OpenPinch

OpenPinch is an open-source Python toolkit for advanced Pinch Analysis and Total Site Integration. It supports direct and indirect heat integration targeting, graph interpretation, Heat Pump and refrigeration screening, exergy and cogeneration post-processing, heat exchanger network synthesis, multi-period analysis, stream piece-wise linearisation (for variable heat capacity and phase change streams), and file-backed or schema-first workflows.

Full documentation is available at https://openpinch.readthedocs.io/en/latest/.

Install

Install the base package for validation, targeting, summaries, and schema-first Python workflows:

python -m pip install openpinch

Install optional extras only for the workflows that need them:

python -m pip install "openpinch[notebook]"      # Jupyter, Plotly graphs, Excel I/O
python -m pip install "openpinch[dashboard]"     # Streamlit dashboard
python -m pip install "openpinch[synthesis]"     # HEN synthesis, then run: idaes get-extensions
python -m pip install "openpinch[brayton_cycle]" # TESPy-backed Brayton-cycle tooling
python -m pip install "openpinch[tespy]"          # TESPy HPR targeting and performance maps
python -m pip install "openpinch[full]"          # all optional surfaces, including synthesis

OpenPinch currently requires Python >=3.14.2.

Both synthesis and full install the IDAES/Pyomo synthesis stack. Complete the IDAES installation before running solver-backed workflows:

idaes get-extensions

First Solve

OpenPinch exposes two package-root workflow classes. Use PinchProblem for one case and PinchWorkspace for named cases and scenarios.

from OpenPinch import PinchProblem

problem = PinchProblem(
    {
        "streams": [
            {
                "name": "Hot feed",
                "zone": "Process",
                "t_supply": 180.0,
                "t_target": 80.0,
                "heat_flow": 1000.0,
            },
            {
                "name": "Cold feed",
                "zone": "Process",
                "t_supply": 20.0,
                "t_target": 120.0,
                "heat_flow": 800.0,
            },
        ],
        "utilities": [],
    },
    project_name="First solve",
)
problem.validate()
problem.target.all_heat_integration()

print(problem.summary_frame())

Analysis is explicit: named methods execute work, while summaries, reports, plots, and exports consume prepared or cached state.

Packaged Resources

OpenPinch ships maintained sample cases and notebook workflows. The resource helpers below are useful repository tooling, but are not compatibility protected:

from OpenPinch.resources import (
    list_notebooks,
    list_sample_cases,
    notebook_metadata,
    sample_case_metadata,
)

print(list_sample_cases())
print(sample_case_metadata("basic_pinch.json").description)
print(list_notebooks())
print(notebook_metadata("01_first_solve_and_core_curves.ipynb").title)

Copy the notebook series from the CLI:

openpinch notebook -o notebooks

Notebooks and sample cases live in OpenPinch/tutorials/notebooks and OpenPinch/tutorials/sample_cases, with discovery and copying available through OpenPinch.resources.

The nineteen-notebook series progresses from first solve through multiperiod HPR, cogeneration, HEN synthesis, and publication workflows.

The CLI intentionally copies notebooks only. Solves, validation, graph export, Excel export, dashboards, and advanced targeting happen through Python.

Documentation Map

• Getting started: https://openpinch.readthedocs.io/en/latest/getting-started.html
• Workflow choice: https://openpinch.readthedocs.io/en/latest/overview/workflow-map.html
• Guides: https://openpinch.readthedocs.io/en/latest/guides/index.html
• API reference: https://openpinch.readthedocs.io/en/latest/api/index.html

Testing

Run the test suite locally:

uv sync --frozen --group dev
uv run --no-sync ruff check .
uv run --no-sync coverage run --branch --source=OpenPinch -m pytest --hypothesis-seed=20260715 -m "not solver"
uv run --no-sync coverage report --fail-under=95
uv run --no-sync python scripts/build_docs.py
uv run --no-sync python scripts/build_dist.py

Ubuntu runs the complete CI suite. Windows and macOS install the generated wheel and verify the core import, CLI, and packaged resources. Tests marked solver require external solver binaries; the release workflow installs the IDAES extensions and runs this gate automatically. Run it locally with uv run pytest -m solver when the required binaries are available.

Release Process

1. For a same-repository pull request targeting main, the PR workflow automatically advances an unchanged release version before validating it. A major, minor, or patch label takes precedence, followed by a matching title marker such as [minor]; the default is patch. The bump updates pyproject.toml, uv.lock, and .bumpversion.toml together without creating a tag. Fork pull requests remain read-only and must provide those synchronized forward-version changes in the contributor branch.
2. The ordered release-version job checks out and validates the updated PR head after the bump. Any later synchronize run or manual rerun recognizes the already-forward version without another commit. Merge only after the validation jobs, external-solver suite, and aggregate pr-gate result pass.
3. The main-branch workflow repeats the test, documentation, solver, build, and cross-platform artifact gates.
4. After those gates pass, it creates the annotated version tag and a draft GitHub release with checksummed release artifacts, then publishes the same distributions to TestPyPI. Preflight and postflight checks require the exact expected filenames and SHA-256 hashes, including safe recovery from a partial upload.
5. After TestPyPI succeeds:
   • it publishes the GitHub release before production PyPI
   • it dispatches the same workflow at the version tag
6. The tag-ref run verifies the source push, workflow, prerequisite jobs, and immutable build artifact ID, digest, and build attempt. It then requires the public release files to match that artifact byte-for-byte without rebuilding and waits at the protected pypi environment.
7. Approve that deployment to use PyPI Trusted Publishing; the workflow uploads the verified distributions and confirms the version through the PyPI API.

The automatic pull-request bump is idempotent: a candidate already greater than the base is validated without another commit, while a candidate behind the base fails for manual reconciliation. Version bumping does not create a tag. The release workflow owns tags and rejects malformed versions, mismatched lock metadata, or an existing tag that points anywhere other than the main-branch release commit. If production publication or its availability check fails after the GitHub Release becomes public, open the original tag run and select Re-run failed jobs. Exact index preflight, skip-existing, and a separately retryable availability check make an absent, partial, or already-complete release recoverable without accepting mismatched files. Do not start a fresh tag dispatch when the upload may already have succeeded.

Build the documentation locally:

uv run scripts/build_docs.py

History and Citation

OpenPinch started in 2011 as an Excel workbook with macros. The Python implementation began in 2021 to make the workflows scriptable and testable.

In publications and forks, please cite and link the foundational article and this repository:

Timothy Gordon Walmsley, 2026. OpenPinch: An Open-Source Python Library for Advanced Pinch Analysis and Total Site Integration. Process Integration and Optimization for Sustainability. https://doi.org/10.1007/s41660-026-00729-6

Contributors

Founder: Tim Walmsley, University of Waikato

Stephen Burroughs, Benjamin Lincoln, Alex Geary, Harrison Whiting, Khang Tran, Roger Padulles, Jasper Walden, Caleb Archer

License

OpenPinch is released under the MIT License. See LICENSE for details.