Python automation platform for FEMM finite element simulations.

py2femm generates Lua scripts from a Python API and executes them in FEMM via a REST server. It supports magnetics, electrostatics, heat flow, and current flow problems in 2D planar and axisymmetric geometries.

pyFEMM uses ActiveX and only works on Windows. py2femm takes a different approach:

• File-based Lua generation -- works with Wine on Linux, in Docker, or natively on Windows
• Geometry separated from physics -- describe a shape once, reuse it for magnetic, thermal, or electrostatic analysis
• REST server -- submit jobs from any client (Python, notebook, CI) to a Windows machine running FEMM
• Parametric workflows -- sweep dimensions, materials, or BCs from Python and collect results programmatically

┌──────────────────┐       HTTP/REST        ┌──────────────────┐
│  Python client   │  ──────────────────►   │  py2femm_server  │
│  (any platform)  │  ◄──────────────────   │  (Windows+FEMM)  │
│                  │      JSON results      │                  │
│  FemmProblem API │                        │  FastAPI + FEMM  │
│  FemmClient      │                        │  subprocess      │
└──────────────────┘                        └──────────────────┘

The client library (py2femm/) runs anywhere: it builds a FemmProblem, serializes it to a Lua script, and hands it to FemmClient. The server (py2femm_server/) runs on Windows where FEMM is installed. It receives Lua scripts over HTTP, launches FEMM as a subprocess, and returns CSV results.

FemmClient auto-detects the best transport: shared filesystem (WSL/local), remote REST, or environment variable override.

• Python >= 3.10
• FEMM 4.2 installed on Windows (default: C:\femm42\)

# Core library (script generation + client)
pip install -e .

# With REST server dependencies (FastAPI, uvicorn)
pip install -e ".[server]"

# Everything including dev tools
pip install -e ".[server,dev]"

Or run the interactive setup on Windows:

setup_femm.bat

start_femm_server.bat

Or manually:

python -m py2femm_server --host 0.0.0.0 --port 8082

To keep FEMM visible for debugging:

python -m py2femm_server --host 0.0.0.0 --port 8082 --show-femm

With the server running:

python examples/heatflow/heatsink/heatsink.py

from py2femm.femm_problem import FemmProblem
from py2femm.general import LengthUnit
from py2femm.geometry import Geometry, Node, Line
from py2femm.heatflow import HeatFlowMaterial, HeatFlowConvection

problem = FemmProblem(out_file="results.csv")
problem.heat_problem(units=LengthUnit.MILLIMETERS, type="planar",
                     precision=1e-8, depth=100, minangle=30)

# Add geometry, materials, BCs...
# Then write the Lua script:
problem.write("my_simulation.lua")

from py2femm.client import FemmClient

client = FemmClient(mode="remote", url="http://localhost:8082")
result = client.run(lua_script, timeout=120)
print(result.csv_data)

# Run a single Lua script
py2femm run my_simulation.lua --output results.csv

# Run all .lua files in a directory
py2femm run-batch simulations/ --output-dir results/

# Check server status
py2femm status

The examples/ directory contains worked examples across all four physics domains.

The heat sink example is split into two notebooks:

Baseline (heatsink_baseline.ipynb) reproduces FEMM Tutorial #7:

1. Build a 5-fin aluminum heat sink geometry and plot it
2. Define material (aluminum k=200) and boundary conditions (heat flux + convection)
3. Generate the Lua script and submit it to FEMM
4. Parse results and display FEMM temperature contour bitmap
5. Summary: T_avg, thermal resistance R_th

Parametric study (heatsink_parametric.ipynb) explores the design space:

1. Define the parameter grid (360 full-factorial combinations, ~176 valid after manufacturability filtering)
2. Run the sweep via FEMM (~10 min runtime)
3. 1D sensitivity plots (R_th and cross-section area vs each parameter)
4. 2D heatmaps of R_th over pitch ratio and duty cycle
5. Scaling analysis: best R_th per base width
6. Contact mode comparison (centered, single-fin aligned, between-fins)
7. Top configurations with geometry overlay

py2femm/
├── py2femm/                  # Core library
│   ├── femm_problem.py       #   FemmProblem — Lua script generator
│   ├── geometry.py            #   Node, Line, CircleArc, Geometry
│   ├── magnetics.py           #   Magnetic materials & BCs
│   ├── electrostatics.py      #   Electrostatic materials & BCs
│   ├── heatflow.py            #   Heat flow materials & BCs
│   ├── current_flow.py        #   Current flow materials & BCs
│   ├── general.py             #   FemmFields, LengthUnit, base classes
│   ├── cli.py                 #   CLI (run, status, run-batch)
│   └── client/                #   FemmClient (auto, local, remote)
│
├── py2femm_server/            # REST server (Windows)
│   ├── server.py              #   FastAPI endpoints
│   ├── executor.py            #   FEMM subprocess runner
│   ├── job_store.py           #   In-memory job state
│   ├── health.py              #   FEMM installation detection
│   └── watcher.py             #   Filesystem watcher mode
│
├── examples/
│   ├── heatflow/heatsink/     #   Heat sink tutorials + parametric study
│   ├── electrostatics/        #   Capacitor and field examples
│   └── magnetics/             #   Motor, solenoid, and machine examples
│
├── docs/superpowers/          # Design specs and implementation plans
├── tests/                     # Test suite
├── start_femm_server.bat      # Launch server on Windows
└── setup_femm.bat             # One-time environment setup

py2femm reads configuration from config/default.yml (generated by setup_femm.bat):

python:
  env_type: conda       # or venv
  env_name: my_env
  conda_root: C:\ProgramData\Anaconda3
femm:
  exe: C:\femm42\bin\femm.exe

Client auto-detection order:

1. Explicit mode/url arguments
2. /mnt/c/ exists -- local shared-filesystem mode (WSL)
3. PYFEMM_AGENT_URL environment variable -- remote mode
4. ~/.py2femm/config.yml -- remote mode

Design documents and implementation plans live in docs/superpowers/. To browse the documentation site locally:

pip install mkdocs-material
mkdocs serve

Then open http://localhost:8000 in your browser.

python -m pytest tests/ -v

Tests that require a live FEMM server are marked with pytest.mark.skip or guarded by server health checks. The unit tests (geometry, config, client mocking) run without FEMM.

Contributions are welcome. Please:

1. Install dev dependencies: pip install -e ".[server,dev]"
2. Run ruff check . and ruff format --check . before submitting
3. Add tests for new features
4. Follow existing code style: type hints, dataclasses, from __future__ import annotations

• FEMM by David Meeker -- the finite element engine that makes this all possible
• pyFEMM -- the original Python-FEMM bridge via ActiveX, which inspired the py2femm approach
• The FEMM community and tutorial authors

AGPL-3.0-or-later

Copyright (c) Riccardo Tinivella (tinix84@gmail.com)