [GeoMechanicsApplication] Add hydraulic discharge calculation in UPw interface element is created.

Fixed error with fluid_flux calculation in SmallStrainUPwDiffOrderElement::CalculateOnIntegrationPoints. A unit test is added.

markelov208 · 2026-03-10T12:20:01Z

Hi @WPK4FEM , @rfaasse and @avdg81 , I've added Permeability to UPwInterfaceElementContributions then building_pit and test_interface_presstress failed. Adding TRANSVERSAL_PERMEABILITY and DYNAMIC_VISCOSITY to MaterialParameters solves them. Is it acceptable?

WPK4FEM · 2026-03-10T13:06:17Z

Hi @WPK4FEM , @rfaasse and @avdg81 , I've added Permeability to UPwInterfaceElementContributions then building_pit and test_interface_presstress failed. Adding TRANSVERSAL_PERMEABILITY and DYNAMIC_VISCOSITY to MaterialParameters solves them. Is it acceptable?

Hi @markelov208, Yes I think that that is acceptable. Its what the water part of the U-Pw interface element needs.

rfaasse · 2026-03-10T14:26:47Z

Hi @WPK4FEM , @rfaasse and @avdg81 , I've added Permeability to UPwInterfaceElementContributions then building_pit and test_interface_presstress failed. Adding TRANSVERSAL_PERMEABILITY and DYNAMIC_VISCOSITY to MaterialParameters solves them. Is it acceptable?

In principle no problem for me. However, since we use IGNORE_UNDRAINED in most of the model, I would not expect this to be necessary. I think if you'd add IGNORE_UNDRAINED: true to the material parameters of the interface (instead of the transversal permeability and dynamic viscosity), it'll also work (and that might be a bit more intuitive, since the rest of the materials also have IGNORE_UNDRAINED set to true.

rfaasse · 2026-03-10T14:28:01Z

+                "MINIMUM_RELATIVE_PERMEABILITY" : 0.0001,
+                "TRANSVERSAL_PERMEABILITY"      : 5.0e-4,
+                "DYNAMIC_VISCOSITY"             : 1.0e-3


As mentioned in the other comment, I think this would also work:

Suggested change

"MINIMUM_RELATIVE_PERMEABILITY" : 0.0001,

"TRANSVERSAL_PERMEABILITY" : 5.0e-4,

"DYNAMIC_VISCOSITY" : 1.0e-3

"IGNORE_UNDRAINED" : true,

Because in that case, the contributions for permeability would be ignored by the interface element

Thank you for the suggestion! It works.

rfaasse

Hi Gennady, thank you for adding the integration tests for the interface! As mentioned the PR will likely change due to the fluid body flow etc, but these are my intermediate comments!

rfaasse · 2026-03-11T14:59:45Z

+            "WATER_PRESSURE", no_pressure_time, no_pressure_output, [5]
+        )[0]
+
+        self.assertGreater(baseline_top_pressure - no_pressure_top_pressure, 900.0)


What is the reason we use an assertGreater instead of assertAlmostEqual here?

The test has been generated by AI and I kept it just in case. Right now I removed it because it checks the imposed pressure. It should be somewhere for the tables/boundaries.

rfaasse · 2026-03-12T07:51:43Z

+Displacement and water pressure values obtained with and without the interface are compared for all nodes. The following pairs are used:
+
+- `column_horizontal_interface` and `column`,
+- `column_vertical_interface` and `column`,
+- `column_horizontal_interface_diff_order_elements` and `column_diff_order_elements`.


If I understand correctly, this means we check that adding the interface does not have an effect on the results. That probably means we have high stiffnesses, such that the interfaces stay closed (correct me if I'm wrong).

I'm not sure, just asking, but would there be a benefit in having at least 1 test which does have an open interface (i.e. would it increase the covered paths within the element)?

You are right these tests show that the solution is the same with/without an interface and the interface stays rather thin. If stiffness values are lower then the interface gets thicker and a solution changes. It is not clear what to check in this case.

The stiffness values of the interface element will have an effect on the obtained displacement (additional deformation compared to the situation where there is no interface, assuming the normal stiffness is not nearly "infinite"). And I suspect the transverse permeability has an impact on the groundwater flow (if it is less permeable than the surrounding soil, I mean).

