Full rewrite of meshing for fractured domains by keileg · Pull Request #1576 · pmgbergen/porepy

keileg · 2026-01-19T14:14:46Z

Proposed changes

This PR is a major update of the functionality for meshing of fractured domains. The main changes are:

The fracture network classes have been almost completely rewritten. Identification of intersecting fractures is now handled by Gmsh's backport to open cascade. This also enabled us to implement truly elliptic fractures instead of the polygon approximations used previously.
The algorithm for tuning of mesh size has been improved to allow for more aggressive refinement towards areas of complex geometries.
The test suite for fracture meshing is to a large degree rewritten. The new (parametrized) tests take a more systematic approach to testing. Some old tests that did not fit into this framework have survived.
The tutorial for fracture meshing has been rewritten to fit in the context of the multiphysics models, and to give some ideas about the new meshing framework. This will require some coordination with Update tutorials #1545; I can give an overview of this in person.

Beyond the module fracs, the changes to the code base should be limited. Some tests have been updated, partly to ensure that the tests are not too computationally expensive.

Migration to the new code

While the update should not be API breaking, the practical implications of the mesh size arguments will change. It is recommended to read the new meshing tutorial carefully before migrating to the new functionality.

Please not that fine tuning of the new algorithm may occur over the coming months.

Suggested workflow for review

The natural starting point is the new tutorial.
The API to the meshing, including names of mesh size controls etc. should be critically reviewed.
The inner workings of the meshing algorithm is highly technical (I have tried to limit the technicalities, but the negotiations with Gmsh must be somewhere), so to some degree this is what it is. However, the interface between the components of the meshing could possibly be improved.
Test coverage for fracture meshing.
Check that tests outside the meshing framework have not been molested.
If the functionality can be used by others as part of the review, that would be ideal.

These are my initial thoughts, they may or may not be good.

EDIT 04.02.26
This PR is downstream from #1573. The merging of that PR (when approved) into this branch will impact the following files:

fracture_importer and its tests are fully rewritten in 1573, there is no need to review the changes in this branch.
The tutorial will get an extra section on importing model geometries. This should be a pure add-on compared to the current text.
There will be minor changes in a few other places, but my expectation is this will not be focal points of the present review.

What remains or have not been considered

We need coordination with the work on Let multiphysics models import meshes (or geometries) from gmsh #1517. This includes both mentioning of that work in the new tutorial and updates of tests. I suggest we coordinate in person.
There are some tests, mainly of convergence type, that fail due to changes in the underlying mesh. These should be updated, but not until we are sure the meshing algorithm is fixed.
Some of the slow functional tests that are skipped in regular runs are now even slower. As with the previous point, I will update these when things are settled. Also note that some of these can be sped up by considering fewer refinement levels - I am not sure if we should do this in the same go.

EDIT: Todo list when merging in #1573

Fix meshing_kwargs in the new load geometry mixin
In set_geometry of the load geometry mixin, the spatial dimension can be inferred from the fracture network class introduced in this PR.
The possibility to load geometries should be mentioned in the meshing tutorial.
Elliptic fractures should be properly used, both in fracture_import and its test.

Types of changes

What types of changes does this PR introduce to PorePy?
Put an x in the boxes that apply.

Minor change (e.g., dependency bumps, broken links).
Bugfix (non-breaking change which fixes an issue).
New feature (non-breaking change which adds functionality).
Breaking change (fix or feature that would cause existing functionality to not work as expected).
Testing (contribution related to testing of existing or new functionality).
Documentation (contribution related to adding, improving, or fixing documentation).
Maintenance (e.g., improve logic and performance, remove obsolete code).
Other:

Checklist

Put an x in the boxes that apply or explain briefly why the box is not relevant.

The documentation is up-to-date.
Static typing is included in the update.
This PR does not duplicate existing functionality.
The update is covered by the test suite (including tests added in the PR).
If new skipped tests have been introduced in this PR, pytest was run with the --run-skipped flag.

Merge work on gmsh object processing into main branch for new gmsh updates

New implementations, based on Gmsh OpenCASCADE to come

This is a helper module which should not be used externally

…cases

…ure functionality

Non-convex domains are no longer supported.

