structured grid and/or increase convergence rate

Ashish Pande

Dear all,

I am Ashish Pande, a new user of Fluent. I am trying to model 3-D, 1 phase, batch/continuous, incompressible, steady state model in a single rushton turbine stirred tank. Latter on, I want to extend this to 2-phase solid suspended model. It is taking too much time for convergence, just for 1-phase calculations. I am using Multiple reference frame model. I have roughly 1,50,000 nodes and 6,50,000 cells. To speed up the convergence, I've tried to play with under-relaxation parameters and changing from SIMPLE to PISO, but that is not not much use. Now my problems are as follows.

1) I have generated unstructured mesh in gambit. Should I switch over to structured mesh to speed things up?

2) I tried to generated structured mesh. In the inner zone (around impeller) I could use quad-map for faces everywhere except the caps of the cylinder enclosing the zone and the bottom of frustum. (To use quad-map, I had to split up the faces and so on). Now the modelling guide says that at least one of the surfaces must not be sub-mappable for cooper scheme to apply (so this condition is fulfilled in my case) and non-source faces must be mappable, which is true in my case. But even then, when I select the source faces as (split faces of) the bottom of frustrum and caps of zone, error says that either the face is not sub-mappable or the choice of source face is not correct. Does it mean that the source faces also have to be sub-mappable for Copper? It wud then violate one of the conditions given above.

3) Is there any other way to use structured/block-structured grid for my case? (I mean without going for cooper scheme)? Most of the literature on rushton turbine modelling uses structured grid (but only half of the geometry of the tank is modelled, for other half symmetry condition is used). Also I don't know how to create block-structured grid in gambit.