I think it's wise to double-check the obtained results with @WPK4FEM.

rfaasse · 2026-03-12T09:33:45Z

+Begin Table 2 TIME WATER_PRESSURE
+    0.0000000000 0.0000000000
+    1.0000000000 1000.0000000000
+End Table


Since the water pressure is time dependent, does it make sense to add asserts for more time steps than just the latest ones?

Correct, now tests check parameters for all times steps.

rfaasse · 2026-03-12T09:34:48Z

+        "max_line_search_iterations": 5,
+        "first_alpha_value": 0.5,
+        "second_alpha_value": 1.0,
+        "min_alpha": 0.1,
+        "max_alpha": 2.0,
+        "line_search_tolerance": 0.5,


I don't think these params are needed, since we don't use line search

Suggested change

"max_line_search_iterations": 5,

"first_alpha_value": 0.5,

"second_alpha_value": 1.0,

"min_alpha": 0.1,

"max_alpha": 2.0,

"line_search_tolerance": 0.5,

It might be GitHub, but I still see these parameters here (and in the other project parameter files)

Sorry it seems they are pop-up due to changing the folder structure. removed

rfaasse · 2026-03-12T09:36:40Z

+                "kratos_module": "KratosMultiphysics.GeoMechanicsApplication",
+                "process_name": "ApplyK0ProcedureProcess",
+                "Parameters": {
+                    "model_part_name": "PorousDomain.settlement_computational_model_part"


Just to validate my understanding, this means it's the intention that K0 is done for the interfaces as well, right?

removed. Initially, the test was falling with Soil. Tried to find the reason.

markelov208

Hi Richard, many thanks for the review. It helped me to improve these tests and to learn better integration tests in general.

markelov208 · 2026-03-12T15:24:51Z

+            "WATER_PRESSURE", no_pressure_time, no_pressure_output, [5]
+        )[0]
+
+        self.assertGreater(baseline_top_pressure - no_pressure_top_pressure, 900.0)


The test has been generated by AI and I kept it just in case. Right now I removed it because it checks the imposed pressure. It should be somewhere for the tables/boundaries.

markelov208 · 2026-03-12T15:41:09Z

+        project_parameters_filenames = (
+            project_parameters_filenames or self.project_parameters_filenames
+        )
+        status = run_geo_settlement.run_stages(case_path, project_parameters_filenames)


Frankly speaking, I just grabbed something that I met first.

markelov208 · 2026-03-12T16:31:06Z

+                "kratos_module": "KratosMultiphysics.GeoMechanicsApplication",
+                "process_name": "ApplyK0ProcedureProcess",
+                "Parameters": {
+                    "model_part_name": "PorousDomain.settlement_computational_model_part"


removed. Initially, the test was falling with Soil. Tried to find the reason.

markelov208 · 2026-03-13T11:45:37Z

+    def test_horizontal_interface(self):
+        file_path = test_helper.get_file_path(os.path.join("UPw_interface", "column"))
+        output_data = self._run_two_stage_soil_case(file_path)
+        self.assertTrue(output_data.get("results"))
+
+    def test_horizontal_interface_diff_order(self):
+        file_path = test_helper.get_file_path(
+            os.path.join("UPw_interface", "column_diff_order_elements")
+        )
+        output_data = self._run_two_stage_soil_case(file_path)
+        self.assertTrue(output_data.get("results"))
+
+    def test_vertical_interface(self):
+        file_path = test_helper.get_file_path(
+            os.path.join("UPw_interface", "column_vertical_interface")
+        )
+        output_data = self._run_two_stage_interface_case(file_path)
+        self.assertTrue(output_data.get("results"))


Yes, these tests just run. Now removed them because these computations are also done in other tests.

markelov208 · 2026-03-13T11:54:22Z

+                    if pair not in seen_pairs:
+                        seen_pairs.add(pair)
+                        node_id_pairs.append(pair)
+
+        return node_id_pairs


a vertical interface consists of two elements and repeated pairs of nodes exist.

markelov208 · 2026-03-13T12:06:11Z

+Begin Table 2 TIME WATER_PRESSURE
+    0.0000000000 0.0000000000
+    1.0000000000 1000.0000000000
+End Table


Correct, now tests check parameters for all times steps.