The tacit support for non-convex domains is hereby dropped

…iers

… plane

…fracture geometry

…://github.com/pmgbergen/porepy into 1477-methods-for-exporting-fracs-domains-gmsh

… and functionality

…ue_1469

…d intersection test

Raise error messages if relevant.

…test class

… PlaneFracture

Also maintenance and documentation

More systematic testing of possible scenarios. Also parametrization.

Implementation must be cleaned up

review-notebook-app · 2026-02-02T09:04:22Z

View / edit / reply to this conversation on ReviewNB

IvarStefansson commented on 2026-02-02T09:04:22Z
----------------------------------------------------------------

Is the writing of all this logging default behaviour? I find it too verbose.

keileg commented on 2026-02-02T11:39:49Z
----------------------------------------------------------------

Should we make the gmsh verbosity level a user parameter?

jwboth commented on 2026-02-02T12:22:31Z
----------------------------------------------------------------

I don't know why, but I like it. So if at least there is a way of turning on info logging from gmsh, OK with me.

review-notebook-app · 2026-02-02T09:04:23Z

View / edit / reply to this conversation on ReviewNB

IvarStefansson commented on 2026-02-02T09:04:22Z
----------------------------------------------------------------

Could we simplify to "It may be desirable to target the mesh resolution towards the vicinity of the fractures." ?

I think the sentence in parentheses is more than important enough to warrant a promotion out of the parentheses!

review-notebook-app · 2026-02-02T09:04:24Z

View / edit / reply to this conversation on ReviewNB

IvarStefansson commented on 2026-02-02T09:04:23Z
----------------------------------------------------------------

This seems to be a fundamental explanation that should come earlier.

jwboth commented on 2026-02-02T12:45:24Z
----------------------------------------------------------------

I agree. See my edits, that I keep for chronological understanding of my thoughts. This of course answers some of my questions.

review-notebook-app · 2026-02-02T09:04:24Z

View / edit / reply to this conversation on ReviewNB

IvarStefansson commented on 2026-02-02T09:04:24Z
----------------------------------------------------------------

I think we should add a section the functionality for reading from file to manually tune grids and/or saving run time for complex geometries.

keileg commented on 2026-02-02T11:38:57Z
----------------------------------------------------------------

Agree, will do when merging the changes from that project into this PR.

keileg · 2026-02-02T11:38:58Z

Agree, will do when merging the changes from that project into this PR.

View entire conversation on ReviewNB

keileg · 2026-02-02T11:39:50Z

Should we make the gmsh verbosity level a user parameter?

View entire conversation on ReviewNB

jwboth · 2026-02-02T12:22:32Z

I don't know why, but I like it. So if at least there is a way of turning on info logging from gmsh, OK with me.

View entire conversation on ReviewNB

jwboth · 2026-02-02T12:45:25Z

I agree. See my edits, that I keep for chronological understanding of my thoughts. This of course answers some of my questions.

View entire conversation on ReviewNB

review-notebook-app · 2026-02-02T12:50:12Z

View / edit / reply to this conversation on ReviewNB

jwboth commented on 2026-02-02T12:50:12Z
----------------------------------------------------------------

Use hyperlink to link to single phase flow tutorial.

It is somewhat unfortunate to communicate two ways of controlling meshing. Somewhat OK. But maybe consolidate and use the same in both tutorials?

review-notebook-app · 2026-02-02T12:50:13Z

View / edit / reply to this conversation on ReviewNB

jwboth commented on 2026-02-02T12:50:13Z
----------------------------------------------------------------

I think this is better than before! Of course, it remains a black box what the effective cell size will be, but I am OK with that. My understanding of the concrete example is. With refinement_threshold=1.0 it may have happened that some strange coarsening would take place with skewed cells with two objects just twice the target cell size apart from each other. With refinement_threshold=2.0 one manages to dictate a uniformly fine mesh with nice cell shapes in between the two objects.

Edit: It somehow does not surprise me based on my initial understanding, but maybe this is not a good example if the parameters are changed and the mesh looks the same. Better would be a visual support of the description of the target effect.

review-notebook-app · 2026-02-02T12:50:15Z

View / edit / reply to this conversation on ReviewNB

jwboth commented on 2026-02-02T12:50:14Z
----------------------------------------------------------------

maybe add some extra interpretation? Something along the lines of aiming to control the mesh within the hull of the fracture network? Just to make sure that one targets here the general mesh and delaying the transition towards the cell_size_boundary.OK also without, if this would be too much.

