[DarioP]

I successfully ran a 3D simulation of a simple rotating fan in a duct using a cyclic AMI modeled on the rotatingFanInRoom tutorial.

One conceptual problem I encountered is that I have a small gap between the fan blades and the duct wall which is sliced half way through by the cylinder representing the rotating region. The proper meshing of this gap requires plenty of refinement and introduces a lot of cells, slowing down the simulation while capturing an effect (the backflow in the gap) which I do not really care about at this stage.

I would like to simplify the study by getting rid of the gap between blade and wall. But if I make a geometry where the two touches, then how do I setup the rotating region? I tried to have a cylindrical rotating region with a radius larger than (and equal to) the one of the wall, but then SHM throws away the side face of the cylinder, and, while the simulation runs, the fan does not rotate.

Any hint would be much appreciated. Thanks.

[GerhardHolzinger]

A halfway solution to your problem, would be to incorporate the duct-wall next to the fan into your AMI region, and prescribe a zero-velocity BC for the duct-wall.

In the attached sketch, everything in blue represents the contents of the AMI region. The dashed red line denotes the boundary of the AMI region.

In the left case, the AMI boundary sits right in between the fan and the wall, which causes all the mesh-refinement requirements you have outlined in your question.

However, if you were to follow the strategy on the right, then fan and duct-wall are within the AMI region. Hence, you have more relaxed meshing requirements. However, special care needs to be take that the blue section of the duct-wall remains static.

[DarioP]

Yes, that is exactly what I had in mind: I have the case on the left of your picture working and I am trying to setup the case on the right.

I have realized that one way to go is by using two pairs of master-slave patches, so I will have two disk-shaped STLs, one upstream and one downstream of the fan, together with an STL cylinder to mark the rotating cells (which in the left case is also used to setup the master-slave pair).

I have no idea about removing the rotation of the duct wall (any suggestion is welcomed!!!) however, this is an higher-order effect and using a fully-slipping boundary condition is already a step towards the ideal setup.

EDIT: I suppose I can cancel the wall rotation by using a rotatingWallVelocity boundary condition with omega opposite to the one of the rotating mesh.