rfaasse

Hi Gennady, apologies for not re-reviewing sooner, I missed the mail. To me, this looks almost good to go, I just have a small suggestion and one question left. Other than that, I think it's a good idea if @WPK4FEM and @avdg81 still have a look 👍

rfaasse · 2026-04-08T06:52:47Z

+        "max_line_search_iterations": 5,
+        "first_alpha_value": 0.5,
+        "second_alpha_value": 1.0,
+        "min_alpha": 0.1,
+        "max_alpha": 2.0,
+        "line_search_tolerance": 0.5,


It might be GitHub, but I still see these parameters here (and in the other project parameter files)

rfaasse · 2026-04-08T07:00:05Z

+        project_parameters_filenames = (
+            project_parameters_filenames or self.project_parameters_filenames
+        )
+        status = run_geo_settlement.run_stages(case_path, project_parameters_filenames)


In that case, since run_geo_settlement is at this moment not that generic, I'd propose to use the python workflow 👍

rfaasse · 2026-04-08T07:06:13Z

                "kratos_module": "KratosMultiphysics.GeoMechanicsApplication",
                "process_name": "ApplyK0ProcedureProcess",
                "Parameters": {
-                    "model_part_name": "PorousDomain.settlement_computational_model_part"
+                    "model_part_name": "PorousDomain.Soil"
                }
            }


Does this mean that K0 should not be applied to the interfaces?

There are no any effects of the model part name. Perhaps, this is a topic for the next study.

rfaasse · 2026-04-08T07:17:08Z

+                    if pair not in seen_pairs:
+                        seen_pairs.add(pair)
+                        node_id_pairs.append(pair)
+
+        return node_id_pairs


Okay, I still don't fully understand the code then, because seen_pairs is a set (so only contains unique items if I'm correct) and a pair is only added to node_id_pairs if it's not in the set. Could you explain this a bit more?

avdg81 · 2026-04-08T07:37:04Z

Hi Gennady, apologies for not re-reviewing sooner, I missed the mail. To me, this looks almost good to go, I just have a small suggestion and one question left. Other than that, I think it's a good idea if @WPK4FEM and @avdg81 still have a look 👍

@rfaasse: I'm looking into it 👍

avdg81

Hi Gennady,

Thanks a lot for all the work that you've done for this PR. I like that you've set things up in a systematic way 👍 I have several suggestions and a few questions.

In addition, I have found two files that define a constitutive law for interface elements, but which don't have IGNORE_UNDRAINED set (is that intentional?):

line_interface_elements/common/MaterialParameters.json
test_mohr_coulomb_with_tension_cutoff/interface_coulomb_2plus2/MaterialParameters.json (this one doesn't have any pore water properties defined)

One more thing: can you also please update this branch with the latest changes from master and then reformat the code (using kp format), to make sure the new JSON files are formatted correctly? Thanks.

avdg81 · 2026-04-07T15:35:35Z

+    // Arrange
+    auto p_element =
+        CreateSmallStrainUPwDiffOrderElementWithUPwDofs(CreatePropertiesForUPwDiffOrderElementTest());
+    SetSolutionStepValuesForGeneralCheck(p_element); // uniform pressure + gravity


When I looked at this function, I noticed that it only sets displacements at three of the six nodes. Not sure what happens at the mid-side nodes. This is not related in any way to your changes, but I was curious about what we prescribe and then I happened to see this. Perhaps just something to think about.

this is a separate action due to replace of auto with Vector . In release mode the flux was always 0. Sorry the goal was to check the non-zero flux only.

avdg81 · 2026-04-08T13:57:31Z

+    def _run_two_stage_case(self, case_path, project_parameters_filenames=None):
+        project_parameters_filenames = (
+            project_parameters_filenames or self.project_parameters_filenames
+        )


When we call _run_two_stage_case in our test cases, we always provide a value for project_parameters_filenames. That is, in our tests, the value will never be None. Therefore, I would suggest to remove the check and just use what is being passed. Then, you also no longer need to define self.project_parameters_filenames.

avdg81 · 2026-04-08T14:11:02Z

+    def _all_times(output_data):
+        times = GiDOutputFileReader.get_time_steps_from_first_valid_result(output_data)
+        if not times:
+            raise RuntimeError("No valid output time steps found")
+        return times


I feel this approach is a bit tricky, since now the time steps depend on the produced output, which may be incorrect. I suspect that we know in advance which time steps will be calculated, so I would prefer to have hard-coded lists of expected time steps.

avdg81 · 2026-04-08T14:29:37Z

+        }
+
+    @staticmethod
+    def _read_mdpa_nodes(mdpa_file_path):


I'm not sure, but perhaps you can use read_coordinates_from_post_msh_file from test_helper.py? If you really need to have a map from node ID to position, we could add a second helper function (e.g. extract_node_map_from_post_msh_file) that does just that (see variable node_map in the body of read_coordinates_from_post_msh_file), and then let read_coordinates_from_post_msh_file call extract_node_map_from_post_msh_file and produce the coordinates list from that.

