Convergence problems
 

Hi, I'm new to CFD and I'm having some convergence problems with my simulation so wonder if someone here can give me some advice.

What I'm trying to do is to simulate the flow in a refiner for processing paper fibers. The refiner consists of two plates facing each other, one with bars and one flat plate. The flat plate rotates at around 750 rpm and the other plate is not moving. The distance between them is around 0.2 mm and the diameter of the plates is 600 mm.

The narrow gap between the plates and the velocity of the flat plate makes it difficult to get a converged solution so I wonder if you have any tips on for instance what visocity model to use or if there are any other things I can do to improve the results?

I've tried using k-epsilon and LES models but only get converged results for velocities around 1 rpm. The mesh I'm using has around 1.2 million elements and I don't think I can add so many more elements or the calculations will take too long.

I'm adding some pictures of the mesh so you can understand the problem better. The first picture shows the whole model (a coarser mesh than used in the simulation) and the second one is an enlargement of the gap-region (actual mesh size used in the simulation).

Thanks!

You said that the solution won't converge, but what specifically is it doing? Also, what are your boundary conditions?

With respect to your grid, it looks like you used a uniform grid spacing along the flat plate. I would suggest clustering grid near the entrance/exit of the gaps, as there are gradients there that you need to resolve. The sharp corners might be causing you some problems, too.

Hi Chris, thanks for your help!

I've tried your suggestion and modified the mesh (hopefully the way you suggested, not sure if I understood you correctly). I am using pressure boundary conditions on both inlet and outlet.

Because the rotating wall is adjacent to the inlet in the model above I got a reversed flow in that area (because of the centrifugal force on the fluid close to the rotor I guess) so I extended the inlet a bit which solved the problem. However, despite the modification of the mesh and the inlet I still have a problem with the convergence. When I use the first order solver I can get the largest residual (continuity) down to approximately 4e-3 (for 600 rpm, when I decrease the under-relaxation factors of the pressure to 0.1 and the momentum to 0.4) but when I try using the second order solver the residuals jump up again.

Below are some pictures to illustrate how the solution behaves. The first 4000 iterations I have used a first order steady solution and then in the first picture I have utilized that results to initialize a second order transient solution with 1 ms time step. In the second picture I've used the first order steady solution to initialize a second order steady solution. Sorry about the monochrome pictures - the largest residual is the continuity in every simulation. I also attached some pictures of the new mesh.

Here is a picture of the model aswell

What are you monitoring besides the residuals? Have you looked at the solution to see what's going on?