Skip to content

Installation & requirements

numgeo-ACT is a Python package. It drives the finite-element solver numgeo to simulate laboratory tests, so a working numgeo installation reachable from your PATH is a prerequisite.

Requirements

Component Purpose
Python 3.9+ Runtime for numgeo-ACT
numgeo (executable on PATH) Runs the element-test simulations
NumPy, SciPy Numerics and optional optimizers
pandas, openpyxl Reading the Excel experimental database
matplotlib Comparison plots and report figures

numgeo must be callable

numgeo-ACT locates numgeo through the system PATH. The helper ACT.utilities.runtime.get_numgeo_command() resolves the executable in a platform-independent way (Windows and Linux are supported). If numgeo cannot be found, calibrations will fail before any simulation is run — make sure numgeo (or the platform-specific executable name) is on your PATH.

Python dependencies

Install the scientific Python stack used by numgeo-ACT:

pip install numpy scipy pandas openpyxl matplotlib

Some optional optimization back-ends rely on additional packages:

  • The DEEM optimizer (the default and recommended choice) ships with numgeo-ACT.
  • Alternative optimizers are available through the mealpy and scipy back-ends, and a Bayesian optimization interface is provided in ACT.bayesian. Install the corresponding package only if you intend to use that back-end.

Obtaining numgeo-ACT

Availability

numgeo-ACT is developed by Jan Machaček. The tool is not yet freely available; release to the wider community is planned in the medium term. In the meantime, our group offers a free calibration service.

For access, collaboration or to have your material calibrated, visit the automatic calibration page or contact jan.machacek@tu-darmstadt.de.

Verifying the setup

Once numgeo is on your PATH and the Python dependencies are installed, a minimal check is to confirm that numgeo-ACT can resolve the solver:

from ACT.utilities import runtime

# Returns the command used to launch numgeo (raises if not found on PATH)
print(runtime.get_numgeo_command())

If this prints a valid command, you are ready to run your first calibration — continue with the Quickstart.