Background: the lid driven flow demo was the oldest in Ocellaris and the CI had not caught that it no-longer ran correctly. It is now tested in CI (running the first time step only).
The demo now works as intended, and the converged results (after 40 minutes running on 1CPU on my laptop) look close to Ghia at al. This was tested in the Ocellaris container based on the FEniCS Docker container.
Issue #9 was reported on Debian. I tested running the first couple of time steps on Debian Buster, with a FEniCS install from Debian repositories. The tested configuration was the following Docker container:
RUN apt-get -qq update && \
apt-get -y --with-new-pkgs \
-o Dpkg::Options::="--force-confold" upgrade && \
apt-get -y install python3-lxml fenics python3-h5py python3-pip git && \
apt-get clean && \
cd && \
rm -rf /var/lib/apt/lists/* /tmp/* /var/tmp/*
RUN useradd --create-home --shell /bin/bash debianuser
I did not run the whole simulation in this configuration, but got no errors.
Some images that may one day be added to a write up about the lid driven cavity demo demo in the documentation:
Paraview, mesh grading towards the upper right corner:
Ocellaris Inspector GUI, horisontal velocity compared to Ghia et al:
Ocellaris Inspector GUI, vertical velocity compared to Ghia et al: