[ELwardi]

Hi all,

I'm using (trying to use ) foam-extend's coupledFvMatrix (https://github.com/Unofficial-Extend...vBlockMatrix.H)

to couple three equations. The coupling terms consist of "laplacians" and Time derivatives .

This thread explains how I'm approaching this and what the current results are. I know I'm doing something (Or more) wrong since these results are not sane enough .

First, construct a volVectorField to hold solved (scalar) Fields:

volVector3Field pSwSg
(
    IOobject
    (   
        "pSwSg",
        runTime.timeName(),
        mesh,
        IOobject::NO_READ,
        IOobject::AUTO_WRITE
    ),  
    mesh,
    dimensionedVector3("zero", dimless, vector3::zero)
);

Later in the code, define each fvScalarMatrix; for example:

// "Main" pressure equation

fvScalarMatrix pEqn
(

     fvm::ddt(M, p) + fvm::laplacian(-D, p) + fvc::div(somePhi)

);


// Coupled term between pEqn and some other SwEqn
fvScalarMatrix SwInp
(   
    - fvm::ddt(M1, Sw) // and maybe fvm::laplacian(D1, Sw)
);

The actual coupling is carried out when creating the block matrix:

// Instantiate the coubled matrix

fvBlockMatrix<vector3> resEqn(pSwSg);



// Add equations
resEqn.insertEquation(0, pEqn);
resEqn.insertEquation(1, SwEqn);
resEqn.insertEquation(2, SgEqn);


// Insert coupling terms
// Sw term in pEqn:
 resEqn.insertEquationCoupling(0, 1, SwInp);
// Note that SwInp is a fvScalarMatrix

 // Add more couplings ...

After this, we solve the blockMatrix and retrieve solutions.


The problem is, inserting the following code:

Info<< "Lower Coupling:" <<
    << resEqn.coupleLower() << nl
    << "Upper Coupling:" << 
    << resEqn.coupleUpper() << nl
    << "Component Coupled:" << 
    << resEqn.componentCoupled() << nl
    << "Coupled Diagonal:" << nl
    << resEqn.diag() << endl;

somewhere before solving the equation, and testing it on a minimal case (1D 5cells mesh), results in the following output:

Lower Coupling:
3
(
unallocated
1
unallocated
1
unallocated
0
)

Upper Coupling:
3
(
unallocated
1
unallocated
1
unallocated
0
)

Component Coupled:
1



Coupled Diagonal:
square
5
(
(-865.137 -7.42132e-07 -7.42132e-07 21.9694 7.42132e-06 0 -129771 -0.00011132 -0.000103899) 

(-14549.7 -1.51723e-13 -1.51723e-13 368.569 1.51723e-12 0 -2.18245e+06 -2.27584e-11 -2.12412e-11) 

(-302060 -4.08706e-29 -4.08706e-29 7645.23 4.08706e-28 0 -4.5309e+07 -6.13059e-27 -5.72188e-27) 

(-1.3066e+07 -8.3987e-74 -8.3987e-74 330590 8.3987e-73 0 -1.95991e+09 -1.2598e-71 -1.17582e-71) 

(-1.27778e+07 -3.88293e-254 -3.88293e-254 323294 3.88293e-253 0 -1.91666e+09 -5.8244e-252 -5.4361e-252)
)



GMRES:  Solving for pSwSg, Initial residual = (0.006936 0.000167427 0.992364), Final residual = (2.11584e-12 7.29283e-15 1.42317e-14), No Iterations 1

Huh, no upper/lower coupling .


Any help or orientation to some reference materials is greatly appreciated.
Best regards,

[hjasak]

Are you looking at the right place? Look at the matrix: there is diag, upper and lower; then there is out-of-core coefficient like processor boundaries.

From your output, you MUST get block clefs for upper and lower; otherwise GMT’s would not be called. I think you are looking at the wrong place.

I also have doubts about your cross-mciupling matrix, but that’s another thing...

[ELwardi]

Mr @hjasak, thank you for your reply.


As you pointed out, the coupled matrix seems to be OK unless I'm making a vague, invalid assumption somewhere (detailed log attached).


I need to be 100% sure of something though; Does

fvBlockMatrix::insertEquationCoupling(id1, id2, 2In1)

insert the matrix 2In1 as coupling term of variable of equation 2 in equation1 ?

(following the same convention of insertBlockCoupling used in pUCoupledFoam) Or is it the other way around.