foam-extend: different empty (2-D) directions
 

Does anybody have any idea how to deal with the following error?



I encounter the error when running foam-extend 4.1 in parallel "Some processors detect different empty (2-D) directions. Probably using empty patches on a bad parallel decomposition".



My empty direction is along x-axis. I tried different decomposing methods in decomposeParDict file and the domain is successfully decomposed as expected. , but it keeps giving the same error. The error pops up at the first time step most of the time, sometime after some iterations. By the way, I am using sixDoFRigidBodyDisplacement boundary condition (0/pointDisplacement) and the solver is interDyMFoam.



The case runs well in serial mode.



decomposeParDict:

numberOfSubdomains 2;
method simple;
//method scotch; //same error
//method metis; //same error
simpleCoeffs
{
n ( 1 2 1 );
delta 0.001;
}

Dear Michael@UW
I have the same problem with cyclic boundary condition and dynamic mesh (fsiFoam). Do you resolve it?

No. I have not solved this issue. But I found the inconsistent empty direction is caused by the dynamic mesh motion; there is an unexpected rotation to make the mesh of the empty patch not on the same plane. I constrain the object but it still produces some motion along the direction which is not supposed to have.

As your tip, I separate cyclic boundary condition to multi part and use "preservePatch" in decomposeDict to add each part to a processor. I think it works.

That's a good way! I may try that later. Recently I learned there is a tool name flattenMesh which can make the patch flat, but I haven’t tried it.

I also get the same error in FE4.1. In the error description there is a mention of increasing 'emptyDirectionTolerance' in controlDict, which I couldn't make it work.
What I did to solve the issue was to reduce my number of processors in decomposeParDict by one. For example changing 6 to 5. I think dividing a number to 5 produces a rounder number so round-off error is increased and more manageable.