Meshgrid

By using the symmetry of the system, the meshgrid need just to model a sixth of the box. You can find the plan of the box here to visualise the studied system. The meshgrid has been created with the sotfware Gambit. At first, I have modelled a sixth of the box. And after I'va tested this meshgrid in single phase simulation to approve it. Representing only the sixth, we reduce the size of the meshgrid and so to reduce the time of the computations. We can have as well a refined meshgrid without going past the quota of the student version of the software Fluent (limitation to 65 000 meshes). The web of hollow fibres has been modeled as a unique membrane in the center of the system. Therefore, it isn't possible to observe the membrane vibration. The study will center on the dynamic aspect. You can see on the right the meshgrid in tetraedric. It contains more than 40 000 meshes. The injector is modelled with a small cylinder. We won't be able to visualize the disturbances caused by the add of the injector. But as the studied phenomenon are located upper, the modelling is acceptable.

Boundary conditions

The model conatins 6 differents types of boundary conditions: walls, membrane, air inlet, waste water inlet, symmetries and the air/water interface on the top. The air and waste water inlet are modeled as velocity inlet. I'va choosen a velocity inlet for the membrane too, but with a negative magnitude to model the water outlet. I've developped a little Matlab programm to calculate the right velocities to fix on the waste water inlet and on the membrane. So we can conserve the flow. The setting of injector velocity can be a problem. The surface of the injector is indeed very small and the meshgrid isn't enough fine to well estimate it. So, I have tested different velocity magnitude to set the desired air flow.

In the beginning, I have choosen a pressure oulet for the boundary condition on the top. The results in single-phase simulation have shown some backflows on this boundray condition, but the mass conservation was good. But I have had some problems with the two-pase simulations. So, I must find a better model for this boundary condition. The best idea is to code it with an UDF (User-Defined Function). The water must be blocked (like a wall) and the air must be pass through without resistances (like a pressure outlet). But I didn't suceed in coding right this UDF.