| FlyBob91 |
December 3, 2016 10:16 |
Adjoint Convergence Problem
1 Attachment(s)
Hi to all,
i wrote my own adjoint solver and compiled it correcty. However i'm meeting some problem running my case because the residuals seem to converge but the solution literally explodes as you can see below
Code:
smoothSolver: Solving for Uax, Initial residual = 4.39575174596e-05, Final residual = 4.16030919667e-08, No Iterations 2
smoothSolver: Solving for Uay, Initial residual = 4.5057555084e-05, Final residual = 3.58842322767e-08, No Iterations 2
smoothSolver: Solving for Uaz, Initial residual = 4.06676312041e-05, Final residual = 3.22296628924e-08, No Iterations 2
GAMG: Solving for pa, Initial residual = 0.000212921685966, Final residual = 1.41332460912e-06, No Iterations 4
Adjoint continuity errors : sum local = 15942358.9466, global = 8202.0330787, cumulative = 9256411904.89
I try to use low relaxation factors but nothing change. I link my fvschemes and fvsolution files.
I also noticed others people have had a problem similar to mine, because the adjoint problem are very sensitive to the boundary conditions.
Code:
solvers
{
"(p|pa)"
{
solver GAMG;
tolerance 1e-08;
relTol 0.01;
smoother GaussSeidel;
nPreSweeps 0;
nPostSweeps 2;
cacheAgglomeration true;
nCellsInCoarsestLevel 10;
agglomerator faceAreaPair;
mergeLevels 1;
}
"(U|Ua|k|epsilon)"
{
solver smoothSolver;
smoother GaussSeidel;
nSweeps 2;
tolerance 1e-08;
relTol 0.1;
}
}
SIMPLE
{
nNonOrthogonalCorrectors 0;
pRefCell 0;
pRefValue 0;
}
relaxationFactors
{
fields
{
"(p|pa)" 0.1;
alpha 0.1;
}
equations
{
"(U|Ua)" 0.1;
"(k|epsilon)" 0.7;
}
}
Code:
ddtSchemes
{
default steadyState;
}
gradSchemes
{
default Gauss linear;
}
divSchemes
{
default Gauss upwind;
div(phi,U) bounded Gauss upwind;
div(phi,k) bounded Gauss upwind;
div(phi,epsilon) bounded Gauss upwind;
div((nuEff*dev(T(grad(U))))) Gauss linear;
div(-phi,Ua) bounded Gauss upwind;
div((nuEff*dev(T(grad(Ua))))) Gauss linear;
}
laplacianSchemes
{
default Gauss linear uncorrected;
}
interpolationSchemes
{
default linear;
}
snGradSchemes
{
default uncorrected;
}
fluxRequired
{
default no;
p;
pa;
}
Furthermore i read this thread
HTML Code:
http://www.cfd-online.com/Forums/openfoam-solving/103229-adjointshapeoptimizationfoam-checkerboarding-inlet.html
where sylvester suggest to use a " pseudo staggered grid approach" but i don't uderstand what he refers to.
The simulation type is a laminar and if you are interested in my code you can find it in a previuos thread
HTML Code:
http://www.cfd-online.com/Forums/openfoam-programming-development/180587-insert-dimensioned-salar.html
i link also my mesh quality
Code:
Create polyMesh for time = 0
Time = 0
Mesh stats
points: 306516
faces: 863850
internal faces: 810150
cells: 279000
faces per cell: 6
boundary patches: 4
point zones: 0
face zones: 0
cell zones: 0
Overall number of cells of each type:
hexahedra: 279000
prisms: 0
wedges: 0
pyramids: 0
tet wedges: 0
tetrahedra: 0
polyhedra: 0
Checking topology...
Boundary definition OK.
Cell to face addressing OK.
Point usage OK.
Upper triangular ordering OK.
Face vertices OK.
Number of regions: 1 (OK).
Checking patch topology for multiply connected surfaces...
Patch Faces Points Surface topology
walls 52500 52619 ok (non-closed singly connected)
inlet1 300 336 ok (non-closed singly connected)
inlet2 300 336 ok (non-closed singly connected)
outlet 600 651 ok (non-closed singly connected)
Checking geometry...
Overall domain bounding box (-0.00015 -0.00165 0) (0.003 0.00165 0.0002)
Mesh (non-empty, non-wedge) directions (1 1 1)
Mesh (non-empty) directions (1 1 1)
Boundary openness (-6.91724010847e-18 1.96638314844e-18 2.57097977866e-17) OK.
Max cell openness = 1.29246970711e-16 OK.
Max aspect ratio = 1 OK.
Minimum face area = 1e-10. Maximum face area = 1e-10. Face area magnitudes OK.
Min volume = 1e-15. Max volume = 1e-15. Total volume = 2.79000000001e-10. Cell volumes OK.
Mesh non-orthogonality Max: 0 average: 0
Non-orthogonality check OK.
Face pyramids OK.
Max skewness = 1.25 OK.
Coupled point location match (average 0) OK.
Mesh OK.
Any idea to solve the problem?
Thanks for help,
Roberto |
