Python Magnet Workflows

Python workflows for coupled electromagnetic-thermal simulations of high-field magnets using Feel++.

Features

• Iterative solver for coupled physics (electromagnetic + thermal + hydraulic)
• Multiple heat transfer correlations (Montgomery, Dittus-Boelter, Colburn, Silverberg)
• Water cooling system simulation with pressure drop calculations
• Support for helical and bitter magnets
• MPI parallel execution via Feel++

Requirements

System Requirements

• Feel++ >= 0.102.1 (must be installed separately)
• Python >= 3.11
• MPI implementation (OpenMPI or MPICH)

Feel installation see installation notes from official feelpp doc

• Add the Feelpp apt repository
• Install the following packages

  sudo apt install python3-feelpp-toolboxes-coefficientformpdes=0.111.0~preview.11-1 \
      libfeelpp-toolboxes1-coefficientformpdes=0.111.0~preview.11-1 \
      libfeelpp-toolboxes1-core=0.111.0~preview.11-1 \
      libfeelpp1=0.111.0~preview.11-1 \
      python3-feelpp-toolboxes-core=0.111.0~preview.11-1 \
      libfeelpp1=0.111.0~preview.11-1 \
      python3-feelpp=0.111.0~preview.11-1 \
      python3-feelpp-toolboxes-thermoelectric=0.111.0~preview.11-1 \
      libfeelpp-toolboxes1-thermoelectric=0.111.0~preview.11-1 \
      libfeelpp-toolboxes1-heat=0.111.0~preview.11-1 \
      libfeelpp-toolboxes1-electric=0.111.0~preview.11-1 \
      python3-feelpp-toolboxes-electric=0.111.0~preview.11-1 \
      python3-feelpp-toolboxes-heat=0.111.0~preview.11-1 \
      python3-feelpp-toolboxes-solid=0.111.0~preview.11-1  \
      libfeelpp-toolboxes1-solid=0.111.0~preview.11-1 \
      libstdc++-14-dev

Python Dependencies

See pyproject.toml for complete list. Major dependencies:

• numpy, scipy, pandas
• iapws (water properties)
• pint (unit management)
• feelpp and feelpp-toolboxes

Installation

Python virtual env

For Linux/Mac Os X:

$ python3 -m venv --system-site-packages magnetworkflows-env
$ source ./magnetworkflows-env/bin/activate
$ python3 -m pip install -e ".[dev]"

Use deactivate to qui the virtual environment

Note

• feelpp python packages must be installed before creating the virtual python env. see bellow for details
• --system-site-packages is mandatory since we cannot install feelpp python packages with pip

Debian/Ubuntu packages

• add lncmi debian repository
• apt update
• apt install ...

Quick Start

# Single configuration run
python -m python_magnetworkflows.cli \
  --mdata '{"M9Bitters":{"value":31000,"type":"bitter","filter":"M9Bitters_","flow":"M9Bitters-flow_params.json"}}' \
  M9Bitters-cfpdes-thelec-Axi-sim.cfg \
  --cooling mean \
  --eps 1.e-5

> **mdata** structure
>
> mdata is a dictonnary that holds informations about the magnets configuration:
> The main keys are the name of the magnet.
> The magnet config is stored in a sub-dictionnaty with:
> * value: the current value in A
> * type: the type of xxx for magnet (helix: Insert|bitter: Bitters|supra: Supras) -- see [python_magnetgeo](https://github.com/MagnetDB/python_magnetgeo) for details
> * filter: optional,
> * flow: a json file that contains parameters for water cooling -- see [python_magnetcooling](https://github.com/MagnetDB/python_magnetcooling) for details
> 
> support for type=supra not implemented

# Commissioning workflow (multiple current steps)
python -m python_magnetworkflows.commissioning \
  --mdata '{"HLtest":{"value":12000,"steplist":[10000,11000,12000],"type":"helix",...}}' \
  config.cfg \
  --cooling gradH

Architecture

python_magnetworkflows/
├── cli.py              # Main CLI entry point
├── commissioning.py    # Multi-step commissioning workflows
├── solver.py           # Feel++ solver interface
├── oneconfig.py        # Single configuration solver
├── cooling.py          # Heat transfer correlations
├── waterflow.py        # Hydraulic calculations
├── error.py            # Convergence error calculations
└── params.py           # Parameter extraction utilities

Configuration

Simulations require:

1. Feel++ CFG file (.cfg) - solver configuration
2. JSON model file - geometry and material parameters
3. Flow parameters JSON - pump curves and hydraulic data
4. Mesh file (.msh or .json)

See README.md examples section for details.

