Hi Jyothis,

What are the mesh quality in each case. Because at the end of the day these effects convergence; during flux calculations. 

Is there any negative volume cells present in the fine mesh, and also if you are using single precision with fine mesh can yield such issues.

Is your mesh independent of mesh ?

Regards,

Rajesh.

With a finer mesh, if you perform a transient simulation, you should reduce the timestep otherwise the results can locally explode causing divergence of the solution.

Franco

Maybe you should change time step. be sure about CFL condition. more info @ attached link.

http://en.wikipedia.org/wiki/Courant%E2%80%93Friedrichs%E2%80%93Lewy_condition

Hi Jyothis

Very often, natural and certainly mixed convection poses a problem with convergence due to the fact, that the flow pattern in natural convection is not stable in time and very often such cases must be solved as transient. But as a first remedy, as said above, keep the Courant number, i.e. time step as low as possible (it is linked with cell dimensions). You can check the force of natural convection by simply calculating ratio of Gr number to second power of Re number and according to this ratio or = 1 you get an information whether the natural convection is strong enough for the problem to be approached as transient. Do play a little. :-)

If you try unsteady problem, as above answered, your problem is related with CFL number.

Generally, there is some optimal value of Courant number that balances the total number of iterations against the time per iteration. This optimum is a function of many factors: whether the flow is compressible, how close the solution is to convergence, mesh quality and the magnitude of the body forces.

For your convinence, I would like to recommend that you consider using ramping to increase the CFL automatically to its final value during the early iterations. Additionally,  if you use a commercial codes, you can use  automatic CFD control option, which automatically modifies the CFL (Courant) number each iteration to ensure that the AMG linear solver converges. 

If you are woking on steady state case, and you are not able to get converged solution, when mesh is getting denser, this may be because that your mesh system becomes very sensitive to geometrical complexity( ex, strong curvature...). In this case, you should improve mesh quality and use strong under-relaxation factors.. in pressure and momentum equations...

Good luck~!  

Thank you all for your Valuable Replies