Skip to content

DLR-RY/TwoPhaseFlow

Repository files navigation

twophaseflow

the TwoPhaseFlow Library adds new surface tension and phase change models to OpenFOAM and provides benchmark cases for verification

Documentation

The available models and solvers are documentated in:

https://arxiv.org/abs/2103.00870

Getting Started

Prerequisites

Requires OpenFOAM v1812:

https://www.openfoam.com/download/release-history.php

The branch of2106 works with of2106

Installing

    git clone https://github.com/DLR-RY/TwoPhaseFlow
    cd TwoPhaseFlow
    ./Allwmake
    ./get-gmsh.sh # will install gmsh version 306 as gmshv306
    # for AMR
    git submodule update --init --recursive
    cd modules/multiDimAMR/
    ./Allwmake

running testsuite

make sure that the desired openfoam installation is sourced e.g. v1812 and that python is installed with a version >= 3.6 (miniconda is a great option, but anaconda works as well)

    python -m venv env # creats virtual python enviroments (optional step)
    pip install oftest

    py.test # runs the tests
    py.test --writeNSteps=1 run/ # test all testcases in run

Authors

  • Henning Scheufler

adaptive mesh refinement with multiple regions

AMR with multiple regions does not work in version of1812 but it is fixed in newer versions.

To fix this apply the patch (assumes openfoam is already source):

    cp  patches/multiRegionAMR.patch $WM_PROJECT_DIR
    cp  patches/tableBase.patch $WM_PROJECT_DIR
    cp  patches/surfaceFieldValue.patch $WM_PROJECT_DIR
    cd $WM_PROJECT_DIR
    git apply multiRegionAMR.patch
    git apply tableBase.patch
    git apply surfaceFieldValue.patch

details see:

https://develop.openfoam.com/Development/openfoam/-/issues/1676

https://develop.openfoam.com/Development/openfoam/-/issues/1753

License

This project is licensed under the GPL v3 License - see the LICENSE.md file for details

running Benchmarks

    ./get-gmsh.sh # install gmsh
    pip install casefoam

The run/benchmark cases are run with

    cd run/benchmark/phaseChange/suckingInterface/
    python genCases.py # generates the study based and template case (here StefanProblem)
    ./Allrun # runs all the created testcases
    python getResults.py # to see results

Alternatively, the runAll.sh can be executed in the folder.

Note:

that some cases use the slurm queuing system and call sbatch Allrun_Slurm in the Allrun script, so you might need to modify it in the template case