Open-source geotechnical software, built in the open.
Slope stability, groundwater flow and rock mechanics — in Python, readable, tested and free.
English · Español
opengeorock.org · Contributing · Rules for AI agents · Licence
Almost every program a geotechnical engineer depends on is closed, expensive and opaque — a black box at the centre of decisions about dams, slopes, tunnels and foundations. When a factor of safety comes out at 1.32, there is no way to read why.
OpenGeoRock rebuilds those tools in the open: a computation core in plain Python, a modern desktop interface, and a test suite anyone can inspect. Every method ships with the reference case it was validated against, so results can be checked instead of believed.
Nine methods, each validated against a reference case:
| Method | Equilibrium satisfied | Validation error |
|---|---|---|
| Ordinary / Fellenius | moment | +0.06 % |
| Bishop simplified | moment | +0.02 % |
| Janbu simplified | force | +0.13 % |
| Janbu corrected | force | +0.10 % |
| Spencer | force + moment | +0.64 % |
| GLE / Morgenstern-Price | force + moment | +0.53 % |
| Lowe-Karafiath | force | +0.09 % |
| Corps of Engineers #1 | force | −0.5 % |
| Corps of Engineers #2 | force | −0.5 % |
The two Corps of Engineers methods — the Modified Swedish procedure — are validated slice by slice against the worked example published in USACE EM 1110-2-1902, Appendix G, whose tables print the interslice force and base normal columns.
Six search algorithms — grid, slope, auto-refine, block, path and simulated annealing — over circular and non-circular surfaces, with composite surfaces, slope limits and random-walk optimisation.
Focused search by window, line, point or tangent, which filters circles before evaluating them: on the reference case, a tangent to a weak layer leaves 17 evaluations out of 206.
Eighteen strength models (Mohr-Coulomb, undrained, Hoek-Brown and generalised Hoek-Brown, power curve, hyperbolic, Barton-Bandis, linear anisotropic, shear-normal and generalised anisotropic functions, SHANSEP, vertical stress ratio…), seven support types, distributed and line loads, pseudo-static seismic loading and tension cracks.
A full seepage engine, not a lookup table:
- T3 meshing over the material regions, conforming across interfaces by construction.
- Saturated steady state with an anisotropic conductivity tensor per material.
- Unsaturated flow with six permeability functions — simple, Brooks-Corey, Fredlund-Xing, Gardner, van Genuchten and user-defined — solved by Picard iteration with relaxation.
- Phreatic surface and seepage faces by nodal switching.
- Transient analysis in stages, using the mixed form of the Richards equation, modified Picard iteration and a factor of safety per stage: a rapid drawdown becomes a stability history.
- Coupling to stability: pressures interpolated at the slice bases, with suction handled through the extended Mohr-Coulomb envelope.
Seven distributions with relative minimum and maximum truncation, Monte Carlo and Latin hypercube sampling, cohesion–friction correlation, Global Minimum and Overall Slope analyses, probability of failure, reliability index, critical probabilistic surface, convergence plots and sensitivity sweeps ranked by influence.
Eurocode 7 partial factors (DA1-C1, DA1-C2, DA2, DA3), support-force back analysis, a Hoek-Brown parameter calculator, DXF import and export with geometry cleanup, an annotation layer, PDF reports, a Spanish / English interface and a command-line interface.
The suite is validated against external references and analytic identities, never against snapshots of its own output — a snapshot locks in whatever bug may be there.
| What | Check | Result |
|---|---|---|
| LEM methods | reference factor of safety | all within 0.7 % |
| Confined seepage | closed-form Darcy solution | 0.000 % |
| Layered media | harmonic and arithmetic averages | 0.000 % |
| Anisotropy | patch test with Kxy ≠ 0 | machine precision |
| Transient | erfc step response | 0.11 % |
| Transient at long time | must reproduce steady state | 0.0000 m |
| FE mesh | area equals the region's area | exact at every refinement |
| Back analysis | active = passive at FS = 1 | exact identity |
| Hoek-Brown | GSI = 100, D = 0 → mb = mi, s = 1, a = 0.5 | exact |
| DXF round trip | identical area, vertices on the originals | < 1e-6 |
Rapid drawdown reproduces the classical result without being told to: the factor of safety is lowest right after the level drops and recovers as the pressures dissipate.
Python 3.11 or newer.
git clone https://github.com/samuelsl27/OGR-Slip2D.git
cd OGR-Slip2D
python -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate
pip install -e .Open the application:
ogr-slip2d # or: python -m ogr_guiOr use it from Python:
from ogr_core.project import Project
from ogr_slip2d import BishopSimplified, GridSearch
project = Project.load("my_slope.ogr")
search = GridSearch(method=BishopSimplified(), num_slices=25)
result = search.run(project)
print(f"critical FoS = {result.critical.fos:.4f}")QT_QPA_PLATFORM=offscreen python tests/_runner.pyQT_QPA_PLATFORM=offscreen is needed because the GUI tests build real Qt
widgets; they simply never reach a display. Note that the project uses its
own runner in tests/_runner.py — pytest is not required, and the test
modules cannot be imported outside the runner.
