Hi all

I've just started using FLUENT, and i am trying to model flow separations over an inverted airfoil in extreme ground effect. I ran some cases in with the steady state solvers and it just would not converge (at high angles of incidence). Thinking that it might be due to the large flow separation and wake formation that the steady state equations break down, i modelled it with an unsteady solver and managed to acheive convergence to the order of 10-5. Question is:-

1) is it right to do so, to use the unsteady solver 2) is it a common practise to grid adaptions (Y+) using unsteady solvers? (i need to monitor the boundary layer and the separation point)

Anyone please enlighten me.

cheers, Matt

Hi,

I will be able to answer the first question. Unsteady solvers are useful when you have a very non-linear problem. then steady state solution is tough but time step in unsteady solution acts as a sort of relaxation factor to control propagation of values in next iterations. moreover, an unsteady state solution can also be seen as a sequence of many steady state solutions. if you observe, you have to supply number of iterations for a time step. it should happen that once you take solution to long enough times (wrt time scales of your problem) you will have achieved a steady state solution. also, you should see that solution convergence is obtained per time jump as early as possible. this criterian should be adhered more stringently for incompressible flow approximations. hope this is on target. Swarup.

Thanks for the info and speedy response Sarup! that really cleared things up. I've tried the unsteady solver with my my other steady solver convereged cases, but it checked out with some discrepancies. Do you suggest that i should let it iterate with more time steps ( i was runnig it with a size of 0.001s for 2000 steps, which will give 2 seconds in total)?

If you mean that the converged solution for unsteady case is not matching with steady case solution then you should check why it is not happening. that is probably a separate case. i say this because if you do not see any change between converged solutions of two successive time steps then you have reached a steady state solution. so running the solution for more number of time steps (say >2000) may not straighten out the discrepancy. On the other hand, if the solver throws some error message, then you need to tackle the problem accordingly with a different scheme/time stepping mechanism (like adaptive stepping)and as a security, check all your inputs and BC values.

Regards,

Swarup.