This repository accompanies the training session on DMD analysis of unsteady flows given at the 17th OpenFOAM workshop in Cambridge (2022).
Without modification, the simulations will only work with OpenFOAM-v2206. You can use the pre-compiled binaries or the provided Singularity container. To build the container, run:
sudo singularity build of2206.sif docker://andreweiner/of_pytorch:of2206-py1.11.0-cpu
When working with the container, use the Allrun.singularity instead of the Allrun scripts and make sure to set the image location in each script.
flowTorch is a Python library for the analysis and modeling of fluid flows. We use flowTorch to access OpenFOAM data in Python and to perform modal analysis (DMD, proper orthogonal decomposition). For this training, you need the latest development version, which can be conveniently installed via pip and git:
pip3 install git+https://github.com/FlowModelingControl/flowtorch.git@aweiner
The simulations require 40 MPI ranks and run for approximately 2 days in total. For the preliminary flow analysis and the DMD analysis in flowTorch, you can also download the following archives:
- training slides, link
- modified surfaceMountedCube tutorial (original setup)
- preliminary flow analysis, notebook
- DMD analysis with flowTorch, notebook
The current setup requires 40 MPI ranks. The suggested workflow to execute the simulations is as follows:
# we are at the top-level of this repository
mkdir run
cp -r surfaceMountedCube run/
cd run/surfaceMountedCube
./Allrun
# or with the container
./Allrun.singularity
For questions, corrections, etc. after the workshop, open a new issue in this repository.