Edit: Looking at the subsequent plot, I am questioning what I just wrote. Which parameter was important for the control of the "hull size"? Based on the above edit, I am not sure whether it was the refinement_buffer or refinement_threshold.The latter would make more sense. Maybe flip the visual demonstration of the two keywords?

review-notebook-app · 2026-02-02T12:50:15Z

View / edit / reply to this conversation on ReviewNB

jwboth commented on 2026-02-02T12:50:15Z
----------------------------------------------------------------

As intersting for some applications as fracture propagation or fracture tip focus, maybe stress that it is neither possible to control the cell size close to fracture tips?

review-notebook-app · 2026-02-02T12:50:31Z

View / edit / reply to this conversation on ReviewNB

jwboth commented on 2026-02-02T12:50:31Z
----------------------------------------------------------------

Use hyperlink to link to single phase flow tutorial.

It is somewhat unfortunate to communicate two ways of controlling meshing. Somewhat OK. But maybe consolidate and use the same in both tutorials?

review-notebook-app · 2026-02-02T12:50:32Z

View / edit / reply to this conversation on ReviewNB

jwboth commented on 2026-02-02T12:50:32Z
----------------------------------------------------------------

I think this is better than before! Of course, it remains a black box what the effective cell size will be, but I am OK with that. My understanding of the concrete example is. With refinement_threshold=1.0 it may have happened that some strange coarsening would take place with skewed cells with two objects just twice the target cell size apart from each other. With refinement_threshold=2.0 one manages to dictate a uniformly fine mesh with nice cell shapes in between the two objects.

Edit: It somehow does not surprise me based on my initial understanding, but maybe this is not a good example if the parameters are changed and the mesh looks the same. Better would be a visual support of the description of the target effect.

review-notebook-app · 2026-02-02T12:50:33Z

View / edit / reply to this conversation on ReviewNB

jwboth commented on 2026-02-02T12:50:32Z
----------------------------------------------------------------

maybe add some extra interpretation? Something along the lines of aiming to control the mesh within the hull of the fracture network? Just to make sure that one targets here the general mesh and delaying the transition towards the cell_size_boundary.OK also without, if this would be too much.

Edit: Looking at the subsequent plot, I am questioning what I just wrote. Which parameter was important for the control of the "hull size"? Based on the above edit, I am not sure whether it was the refinement_buffer or refinement_threshold.The latter would make more sense. Maybe flip the visual demonstration of the two keywords?

review-notebook-app · 2026-02-02T12:50:34Z

View / edit / reply to this conversation on ReviewNB

jwboth commented on 2026-02-02T12:50:33Z
----------------------------------------------------------------

As intersting for some applications as fracture propagation or fracture tip focus, maybe stress that it is neither possible to control the cell size close to fracture tips?

IvarStefansson

Some effort! Only relatively minor comments, but quite a few of them. Let's discuss where indicated. Did not consider tests in this iteration.

IvarStefansson · 2026-02-05T10:03:20Z

src/porepy/fracs/_structured.py

+    gmsh.initialize()
+
+    domain_tag = network.domain_to_gmsh()
+    fracture_tags = network.fractures_to_gmsh()


Am I being unreasonable if I claim this is somewhat sparsely documented?

IvarStefansson · 2026-02-05T10:04:24Z

src/porepy/fracs/ellipse_fracture.py

@@ -0,0 +1,150 @@
+"""Contains classes representing two-dimensional fractures in 3D."""


Is there a reason the class is ellipTIC and the file ellipSE?

IvarStefansson · 2026-02-05T10:12:42Z

src/porepy/fracs/ellipse_fracture.py

+        )
+
+    def copy(self) -> Fracture:
+        """Return a copy of the fracture with the current vertices.


Is "current vertices" accurate?

IvarStefansson · 2026-02-05T10:13:12Z

src/porepy/fracs/ellipse_fracture.py

+            major_axis_angle=self.major_axis_angle,
+            strike_angle=self.strike_angle,
+            dip_angle=self.dip_angle,
+            index=self.index,


What does this imply about the interpretation of the index?