Examples

Running in container

Note

To check whether or not python_magnetworkflows is installed in the container

singularity exec /home/singularity/feelpp-v0.110.2.sif \
    dpkg -L python3-magnetworkflows

In devcontainer, you shall mount your data directory to /home/vscode to run singularity container,eg:

singularity exec -B /data/M9_M19020601-cfpdes-thmagel_hcurl-nonlinear-Axi-sim/data/geometries:/home/vscode \
    /home/singularity/hifimagnet-salome-9.8.4.sif \
    salome -w1 -t /opt/SALOME-9.8.0-UB20.04/INSTALL/HIFIMAGNET/bin/salome/HIFIMAGNET_Cmd.py args:M9_M19020601.yaml,--axi,--air,4,6

singularity exec -B /data/M9_M19020601-cfpdes-thmagel_hcurl-nonlinear-Axi-sim/data/geometries:/home/vscode \
    /home/singularity/hifimagnet-salome-9.8.4.sif \
    python3 -m python_magnetgeo.xao M9_M19020601-Axi_withAir.xao mesh  --group CoolingChannels --geo M9_M19020601.yaml []

mpirun -np 2 python -m python_magnetworkflows.workflows.cli HL-test-cfpdes-thelec-Axi-sim.cfg --eps 1.e-5

Running multiple configs at once

python -m python_magnetworkflows.run cli \
   --cfgfile M9Bitters/mean/M9Bitters-cfpdes-thmag_hcurl-nonlinear-Axi-sim.cfg
   --mdata '{"M9Bitters":{"value":31000,"type":"bitter","filter":"","relax":0,"flow":"M9Bitters/M9Bitters-flow_params.json"}}'
   --coolings "mean" "meanH" "grad" "gradH" "gradHZ" --hcorrelations "Montgomery" --frictions "Constant"

python -m python_magnetworkflows.run commissioning \
   --cfgfile M9Bitters/mean/M9Bitters-cfpdes-thmag_hcurl-nonlinear-Axi-sim.cfg \
   --mdata '{"M9Bitters":{"value":31000,"step":31000,"stepmax":1,"type":"bitter","filter":"","relax":0,"flow":"M9Bitters/M9Bitters-flow_params.json"}}' \
   --coolings "mean" "meanH" "grad" "gradH" "gradHZ" --hcorrelations "Montgomery" --frictions "Constant"

You need a tree structure like :

M9Bitters
|_ mean/M9Bitters-cfpdes-thmag_hcurl-nonlinear-Axi-sim.cfg
|_ meanH/M9Bitters-cfpdes-thmag_hcurl-nonlinear-Axi-sim.cfg
|_ grad/M9Bitters-cfpdes-thmag_hcurl-nonlinear-Axi-sim.cfg
|_ gradH/M9Bitters-cfpdes-thmag_hcurl-nonlinear-Axi-sim.cfg
|_ gradHZ/M9Bitters-cfpdes-thmag_hcurl-nonlinear-Axi-sim.cfg

testsuite

Create a setup for thelec axi sim for 'M9Bitters_HLtest' site

export MAGNETDB_API_KEY=
python -m python_magnetapi ...

Edit 'tmp/M9Bitters_HLtest/M9Bitters_HLtest-cfpdes-thelec-Axi-sim.json' and modify a U_ parameter

Run the workflow:

python -m python_magnetworkflows.cli run \
  '{HLtest":{"value":12000,"type":"helix","filter":"HLtest_","flow":"tmp/M9Bitters_HLtest/HLtest-flow_params.json"}, \
            "M9Bitters":{"value":31000,"type":"bitter","filter":"M9Bitters_","flow":"tmp/M9Bitters_HLtest/M9Bitters-flow_params.json"}}' \
  tmp/M9Bitters_HLtest/M9Bitters_HLtest-cfpdes-thelec-Axi-sim.cfg --cooling mean 

It must converge in 2 iterations