[kar]

Hello,

I would be happy to read someone's experience in thermo-convection modeling - I solved some cases, but when Ra >= 26000 for air, U matrix solving takes 1000 iterations at some moment (some max 10 otherwise). In the next step Co is very big, so solver crashes with floating point exception. Note that in previous iteration there seams to be no problem.

this is a piece of solver output:

Courant Number mean: 0.14292815354 max: 0.79327506765
DILUPBiCG: Solving for Ux, Initial residual = 7.6037581268e-05, Final residual = 5.3528418586e-09, No Iterations 4
DILUPBiCG: Solving for Uy, Initial residual = 1.7520153951e-05, Final residual = 118.26880138, No Iterations 1001

UEqn = new fvVectorMatrix
(
fvm::ddt(U)
+ fvm::div(phi, U)
- fvm::laplacian(nu, U)
);

solve
(
*UEqn == -fvc::grad(p)/rho0 + alpha*(T - T0)*g
);

Why matrix solving could be a problem?

Kārlis