IvarStefansson · 2026-02-05T10:16:01Z

src/porepy/fracs/fracture.py

-    order/sorting of vertices has to be implemented explicitly.
+    """
+
+    def set_index(self, index: int) -> None:


See comment in ellipse_fracture.py. Should we specify intended use/interpretation?

IvarStefansson · 2026-02-12T10:06:42Z

src/porepy/fracs/fracture_network_3d.py

+
+        # For a boundary line to be an intersection, it must be shared by at least two
+        # fractures.
+        num_lines_occ = np.bincount(np.abs(bnd_lines).astype(int))


IvarStefansson · 2026-02-12T10:09:11Z

src/porepy/fracs/fracture_network_3d.py

+
+        ### Get hold of lines representing fractures and boundaries.
+        domain_entities = gmsh.model.get_entities(nd)
+        # TODO: If there is more than one domain entity (the domain is split into parts


Please resolve.

IvarStefansson · 2026-02-12T10:09:51Z

src/porepy/fracs/fracture_network_3d.py

+        # and assign the minimum mesh size among all occurrences of the point.
+        self._uniquify_mesh_size_dictionary(mesh_size_points)
+
+        # For lines that with no extra mesh size control points, assign an empty list.


Suggested change

# For lines that with no extra mesh size control points, assign an empty list.

# For lines with no extra mesh size control points, assign an empty list.

IvarStefansson · 2026-02-12T10:13:10Z

src/porepy/fracs/fracture_network_3d.py

@@ -2721,6 +1147,9 @@ def to_csv(self, file_name: Path, domain: Optional[pp.Domain] = None) -> None:

                Domain specification.

+        Raises:


I don't know if it is a priority, but we could create an issue about implementing it as exporting the parameters defining an elliptic fracture.

IvarStefansson · 2026-02-12T10:37:13Z

src/porepy/fracs/plane_fracture.py

+
+        """
+        pts = self.pts.T
+        num_pts = len(pts)


Suggested change

num_pts = len(pts)

num_pts = pts.shape[0]

IvarStefansson

Some effort! Only relatively minor comments, but quite a few of them. Let's discuss where indicated. Did not consider tests in this iteration.

keileg and others added 30 commits October 13, 2025 07:32

Merge branch 'develop' into issue_1478

76ddcdd

Merge branch 'issue_1469' into issue_1478

e5f653f

Merge pull request #1505 from pmgbergen/issue_1478

299f57f

Merge work on gmsh object processing into main branch for new gmsh updates

MAINT: Retire FractureNetwork3d find/split intersection methods

d6c6d98

New implementations, based on Gmsh OpenCASCADE to come

MAINT: FractureNetwork3d intersection processing in separate method

a3b9b0d

MAINT: Structured 3d fracture meshing makes use of gmsh

8036355

MAINT: Rename fracs/structured.py -> _structured

213a093

This is a helper module which should not be used externally

ENH: Add tests for ellipse_fracture functionality with parameterized …

1ee5a8d

…cases

DOC: Update test module docstring to accurately reflect ellipse_fract…

b8a93eb

…ure functionality

ENH: FractureNetwork2d exports general polygon domains in new format

742405e

TST: Update test of fracture network 2d meshing of general polygons

a9c44c3

Non-convex domains are no longer supported.

TST: Update test of 3d fracture network meshing with non-box domains

2ee6af6

The tacit support for non-convex domains is hereby dropped

WIP: FractureNetwork3d meshing of polytope domains

c384eba

ENH: Add test for ellipse_fracture tags to verify unique GMSH identif…

591b2aa

…iers

ENH: Add the test for checking elliptical boundary nodes in the given…

6ca89dd

… plane

DOC: Update docstring for test_gmsh_ellipse_nodes

7657192

DOC: Change the name of the method for the plane test

e19f211

ENH: Refactor tests for ellipse_fracture module and add new test for …

6355bce

…fracture geometry

Merge branch '1477-methods-for-exporting-fracs-domains-gmsh' of https…

297a25c

…://github.com/pmgbergen/porepy into 1477-methods-for-exporting-fracs-domains-gmsh

ENH: Rename and update test for Gmsh ellipse nodes to improve clarity…

c4e4d4a

… and functionality

Merge branch '1477-methods-for-exporting-fracs-domains-gmsh' into iss…

f89a04d

…ue_1469

ENH: Add a test for pp.create_mdg with elliptical fractures

505d6f6

ENH: Update tests for ellipse fractures, improve bounding box, and ad…

5c7ef95

…d intersection test

MAINT: Update test domain bounds for ellipical fracture meshing

fe63779

MAINT: create_mdg function parses elliptic fractures

b9fd877

Raise error messages if relevant.

TST: Move EllipticFracture mesh generation tests to fracture network …

1407e49

…test class

FEAT: ABC class PointBasedFracture is superclass for LineFracture and…

c4831c2

… PlaneFracture

FEAT: EllipseFracture inherits from ABC Fracture.

4fcbea8

Also maintenance and documentation

TST: Reworked tests for 2d fracture meshing

92167f0

More systematic testing of possible scenarios. Also parametrization.

TST: New test for 3d fracture mesh generation

0993686

DOC: Minor comment in fracture network meshing

acf4191

keileg force-pushed the meshing_rework branch from 8af4b02 to 8f3feb9 Compare January 29, 2026 14:29

keileg added 2 commits January 29, 2026 14:36

WIP: Remove fracture tips as mesh size points in 2d domains

ec4306e

Implementation must be cleaned up

TUT: Revised tutorial for meshing

07444d6

keileg force-pushed the meshing_rework branch from 8f3feb9 to 07444d6 Compare January 29, 2026 14:36

keileg mentioned this pull request Feb 3, 2026

Proper importing of 3D plane and elliptic fractures #1584

Open

7 tasks

keileg and others added 6 commits February 5, 2026 08:54

MAINT: Allow the user to set theGmsh verbosity level

9890235

MAINT: Change keywords for advanced mesh size control

6fbe6da

TUT: Update of meshing tutorial

7bf6c04

TST: Update tests after changes to mesh size parameters

f4b0fd1

DOC: Minor clarification and line lengths

3551d61

DOC and STY

b044eda

IvarStefansson reviewed Feb 12, 2026

View reviewed changes

		@@ -0,0 +1,150 @@
		"""Contains classes representing two-dimensional fractures in 3D."""

	# For lines that with no extra mesh size control points, assign an empty list.
	# For lines with no extra mesh size control points, assign an empty list.

		@@ -2721,6 +1147,9 @@ def to_csv(self, file_name: Path, domain: Optional[pp.Domain] = None) -> None:

		Domain specification.

		Raises:

Conversation

keileg commented Jan 19, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Proposed changes

Migration to the new code

Suggested workflow for review

What remains or have not been considered

Types of changes

Checklist

Uh oh!

review-notebook-app bot commented Feb 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

review-notebook-app bot commented Feb 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

review-notebook-app bot commented Feb 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

review-notebook-app bot commented Feb 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

keileg commented Feb 2, 2026

Uh oh!

keileg commented Feb 2, 2026

Uh oh!

jwboth commented Feb 2, 2026

Uh oh!

jwboth commented Feb 2, 2026

Uh oh!

review-notebook-app bot commented Feb 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

review-notebook-app bot commented Feb 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

review-notebook-app bot commented Feb 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

review-notebook-app bot commented Feb 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

review-notebook-app bot commented Feb 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

review-notebook-app bot commented Feb 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

review-notebook-app bot commented Feb 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

review-notebook-app bot commented Feb 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

IvarStefansson 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!

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!

IvarStefansson left a comment

Choose a reason for hiding this comment

Uh oh!

keileg commented Jan 19, 2026 •

edited

Loading

review-notebook-app bot commented Feb 2, 2026 •

edited

Loading

review-notebook-app bot commented Feb 2, 2026 •

edited

Loading

review-notebook-app bot commented Feb 2, 2026 •

edited

Loading

review-notebook-app bot commented Feb 2, 2026 •

edited

Loading

review-notebook-app bot commented Feb 2, 2026 •

edited

Loading

review-notebook-app bot commented Feb 2, 2026 •

edited

Loading

review-notebook-app bot commented Feb 2, 2026 •

edited

Loading

review-notebook-app bot commented Feb 2, 2026 •

edited

Loading

review-notebook-app bot commented Feb 2, 2026 •

edited

Loading

review-notebook-app bot commented Feb 2, 2026 •

edited

Loading

review-notebook-app bot commented Feb 2, 2026 •

edited

Loading

review-notebook-app bot commented Feb 2, 2026 •

edited

Loading