[blaise]

Hello all,

I am facing a weird problem with mapFields when called automatically from a python script.

Setup:
OpenFOAM 2.2.x, CentOS 6.4

I am running an optimisation loop for a bulbous bow of a ship using LTSInterFoam. In each iteration of the optimisation loop the geometry is slightly changed and the simulation is then initialized with mapFields using a case with pre-calculated data. The automatization of the whole process (mesh generation, field mapping, partitioning, running the simulation, evaluating the results) works just fine and is done with python.

But:
Quite often mapFields just stops during the interpolation of the first field, without any error message. Consequently the fields are just uniformly initialized instead of having the pre-calculated data. This seems to happen randomly and does not depend on the new geometry the initial case is mapped onto. I know this because I run 12 different simulations with the same new geometry simultaneously, the difference being the draught and speed of the ship, and the problem only shows up on some simulations (sometimes none, sometimes all, mostly a few).

Here is what I do in the python script when calling mapFields:

cmd_line = 'mapFields -noFunctionObjects %s' % LOCAL_STARTDATA_SOURCE_DIR
cmd_args = shlex.split(cmd_line)
p = subprocess.Popen(cmd_args)
p.wait()

This is the content of system/ mapFieldsDict:

patchMap ( xMax xMax xMin xMin zMax zMax zMin zMin yMax yMax );
cuttingPatches ( rumpf_ship rumpf_ship yMin yMin );


Here is the output of mapFields:

Source: "/scratch_local" "startdata_01"
Target: "/scratch_local/ResMngr/ofuser/85b6ca4a-b9ba-453c-af61-5684e3085c9a_1414404186" "wulst_opt_Nsga2_1b
is8_02_des0007_01"

Create databases as time

Source time: 10000
Target time: 0
Create meshes

Source mesh size: 555227 Target mesh size: 555060

--> FOAM Warning :
From function Foam::List<Foam::tetIndices> Foam:olyMeshTetDecomposition::faceTetIndices(con st polyMes
h&, label, label)
in file meshes/polyMesh/polyMeshTetDecomposition/polyMeshTetDecomposition.C at line 570
No base point

... a bunch of similar warnings...
The following lines seem to be specific to the failure of mapFields, as these warning do not appear among the warnings mapFields issues when running correctly:

--> FOAM Warning :
From functi in "/usr/local/OpenFOAM/OpenFOAM-2.2.x/platforms/linux64Gcc47DPOpt/bin/mapFields"
#6
in "/usr/local/OpenFOAM/OpenFOAM-2.2.x/platforms/linux64Gcc47DPOpt/bin/mapFields"
#7
in "/usr/local/OpenFOAM/OpenFOAM-2.2.x/platforms/linux64Gcc47DPOpt/bin/mapFields"
#8
in "/usr/local/OpenFOAM/OpenFOAM-2.2.x/platforms/linux64Gcc47DPOpt/bin/mapFields"
#9
in "/usr/local/OpenFOAM/OpenFOAM-2.2.x/platforms/linux64Gcc47DPOpt/bin/mapFields"
#10 __libc_start_main in "/lib64/libc.so.6"
#11
in "/usr/local/OpenFOAM/OpenFOAM-2.2.x/platforms/linux64Gcc47DPOpt/bin/mapFields"
on Foam::List<Foam::tetIndices> Foam:olyMeshTetDecomposition::faceTetIndices(con st polyMesh&, label, labe
l)
...
--> FOAM Warning :
From funct#0 Foam::error:rintStack(Foam::Ostream&) in "/usr/local/OpenFOAM/OpenFOAM-2.2.x/platforms/
linux64Gcc47DPOpt/lib/libOpenFOAM.so"
#1 Foam::sigFpe::sigHandler(int) in "/usr/local/OpenFOAM/OpenFOAM-2.2.x/platforms/linux64Gcc47DPOpt/lib/li
bOpenFOAM.so"
#2
at sigaction.c:0
#3 Foam::meshToMesh::calculateInverseDistanceWeights( ) const in "/usr/local/OpenFOAM/OpenFOAM-2.2.x/platfo
rms/linux64Gcc47DPOpt/lib/libsampling.so"
#4 Foam::meshToMesh::inverseDistanceWeights() const in "/usr/local/OpenFOAM/OpenFOAM-2.2.x/platforms/linux
64Gcc47DPOpt/lib/libsampling.so"
#5
ion Foam::List<Foam::tetIndices> Foam:olyMeshTetDecomposition::faceTetIndices(con st polyMesh&, label, lab
el)
in file meshes/polyMesh/polyMeshTetDecomposition/polyMeshTetDecomposition.C at line 570
No base point for face 50858, 4(13262 16868 17830 17196), produces a valid tet decomposition.
--> FOAM Warning :
...

Mapping fields for time 10000

interpolating p_rghResidual
/*---------------------------------------------------------------------------*\
| ========= | |
| \\ / F ield | OpenFOAM: The Open Source CFD Toolbox |
| \\ / O peration | Version: 2.2.x |
| \\ / A nd | Web: www.OpenFOAM.org |
| \\/ M anipulation | |
\*---------------------------------------------------------------------------*/
Build : 2.2.x-68d8dc9f4acc
Exec : renumberMesh -overwrite -latestTime

Here the python script starts renumberMesh while mapFields is not ready yet (or maybe it is ready but is giving up for some occult reason). And as the python code snippets show, the subprocess is not called in the background.

Can someone tell me why on earth this happens??? This never occurs when I start the mapping of fields manually from a python shell (for testing purposes).

Any idea is appreciated!

Thank you!