CFD Online Discussion Forums - Laminar gas flow under slight suction

Hi all,

I am working on a large CFD project and this portion of the work is just a small step to help me build the final model. I was assuming that this would be quite a straight forward and easy simulation, but here I am...

Essentially, I have a low flow of gas entering a rotary kiln (long horizontal pipe) under laminar conditions. The walls on the kiln are rotating at 2rpm (although I've also tried the simulation without rotation). The inlet condition specifies the normal gas velocity to 0.00336 m/s based on lab conditions. I am trying to run this as a steady simulation because it will get quite complex once fully built.

The outlet here is a series of 8 openings around the pipe wall at a distance of 3m from the inlet. I originally assumed (before lab work started) an outlet static pressure of 0 Pa and my simulations converged well and quickly, plus the results seemed perfectly reasonable (not fully checked against analytical solution yet).

Now, when doing experiments, we noticed that the pressure condition at the outlet is actually under a negative pressure of 29 Pa (caused by the afterburner that follows the kiln). I thought this would just be a matter of changing the pressure at the bc and voila, but no. I couldn't get convergence. Tried different time step, tried decreasing the pressure slowly, tried starting from different previous simulations, nothing helped.

I decided to run a transient analysis of the system. After fiddling with the time step, I was able to get the residuals to decrease to very low levels. I had a good look at the results from the transient analysis, and variations in velocities and pressure during the simulation are very small - can't be noticed by just looking at figures. I tried running a steady simulation from the results of the transient analysis, but no luck.

I don't think this should be a transient situation, but I may be wrong. I'm including a figure that shows the system layout (mesh is tet mesh with 1.2 million elements and passing all CFX tests). The other 2 figures show the RMS history plots for my steady and transient runs. I'm a bit puzzled by the behaviour of the transient run RMS (very bouncy, steps are caused by changes in timestep size), if anybody can say anything about that I'd be interested.

Any advice/suggestions/comments are welcome here, I'm at a loss.