[siefer92]

Howdy Yall:

Still working on assembling a solver for supersonic solid-solid abrasion. Currently the lagrangian libraries are coupled with rhoCentralFoam and I have a test case that runs and is stable.

Below is a visual overview of the case:

Vel&particles.jpg

The test case is of a flow past a 2-dimensional cylindrical obstruction. pretty standard case.

I have ran some cases that include in-situ deformation of the cylindrical boundary caused by erosion but doesn't include any sort of boundary refinement. This code doesn't require a dynamicMeshDict and is governed through the solver source. The source file governing the erosion is based on Finnie's law and can be found in thesrc/lagrangian/intermediate/CloudFunctionObjects/ParticleErosion/ParticleErosion.C)

looking at the figure titled 'boundary erosion', you can see the extension of the boundary layer cells as the obstruction is eroded.

BoundaryErosion.jpg

With enough erosion and no successive boundary refinements, the solver crashes due to negative temperature in the overly deformed boundary cells. This is behavior that one would expect.

I have incorporated yPlus based boundary refinement using a dynamicMeshDict in the constant/ directory. The contents are shown below:

/*--------------------------------*- C++ -*----------------------------------*\
| =========                 |                                                 |
| \\      /  F ield         | OpenFOAM: The Open Source CFD Toolbox           |
|  \\    /   O peration     | Version:  v2012                                 |
|   \\  /    A nd           | Website:  www.openfoam.com                      |
|    \\/     M anipulation  |                                                 |
\*---------------------------------------------------------------------------*/
FoamFile
{
    version     2.0;
    format      ascii;
    class       dictionary;
    location    "constant";
    object      dynamicMeshDict;
}
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //



dynamicFvMesh dynamicRefineFvMesh;
   dynamicRefineFvMeshCoeffs
   {
       refineInterval 500;
       field yPlus;
       lowerRefineLevel   350;
       upperRefineLevel   1000;
       unrefineLevel 100; 
       nBufferLayers 3;
       maxRefinement 10;
       maxCells 10000000;
       correctFluxes 	(
			( phi none )
                        ( neg none )
                        ( pos none )
			( dustCloudQ none)

			);
       dumpLevel true;
   }

The boundary refinement is occurring as expected without error and in the locations one would expect. The interpolation of the erosion field dustCloudQ, however, is not functioning. The values shoot up from initial values in the e-8 magnitude to values in e+8 magnitude.
Here is the Erosion volume removal at a timestep where it is functioning
Q_functional_at_wall.jpg

Below is an image when it is not functioning.
Q_NON_functional_at_wall.JPG



Additionally the location where the volScalarField, dustCloudQ, increases in value is in an area between cell sizes on the boundary surface.

Does anybody have any thoughts as why this is happening?

My thought is that it has something to do with correctFluxes in the dynamicMeshDict and how the field is mapped between meshes.

I'd appreciate yalls thoughts on the matter.

[siefer92]

As an update I am looking into mapFields at the moment.

Long story short, the volScalarField dustCloudQ is a volume. I need the dustCloudQ field to be mapped in a conservative manner.

E.g. a cell of .1m^3 eroded volume is refined into 4 cells based on the yPlus criteria near the boundary. The eroded volume of each of the refined subcells is .025.

This is speculation ATM I am running a case with a I/O time at every refinement step to track the growth of the Eroded volume. If the growth is fast & consistent with each refine/unrefine then I am on the correct track I think.

if, in one I/O, it explodes to e36 then this is a different problem. Will post thoughts soon