for buoyant turbulent flows which model is better ? k-omega, RSM are not converging even after 30000 iteration minimum face area (m2): 1.000000e-004 maximum face area (m2): 2.500730e-004 is this fine ??

one more question as i am using boundary layer mesh cell size will depend on 1st cell and b/a ratio, so what can be possible size of first cell taking decision based on no of cells and change in Nu no. may not be correct because of no of bl steps (am using ratio 1.2)

is there any way by which i can say grid is fine based on results of simulations ? i am getting max Y+ value 0.14 (RSM MODEL)

thanks in advance

[Micael]

Hi Sudhir,

I am just experienced with k-epsilon yet.

Here are some strategies I successfully used to converge turbulent flow in mixed convection.

- first, converge without gravity
- use "coupled" for pressure-velocity coupling
- use "PRESTO" for pressure discretization (second order upwind for others)
- reduce under-relaxation factor, slowly and by increment between some hundred iterations, especially turbulence properties (like k and epsilon)
- reduce momentum explicit relaxation factors of the coupled algoritm

Before reducing any relaxation factor, wait for stagnation during iterations. Reducing those factors also reduce convergence rate, so it is better to apply it only when there are stagnation or instability in residuals.

Mesh quality is very important. y+ = 0.14 is very fine. However, free convection is challenging the solver everywhere in the domain, not just at wall.

I hope this will help you.

Micael

[Micael]

I just think that if your problem is purely free convection, then is it not usefull to solve without gravity, since that will produce no flow at all. In that case, you can start with a low gravity value (like 0.98 instead of 9.8).