[C_J]

Hi!

Parallel processing worked fine before I implemented a boundary condition that changed with every time step. The boundary is defined as a fixed value (the error occurs with the groovy boundary too) and the boundary is changed using "variable.boundaryField()[patchID]==newvalue;
variable.correctBoundaryConditions();"

When I try to run parallel processing with changing boundary I get a "possible face ordering problem" after running three time steps.

I suspect the issue is in the dynamic mesh (a spherical mesh that expands with time). The type of dynamic mesh selected is the dynamicMotionSolverFvmesh, solver is velocityLaplacian. The mesh is expanded based on a value calculated from the solver. Could someone please provide a solution?

BTW the removeSlivers is set to "true" and edgeRefinement to "false" in the dynamicMeshDict, if that has anything to do with this error. Not sure where to look. Help please!

[simrego]

Hi!


Have you tried the coded BC? It should work with dynamicRefinement too which is a more complex "motion" solver than a laplacian solver.
There are plenty of examples about how to use a coded BC around the web.
(NOT A NEW BC, but a coded BC, they are not exactly the same!!! The result is the same but i like coded BC more since you just modify the code in the BC and it'll recompile it in the next run automatically.)

[C_J]

Originally Posted by simrego

Hi!


Have you tried the coded BC? It should work with dynamicRefinement too which is a more complex "motion" solver than a laplacian solver.
There are plenty of examples about how to use a coded BC around the web.
(NOT A NEW BC, but a coded BC, they are not exactly the same!!! The result is the same but i like coded BC more since you just modify the code in the BC and it'll recompile it in the next run automatically.)

Thank you for your answer. I'm using Foam-extended to make use of the groovy boundary, as I have to expand the mesh using a realtime value calculated by the solver in each time-step. Parallel processing worked well when an average constant value was used for the boundary condition. I tried using the coded boundary condition to change the boundary condition in each time-step on the Foam-extended version, but it seem to be implementable only in openfoam-dev. Could you suggest a method to implement a changing boundary condition that will enable parallel processing in the Foam-extended version?



Thank you very much in advance!