hi, I have a model to be stimulated in fluent. i have successfully export it to fluent from gambit and grid check is alright too.However, when i initialize and iterate, the result seems to be several residual ( i am using k-epsilon) vibrating up and down in an amplitud (seemingly constant) even till a very large maximum iteration per-time step. It cannot converge even i input 10 000 for maximum iteration per-time step.However, it does not diverge too, can i accept such model and what are the assumption ?

I do not use Fluent and for that matter any other pack. I do not know what is meant by grid check.

However, depending on the boundary conditions (i.e., the flow situation being modelled), such non-convergence may arise because of : (a) improper imposition of boundary conditions (b) under relaxation of k-eps - its parity with that used for momentum eqns - u/v/w (c) discretisation of convection terms(upwind order/type) (d) choice of solution scheme for continuity eqn (e) time step (f) flow stability

- and may be many others. If solution has not converged, if sensitivity to grid variation has not been studied, etc., how can we say results are acceptable.

Thank you for your advise. When i keep the time step size small and increase the number of time step gradually.Plus the maximum iteration per time step is reduced, the result will converge due to these changes.How would i interprete these phenomena? please advise

If you decrease time step(small time step size), diagonal dominance can be assured in your CFD matrix. So, in general, the convergence becomes better than the case of large time step. For transient calculation, flow variation at the initial stage may be very large. In such case, time step should be small for the convergence. How much small time step ? It depends on your flow situation.

Sincerely, Jinwook

'How much small time step ? It depends on your flow situation' -

True. Must take at least Courant (CFL) and Grid Fourier stability check. Start with about half of the minimum of these - should work well for 'not very complex' situations.