[al_th]

Hi everyone,

[OF 2.3]

I am encountering a weird (but maybe intentional) behaviour using mesquite as a motion solver for a dynamic mesh problem.

I am simulating a cylinder where one of the base is moving in and out. At first, I was using the velocityLaplacian solver but due to some of its behaviour on cyclic patch, I had to find an alternative (basically it replaces the pointMotionU at each time steps and force cyclic patch to be cyclic even if I do not want the patches to move).

Anyway, I finally decided to use the mesquiteMotionSolver that is in the dynamicTopoFvMesh library of Sandeep. I use the 2.3 port of the library, compiled it with no apparent problem, and I am able to use a prescribed motion on my boundary. Here is my dynamicMeshDict :

/*--------------------------------*- C++ -*----------------------------------*\
| =========                 |                                                 |
| \\      /  F ield         | OpenFOAM: The Open Source CFD Toolbox           |
|  \\    /   O peration     | Version:  2.3.0                                 |
|   \\  /    A nd           | Web:      www.OpenFOAM.org                      |
|    \\/     M anipulation  |                                                 |
\*---------------------------------------------------------------------------*/
FoamFile
{
    version     2.0;
    format      ascii;
    class       dictionary;
    location    "constant";
    object      dynamicMeshDict;
}
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //


dynamicFvMesh dynamicMotionSolverFvMesh;


motionSolverLibs ("libmesquiteMotionSolver.so");
solver mesquiteMotionSolver;

mesquiteOptions
{
    surfInterval 1;
    optMetric MeanRatio;
    objFunction LPtoP;
    optAlgorithm            TrustRegion;

    tcInner
    {
        absGradL2            1e-5;
        cpuTime              0.05;
    }

    //- Power value for the LPtoP objective function
    pValue                  2;
    power                   2;

    //- Specify a tolerance for the surface-smoothing CG solver
    tolerance               1e-2;

    //- Specify number of CG sweeps for surface-smoothing
    nSweeps                 2;


	fixedValuePatches{
		cylinderBasis
		{
	           type           	timeVaryingUniformFixedValue;
	           value 			uniform (0 0 0);
	           outOfBounds 	repeat;
	           fileName 		"$FOAM_CASE/values";
   		}
	}
	
}

And the file "values" :

(
(0 (0 0 0))
(0.001 (0 0 0))
(0.002 (0.0000104716405696524 0 0))
(0.003 (0.0000104710663911498 0 0))
(0.004 (0.0000104702051430726 0 0))
(0.005 (0.0000104690568490324 0 0))
(... (... ... ...))
(0.996000000000001 (-0.0000000553892191113015 0 0))
(0.997000000000001 (-0.0000000453188323409903 0 0))
(0.998000000000001 (-0.000000035248217367459 0 0))
(0.999000000000001 (-0.000000025177424900838 0 0))
)

My problem is the following : If I change the deltaT value in my controlDict file, then the motion that is prescribed is not good anymore (the values are wrong, by a coefficient factor equal to the one applied on the deltaT value). Is it a normal behaviour ? If not, does anyone have a clue of what could be wrong ? I suspect the "boundaryConditions_().correctBoundaryConditions() ;" correct the BC wrongly in "mesquiteMotionSolver.C" but I am farm from being a pro of OpenFoam.

I would be glad to find a solution to this because it would allow me to use dynamic deltaT instead of a fixed one. My simulation does have two different states, one where the motion of the wall is really slow, therefore the courant number drops drastically and I could save a lot of time by increasing the deltaT during those states.

[al_th]

Anyone please ?

[Zane]

The prescribed motion of a patch is relatived to deltaT.

Because the value of pointPatchField, which you set in timeVaryingUniformFixedValue, is the relative displacement, not the absolute position of every mesh point.

That is, the data you set in "values" file means the displacement vector of every mesh point at every specific time step.