Simulating a forced external flow loop with 2 phases
 

Hi all,

I am trying to simulate a bubble column with external flow loop (leaving at the bottom, re-entering halfway). There is a pump in this external loop, and only liquid runs through - the inlet is below the sparger and no considerable amount of gas is being entrained. The average flowrate through the external loop is known. In my simulations, I need to take this flow-loop into account; Lagrangian particle tracking is being done for a component in the liquid phase, and these particles can and will enter the flow-loop. Hence, simply modeling it as an in/outlet is not an option.

As general setup I use: euler-euler multiphase (fixed bub. size), k-e realizable, mixture formulation (dispersed diverged). Top of the column may be pressure outlet or degassing (for the latter, headspace section is removed).

So far, I have great difficulties in modeling the flow through this loop and hope you have some suggestions. I have tried the options below. In all cases, a small cell-zone (about 1/50th of the tube length in the center) was designated as the "pump-zone". I have tried enforcing the flow by:

- Setting a fixed velocity X = Qliq/A in the flow direction. This worked, but only if I (1) set the other velocities to 0 and (2) also fixed k and epsilon in the zone (based on the loop-average values). It only works when I use pressure-outlet; with degassing, gas starts to be generated out of nowhere inside this pump section (!) and the whole thing crashes. Even with pressure-outlet, things go wrong when I add particles: a considerable fraction of them gets stuck at the pump-zone; they stick at the face-centers for some reason, even though the particle velocity is finite and equal to the liquid velocity! I have no idea why this is, or how to fix it.

- Setting a momentum source term counteracting the loop pressure drop: Setting a momentum source in the flow direction works fine in single phase. However, in multi-phase, the pressure drop over the pipe changes considerably and as soon as I add gas, the flow in the pipe reverses, gas gets entrained, and the whole thing goes to hell (sudden spikes in velocity to 200 m/s or so randomly occur). This happens both with degassing and pressure-outlet. I have tried using an UDF that adapts the source term size, but the random velocity jumps that occur made it unmanagable.

- Using a fan-zone: not compatible with Eulerian multiphase, so no option.


I am running out of ideas, so I hope there are some solutions to be found here. Hope to hear from you!

Cees

Ok, I've made some steps (but ideas are still welcome).

The degassing setup instead of Pressure outlet works works, but only when using 1st order discretization schemes for turbulence and momentum (maybe not all are required to be first order, I have not yet tested them individually).
The issue with particles sticking in the "pump", however, remains.

This second issue is solved when I select massive particles instead of massless tracers. Giving the particles the same density as the fluid, and a very low diameter, ensures they will still follow the flow. However, computation is a bit slower with this model due to the additional operations.