Thank you for the info. I tried to use read_coordinates_from_post_msh_file and met a situation that in column_vertical_interface folder .post.msh file does not include all nodes. Nodes used only in the vertical interface are missed. Perhaps, we can discuss it.

avdg81 · 2026-04-08T14:54:35Z

Wow, this is a lot of test code. I see that many things have been set up in a very generic way, which is generally a good idea. For test code, however, and in particular for asserting certain results, it may be clearer to hard-code, for instance, which times or at which nodes you'd like to check something. That may help to reduce this test code a bit. For now I would suggest to keep the code "as-is", but we may want to simplify it a bit more in a separate PR. What do you think?

That is AI. Now I reduced the code a bit.

markelov208 · 2026-04-09T10:50:39Z

Hi Gennady,

Thanks a lot for all the work that you've done for this PR. I like that you've set things up in a systematic way 👍 I have several suggestions and a few questions.

In addition, I have found two files that define a constitutive law for interface elements, but which don't have IGNORE_UNDRAINED set (is that intentional?):

line_interface_elements/common/MaterialParameters.json

test_mohr_coulomb_with_tension_cutoff/interface_coulomb_2plus2/MaterialParameters.json (this one doesn't have any pore water properties defined)

One more thing: can you also please update this branch with the latest changes from master and then reformat the code (using kp format), to make sure the new JSON files are formatted correctly? Thanks.

the new tests fail when IGNORE_UNDRAINED is removed

markelov208

Hi Richard and Anne, many thanks for the reviews and suggestions how to improve the tests. Unfortunately, I was not able to use all these suggestions. Hope we can discuss them.

markelov208 · 2026-04-08T07:52:28Z

+                    if pair not in seen_pairs:
+                        seen_pairs.add(pair)
+                        node_id_pairs.append(pair)
+
+        return node_id_pairs


As far as I understand a set is faster and a list preserves the order. The number of pairs is small in these tests so only the list can be used. Removed seen_pairs.

markelov208 · 2026-04-08T07:54:12Z

+        "max_line_search_iterations": 5,
+        "first_alpha_value": 0.5,
+        "second_alpha_value": 1.0,
+        "min_alpha": 0.1,
+        "max_alpha": 2.0,
+        "line_search_tolerance": 0.5,


Sorry it seems they are pop-up due to changing the folder structure. removed

markelov208 · 2026-04-09T08:55:00Z

+        project_parameters_filenames = (
+            project_parameters_filenames or self.project_parameters_filenames
+        )
+        status = run_geo_settlement.run_stages(case_path, project_parameters_filenames)


switched to run_multiple_stages, this required some changes in it and adding K0 parameters in material files.

markelov208 · 2026-04-09T08:57:36Z

                "kratos_module": "KratosMultiphysics.GeoMechanicsApplication",
                "process_name": "ApplyK0ProcedureProcess",
                "Parameters": {
-                    "model_part_name": "PorousDomain.settlement_computational_model_part"
+                    "model_part_name": "PorousDomain.Soil"
                }
            }


There are no any effects of the model part name. Perhaps, this is a topic for the next study.

markelov208 · 2026-04-09T09:22:32Z

+    // Arrange
+    auto p_element =
+        CreateSmallStrainUPwDiffOrderElementWithUPwDofs(CreatePropertiesForUPwDiffOrderElementTest());
+    SetSolutionStepValuesForGeneralCheck(p_element); // uniform pressure + gravity


this is a separate action due to replace of auto with Vector . In release mode the flux was always 0. Sorry the goal was to check the non-zero flux only.

markelov208 · 2026-04-09T11:01:30Z

+- Constitutive law: `GeoIncrementalLinearElasticInterfaceLaw`
+- `INTERFACE_NORMAL_STIFFNESS = 6.4e10`
+- `INTERFACE_SHEAR_STIFFNESS = 2.6000869565e10`
+- `TRANSVERSAL_PERMEABILITY = 5.0e-4`


