[cob]

Hi everyone,

I am trying to model a gas compression process in a simple closed cylinder. It is a pretty big geometry with a diameter of 19 mm and a length of 2m (see geometry). As the problem is asssumed to be axisymetric I model only a small wedge and generated a structured mesh in icem (see picture). I model only the gas phase so the compression is performed by mesh deformation, moving the bottom boundary in upward direction for 1m in a transient simulation. There is also some heat transfer taking place, since the side wall has a fixed temperature and the gas temperature rises during the compression process. The boundarys near the centerline have symmetrie conditions. The buoyancy model is activated to catch convection effects. Since this is a very slow stroke of 60s there is no turbulence model used. The material is Helium Ideal Gas.

I am interestet in the heat flux at the side wall for different initial pressures, temperatures and cycle times. For low intial pressures I get good results. But for higher initial pressure (25bar) I get some wiggles in the heat flux trend (see picture). I think the simulation is still converging since the residuals are met 10^(-5) and the imbalances are very low.

I have done some sensivity studies, but when increasing mesh cells and the number of timesteps the results get even worse.

So I would like to know what couses the wiggling trend? Is it because bad meshing (bad cells near the center line), no convergence or should I apply a turbulence model to resolve the natural convection?

Thank you for help!

Cob

[ghorrocks]

Have a look at the mesh in the wiggly region. Is it staying straight? But the tight wedge angle on this thing will cause issues in itself. Try a larger wedge angle (but still only 1 element thick).

If you are getting psuedo-turbulent flows then you will see that in the velocities. So you can check that one yourself as well.

[cob]

Hi ghorrocks,

thanks for your quick response. I checked the mesh in the wiggly region. You are right, it is badly distorted. The attached picture shows the upper boundary of the cylinder in the last timestep. The distortion process starts very slight right from the beginning of the simulation and gets bigger. It also looks like if the cells in the center position (left side) are distorted more quickly.

Do you have any idea whats causing this problem?

Thank you for help.

[ghorrocks]

The mesh diffusion algorithm in CFX causes problems when you compress a mesh by a large amount. The V15 mesh diffusion algorithm is meant to improve this, but I have never tested it so cannot say. You might be able to improve the situation by adjusting the mesh diffusion parameters.

If that does not work you might need to explicitly define the mesh motion using a fortran routine.