sponge

Flow past a sponge, a porous cylindrical lattice (parameters in geom/README.md), simulated with Basilisk. The sponge geometry is an STL file converted to a Basilisk dump, which the solver cylinder/cylinder.c restores with -d. The solver also has analytic shapes (-S cylinder, -S sphere) used for validation.

real.stl: the scanned sponge, centred and scaled to unit diameter by stl/center.py.

Files:

• geom/gen.py: sponge parameters (config.py) to ver.stl
• stl/center.py: center the STL, scale the diameter to 1, axis along z
• stl/size.py: print z extent of an STL
• dump/stl2dump.c: STL to Basilisk dump, -w z LEVEL Z0 Z1 adds walls at the sponge ends
• dump/dump2xdmf.c: dump to XDMF for ParaView
• cylinder/cylinder.c: solver, cylinder -h lists options
• cylinder/deploy/cylinder.c: preprocessed solver source, needs only mpicc
• geom/0pre.sh, geom/0run.sh: SLURM scripts used for production runs

Install

Requires MPI, GNU make, Python 3 with numpy and scipy.

wget -q http://basilisk.fr/basilisk/basilisk.tar.gz
tar zxf basilisk.tar.gz
cd basilisk/src
cp config.gcc config    # macOS: cp config.osx config
(MAKEFLAGS=-j`nproc`; make ast && make qcc)
mkdir -p "$HOME/.local/bin" && cp qcc "$HOME/.local/bin/"

The path to basilisk/src is compiled into qcc, do not move the directory after building.

(cd cylinder && make cylinder)
(cd dump && make stl2dump dump2xdmf)

On macOS use make cylinder QCCFLAGS=-D_DARWIN_C_SOURCE and make stl2dump dump2xdmf CC=gcc-15 (Homebrew gcc, for OpenMP).

Run

mkdir sponge0 && cd sponge0
cat > config.py <<'EOF'
NV = 20
NC = 30
AA = 0.25
BB = 0.5
LOOP_NO = 1
R_E = 1.2
NumBelix = 1
NumBelix2 = 0
EOF
PYTHONPATH=. python3 ../geom/gen.py
../stl/center.py -v ver.stl center.stl
set -- $(../stl/size.py center.stl)
../dump/stl2dump -v -o -w z 6 $1 $2 -- -5 -6.25 -6.25 12.5  4 7  4 center.stl basilisk.dump
mpiexec -n 4 ../cylinder/cylinder -v -F -r 100 -p 100 -e 200 -Z $2 \
        -f force.dat -d basilisk.dump -o h -b pp

stl2dump arguments: X0 Y0 Z0 L minlevel maxlevel npe file.stl out.dump. The mesh is fixed by stl2dump and does not change during the run: minlevel everywhere (with -o the last 10% of the domain in x is coarser), maxlevel at the sponge surface, LEVEL at the walls. The wake is resolved at minlevel. Production runs use levels 10 13, see geom/0pre.sh.

Output: force.dat (columns: step, t, dt, total force, pressure force, viscous force, three components each), h.N.xdmf2 full field (with -F), h.y.N.xdmf2 and h.z.N.xdmf2 slices for ParaView, h.N.dump restart files (restart with -d h.N.dump -i). Cell attributes in h.N.attr.raw: p, lambda2, Q, u (3), omega (3); the slices also have cs and phi after Q.

Render

Iso-surfaces of |omega| (or the Q criterion with qVALUE, lambda2 with lVALUE) from the full-field output (-F), extracted with amriso (pip install git+https://github.com/cselab/amriso) and rendered with ParaView. The sponge surface comes from the STL. -Z ZCUT drops cells with |z| > ZCUT, to hide the boundary layers on the end walls (the sponge above spans |z| < 2.06).

python3 ../tool/iso.py -Z 1.6 1 h.[0-9]*.xdmf2
pvpython ../tool/pvsurf.py 1 center.stl h.[0-9]*.xdmf2
ffmpeg -framerate 4 -pattern_type glob -i 'h.*.omega1.png' -pix_fmt yuv420p omega1.mp4

tool/iso.py writes h.N.omega1.xdmf2 triangle meshes colored by omega_z, tool/pvsurf.py writes h.N.omega1.png with a fixed camera and color range over all frames. tool/pvmulti.py draws several levels of one frame with their own opacity and color (c colored by omega_z, b pale blue, p pink, or R/G/B), for example two Q levels:

python3 ../tool/iso.py -Z 1.6 q1,q5 h.[0-9]*.xdmf2
pvpython ../tool/pvmulti.py center.stl q1:0.3:b,q5:1.0:p h.[0-9]*.xdmf2