To work on one area, run only that part of the suite:
python tests/_runner.py transient # files whose name contains it
python tests/_runner.py -k erfc # tests whose name contains it
python tests/_runner.py --list transient # show the selection, run nothingA pattern that matches nothing exits 2 rather than announcing an empty
success, and a filtered run prints a FILTERED RUN warning around its
totals — it is not evidence for a release.
About 71,500 lines of implementation and 61,800 of tests, 3188 of them passing, with no known failures. The full suite takes between 5 and 7½ minutes; the same code, unchanged, has taken 5:55, 6:19, 6:41 and 7:22 on the same machine.
| Package | Contents |
|---|---|
ogr_core/ |
Geometry, materials, loads, supports, project model, hydraulics, statistics, annotations, DXF |
ogr_slip2d/ |
Limit-equilibrium engine: methods, searches, slicer, focus, optimisation, back analysis |
ogr_fem2d/ |
Finite elements: meshing and seepage solvers |
ogr_gui/ |
PySide6 interface: canvas, dialogs, interpret windows, translations |
ogr_cli/ |
Command-line interface |
ogr_data/ |
Placeholder for the shared materials database (planned for v0.2.0) |
validacion/ |
Validation cases: model, expected values and their source |
spec/ |
Specifications (project constitution and features) |
docs/ |
Plans, audits and the changelog |
tests/ |
One file per feature area |
OGR Slip2D is the first of five programs.
| # | Program | Scope | Status |
|---|---|---|---|
| 01 | OGR Slip2D | 2D stability, FE seepage, probabilistic analysis | in development |
| 02 | OGR Data | Property database, correlations and charts | planned |
| 03 | OGR FEM2D | 2D elastoplastic finite elements | planned |
| 04 | OGR Slip3D | 3D limit-equilibrium surfaces | planned |
| 05 | OGR FEM3D | 3D finite elements with coupled flow | planned |
Next up in Slip2D: installers, more validation cases and the materials database.
Any help is welcome: code, validation against real cases, bug reports, or simply running the program on an actual project and reporting what broke.
Read CONTRIBUTING.md first: it explains the seven
conditions a change must satisfy and why each one exists — every one of
them traces back to a real problem in this project's history. The first is
the most important: numerical work is validated against something external,
never against a snapshot of the current output.
The most valuable thing you can contribute is a validation case. If you have a slope whose factor of safety you know from a publication, another program or a hand calculation, open an issue with the Validation case template. That is how confidence in a computation engine gets built.
The repository includes AGENTS.md with the full contract:
stack, commands, the seven rules, the workflow and an explicit list of what
not to do. There are also commands and skills in .claude/, and
specification templates in spec/.
AI-assisted code is accepted by the same standard as everything else: if you cannot explain why a line is there, do not submit it. Responsibility for what gets merged belongs to whoever signs it.
Contributions are accepted under the terms of CLA.md.
GNU Affero General Public License v3.0 or later (AGPL-3.0-or-later). The
full text is in LICENSE.
In practice:
- You may use OGR Suite freely, including for professional engineering work, and bill that work without paying or publishing anything. Running a program is not a restricted activity.
- If you modify it and let others use your version remotely over a network (a web service or SaaS), you must offer them the source of your version. Charging for that service is allowed; keeping it closed is not.
- There is no warranty of any kind. Results must be checked against independent calculations before anything is based on them. A project is signed by a person, not by a program.
Every file carries its SPDX identifier, and a test fails if any is missing.
Note on the licence change. Until v0.1.42 the project was GPL-3.0. It moved to AGPL in v0.1.43 so that a modified version offered as a network service has to publish its source — something the GPL does not require, because serving a program is not distributing it. Earlier changelogs still say GPL-3.0 on purpose: they record what was true at the time.
Samuel Sáez López — author, copyright holder and maintainer. Mining engineer and PhD candidate at the Universidad Politécnica de Cartagena (UPCT), in the Technology and Modelling in Civil, Mining and Environmental Engineering programme. His doctoral research on slope stability and rock mechanics feeds directly into the numerical methods published here.
Prof. Emilio Trigueros Tornero — academic supervisor of the doctoral research at UPCT and collaborator in the scientific direction of the project, contributing judgement in rock mechanics and research methodology.
Collaborators: Universidad Politécnica de Cartagena (academic framework) and IMGA S.L.P. — Ingeniería Minera, Geológica y Ambiental (professional collaborator).
So that the split of contributions leaves no room for doubt: Samuel Sáez López is the author and copyright holder of the source code; Emilio Trigueros Tornero contributes academic supervision and scientific collaboration.
If OGR Suite contributes to published work:
@software{ogr_suite,
author = {Sáez López, Samuel},
title = {{OGR Suite (OpenGeoRock)}: open-source geotechnical analysis
of slope stability and groundwater flow},
year = {2026},
url = {https://opengeorock.org},
note = {Universidad Politécnica de Cartagena. AGPL-3.0-or-later}
}© 2026 Samuel Sáez López — Universidad Politécnica de Cartagena · AGPL-3.0-or-later