Wall Function Issue

[kingdomkarl]

Hey guys,

congratulations on the great new release! I’ve encountered an issue when using the new wall functions in an internal flow geometry. My prepareLattice function looks like:

void prepareLattice( SuperLattice<T, DESCRIPTOR>& sLattice,
UnitConverter<T,DESCRIPTOR> const& converter,
STLreader<T>& stlReader,
SuperGeometry<T,3>& superGeometry)
{
WallModelParameters<T> wallModelParameters;
wallModelParameters.bodyForce = false;
wallModelParameters.rhoMethod = 0;
wallModelParameters.fNeqMethod = 0;
wallModelParameters.samplingCellDistance = 3.5;
wallModelParameters.interpolateSampleVelocity = true;
wallModelParameters.useVanDriest = false;
wallModelParameters.wallFunctionProfile = 1;
wallModelParameters.movingWall = false;
wallModelParameters.averageVelocity = false;

OstreamManager clout( std::cout,”prepareLattice” );
clout << “Prepare Lattice …” << std::endl;
const T omega = converter.getLatticeRelaxationFrequency();

// material = 1: bulk dynamics
if ( useWallFunctions ) {
setTurbulentWallModelDynamics(sLattice, superGeometry.getMaterialIndicator({1}), wallModelParameters);
} else {
sLattice.defineDynamics<BulkDynamics>(superGeometry, 1);
}

// material = 2: walls
setBouzidiBoundary(sLattice, superGeometry, 2, stlReader);
if ( useWallFunctions ) {
setTurbulentWallModel(sLattice, superGeometry, 2, wallModelParameters);
}

// material = 3: inlet
if ( useWallFunctions ) {
boundary::set<boundary::InterpolatedVelocity>(sLattice, superGeometry, 3);
} else {
setBouzidiBoundary<T,DESCRIPTOR,BouzidiVelocityPostProcessor>(sLattice, superGeometry, 3, stlReader);
}

// material = 4: outlet
boundary::set<boundary::InterpolatedPressure>(sLattice, superGeometry, 4);
if ( !useWallFunctions ) {
sLattice.defineDynamics<BulkDynamics>(superGeometry.getMaterialIndicator(4));
}

// Initial conditions
AnalyticalConst3D<T,T> rhoF( 1 );
std::vector<T> velocity( 3, T(0) );
AnalyticalConst3D<T,T> uF( velocity );

// Initialize all values of distribution functions to their local equilibrium
sLattice.defineRhoU( superGeometry.getMaterialIndicator({1, 2, 3, 4}), rhoF, uF );
sLattice.iniEquilibrium( superGeometry.getMaterialIndicator({1, 2, 3, 4}), rhoF, uF );

AnalyticalConst3D<T, T> u0(0.,0.,0.);
sLattice.defineField<descriptors::VELOCITY>(superGeometry.getMaterialIndicator({0,1,2,3,4}), u0);

sLattice.setParameter<descriptors::OMEGA>(omega);
sLattice.setParameter<collision::LES::SMAGORINSKY>(T(0.15));

// Lattice initialize
sLattice.initialize();
clout << “Prepare Lattice … OK” << std::endl;
}

The behavior is the following:
– Without WF everything works as intended
– When I switch on the WF and use only one cuboid everything works as intended
– As soon as I increase the number of cuboids I get the following error:

terminate called after throwing an instance of ‘std::runtime_error’
what(): an illegal memory access was encountered
Aborted (core dumped)

I traced it back to the sLattice.initialize function but am very much lost now as I am not super experienced in openLB. Any help here would be greatly appreciated.

I run the case on a single GPU using the gpu_only.mk and tried CUDA 12.4 and 12.6.

Thanks a lot for any help on this.

[FBukreev]

Hi,

you can check whether your overlap is bigger then the sample distance in WM and you can give stlReader into setTurbulentWallModel. Also please add POROSITY field into descriptor and define it to 1 in the bulk and 0 at the boundaries.

[kingdomkarl]

Hey,

thank you very much for the quick answer. With your suggestion I managed to get my case running. What helped:
– The WFs work fine with multiple cuboids as long as they cut the walls perpendicularly
– The WFs do not seem to work when they cut a wall at a non-perpendicular angle. This can be remedied by setting the samplingCellDistance to a small value or disabling it. Then no sampling outside a particular cuboid is done. I am not sure which impact this has on the performance of the WFs.

Feel free to let me know if I got something wrong.

[FBukreev]

Hi,

I am happy that it runs!
I think if you make the size of overlap larger then probably WF will also work on non-perpendicular angle cuts. In the not very far future we plan to improve normal and distance calculation at the wall, so that hopefully the WF at all complex geometries will be more stable.

[Author]

Posts