BUG: applyBoundaryLayer - inconsistent at processor boundaries and refinement regions

The transition cells for refinement and processor boundaries seem to have inconsistnet results when using applyBoundaryLayer utility.

Attached example and obtained result.

@andy @Mattijs

pitzDaily_appyBoundaryLayer_issue.tgz

Hi, The inconsistency on the proc patches is created by the reconstruction. If you check k, nut, etc on the individuals processors the fields are fine. As a preprocessor tool to start a calculation the fields on each processor are OK to start the simulation in parallel.

The inconsistency still remains even at cell refinements. Attached modified case and example result. pitzDaily_appyBoundaryLayer_issue-refinementCells.tgz

Hi, I came up with this error (exchange ID 110) and I don't see the fields are fine on each processor. First when I load the decomposed case (done in EnSight, please excuse my lack of experience with paraView) I clearly see the peaks for the turbulent quantities in the cells next to processor boundaries. Second I see these peaks are getting smeared over the iterations, so they must be contained in the initialized fields. The attached pictures show the initialization and the first iteration. The decomposed fields were loaded and no reconstruction was involved. The case Prashant has provided was used with simpleFoam and kOmegaSST model.

Hi, I pushed some changes to applyBoundaryLayer. Now the turbulent is corrected after being constructed. The refinement pattern jump is improved using a different grad(U) schemes such as leastSquares. As well, in order to have a better potentialFoam U field use Gauss linear uncorrected for the laplacian. With these changes the value jump on processor patches is small although is still there. Sergio

mentioned in commit b529cd7a

mentioned in commit 92ae4fbe

The refinement pattern issue is attributed to potentialFoam, and more sensitive to the schemes used. Attached a simple case (its just refinement boxes!) which replicates the issue and also suggests the scheme suitable for such refinement pattern jumps. potentialFoam_issue_at_refinementTransistion.tgz

Status changed to closed

General Note:

To help users from OF231 version, the bugfix for plus version from commit 92ae4fbe is backported.

The following attachment works with OF231. It creates a new executable (which can happily sit together with original utility)

applyBoundaryLayer-231_11032016.tgz

Best Regards, Prashant