Are k_{\perp} and m^2 OK?

markelov208 · 2026-04-09T11:05:15Z

It is just happened because a one stage calculation did not work due to my mistakes in settings and I looked at building_pit model as an example.

markelov208 · 2026-04-09T11:06:52Z

That is AI. Now I reduced the code a bit.

markelov208 · 2026-04-09T12:25:30Z

+        }
+
+    @staticmethod
+    def _read_mdpa_nodes(mdpa_file_path):


Thank you for the info. I tried to use read_coordinates_from_post_msh_file and met a situation that in column_vertical_interface folder .post.msh file does not include all nodes. Nodes used only in the vertical interface are missed. Perhaps, we can discuss it.

avdg81

I feel this PR is good to go. Thanks a lot for processing the review suggestions. 👍

WPK4FEM

Hi Gennady,
Lets put this in master.
Wijtze Pieter

added integration tests

2a8afc8

markelov208 self-assigned this Mar 9, 2026

markelov208 requested a review from a team as a code owner March 9, 2026 14:06

formatted with black

ea33682

markelov208 linked an issue Mar 9, 2026 that may be closed by this pull request

[GeoMechanicsApplication] Integrate the functionality for UPw interfaces #14042

Closed

markelov208 added 6 commits March 9, 2026 17:07

Merge branch 'master' into geo/14042-integration-tests-for-UPw-interface

0c236f3

use small letters for test_UPw_interface.py

d1234f1

renamed class calls

cac24a3

added extra Material parameters

27e74d4

removed code smells

38a8548

added interface material parameters to building_pit

06a2f1b

markelov208 requested review from WPK4FEM, avdg81 and rfaasse March 10, 2026 12:17

rfaasse reviewed Mar 10, 2026

View reviewed changes

markelov208 added 2 commits March 10, 2026 21:15

used IGNORE_UNDRAINED true

3f0891a

Merge branch 'master' into geo/14042-integration-tests-for-UPw-interface

36c8284

rfaasse reviewed Mar 12, 2026

View reviewed changes

processed review comments

6cfc88b

markelov208 commented Mar 13, 2026

View reviewed changes

markelov208 requested a review from rfaasse March 13, 2026 12:57

markelov208 added 6 commits March 13, 2026 14:32

Merge branch 'master' into geo/14042-integration-tests-for-UPw-interface

905d1b1

fixed error with fluid_flux and added unit test

f8422f3

fixed code smells

5ee658d

added IGNORE_UNDRAINED true

98684ee

added checks

c9d9ef6

minor change

103bc92

markelov208 added 3 commits March 14, 2026 00:37

reverted checks

d7179d3

clang-format

ac5b9ec

reverted test_U_Pw_interface_element.cpp

c6669aa

rfaasse reviewed Apr 8, 2026

View reviewed changes

avdg81 reviewed Apr 8, 2026

View reviewed changes

markelov208 added 10 commits April 9, 2026 11:02

processed 2 batch of the review comments

a847b74

Merge branch 'master' into geo/14042-integration-tests-for-UPw-interface

6201df8

cleaned mdpa files

124da4c

formatted project and material files

1a53949

removed max_line_search_iterations etc

f8c365b

removed non-soil nodes from soil node list

2cfb784

removed empty Tables

60ad84d

moved filename patters to where they used

d87418d

cleaned a bit test_upw_interface.py

46d351e

inlined _top_vector_component_differences

5adc47a

corrected README.md

0f84e0e

markelov208 commented Apr 9, 2026

View reviewed changes

markelov208 requested review from avdg81 and rfaasse April 9, 2026 12:31

avdg81 approved these changes Apr 10, 2026

View reviewed changes

WPK4FEM approved these changes Apr 13, 2026

View reviewed changes

markelov208 merged commit 6a81a1f into master Apr 13, 2026
10 checks passed

markelov208 deleted the geo/14042-integration-tests-for-UPw-interface branch April 13, 2026 09:51

Conversation

markelov208 commented Mar 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

markelov208 commented Mar 10, 2026

Uh oh!

WPK4FEM commented Mar 10, 2026

Uh oh!

rfaasse commented Mar 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

rfaasse left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

markelov208 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

rfaasse left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

markelov208 commented Mar 9, 2026 •

edited

Loading

rfaasse commented Mar 10, 2026 •

edited

Loading