centrifugal pump self-priming simulation
 

Setup for self-priming CFD
1.Turbulence model:
SST
2.Two-phase simulation:
Fluid1: water at 25C, continuous fluid;
Fluid2: air at 25C, continuous fluid;
mixture model and interface length scale is 1mm;
3.Boundary conditions:
Inlet pressure is 5kPa;
Outlet is Opening;
4.Impeller:
Rotating speed: 2900rpm, rotating axis is Z;
5.Buoyancy model:
Gravity of X: 9.81m/s2;
Gravity of Y: 0;
Gravity of Z: 0;

Problems of CFD
1.The overall self-priming process takes 2 seconds. But in physical test, it needs 60 seconds.
2.In the simulation, the air and water are discharged together. But in physical test, air is discharged firstly and water is left in pump. After all the air is discharged completely, water is pumped out.

Show us the pressure field at the start. Did you include your hydrostatic pressure?

centrifugal pump self-priming simulation
 

Dear Gert-Jan,
Thanks for your reply. Here is pressure of start time. I choose the transient simulation, releative pressure of domain is 0 Pa, the inlet pressure is 5kPa. The total simulation time is 60 seconds, and the time step is 0.005s.
/Denghao Wu

You have a nice initial guess for the volume fractions. But you forgot to include the hydrostatic pressure.
Read the paragraph "Modeling Advice for Free Surface Flow" in the help of CFX. It is paragraph 7.18.5.4 in v19.0

centrifugal pump self-priming simulation
 

Dear Gert-Jan,
Thanks for your suggestion. I take the hydrostatic pressure into consideration. However, the air and water are still discharged together. I will really appreciate if possible you can help me check the CFX file? It is an important task for me, I put my energy on this case lasting one month.
/Denghao Wu

Water comes up? So that is ok, not?
It is also transported to the top. And you don't expect that?

This can be caused by a lot of parameters......
I try to image a pump that pumps up liquid in an empty duct. This could be quite violent. A lot of splashing. So maybe it is not so unrealistic. Don't expect a very distinct free surface.

Questions I have:
- do you have experimental verification?
- do you ramp up from 0 to 2900 rpm?
- quite fast rotation, not? What is your tip speed?
- pump diameter?
- mesh size? Mesh type? Please show the mesh in the duct above the pump.
- typical model settings. Both continuous models.
- did you switch on free surface with compression? Could help a (little) bit....

centrifugal pump self-priming simulation
 

Dear Gert-Jan,
Yes, I did the test. In the physical experiment, we can see the water is sucked into pump inlet pipe, air is discharged to outlet, but the water is still left in pump chamber. After all the air is discharged completely, the water is pumped out.
1. for the pump speed: I fixed the speed to 2900rpm as constant.
2. the tip speed of impeller outlet is 21.56m/s.
3. the diameter of impeller is 142mm.
4. total mesh size is around 4218158. It is hex structured mesh type.
5. Yes, water and air are both continuous.
6. I will try to switch on free surface.

I think you are not that far of. I think the water in the upward duct would more or less behave the same in an experiment if you would be able to rotate your impeller with 2900 rpm. Instantly!

Btw, this equals 50 rotations per scond! Are you sure about that? You cannot even see the impeller rotate that with your own eyes. Did you calculate the dynamic pressure? You will create a vacuum, so cavitation starts to play a role here too.......

Moreover, there will be a lot of violent mixing of air and water, before the water has reached your impeller. Also, all the air in the duct upstream the pump has to go through the pump. Even more violent mixing? Would you expect a flat surface around the impeller? I would not believe it.