Learning Forum

[Anuhar.Nesme]

Hi everyone,

I am simulation with Fluent one-phase, laminar, flow in a container with dense fluid (rho= 6440 kg/m³, visc= 2000 Pa s), which is stirred by 2 different elements: one rotating at 28 rpm, the second at 100 rpm, both also rotating together around container axis with 9.33 rpm. The container is not filled completelly, then, instead of using costly VOF-Simualtion, I decide to approximate the interface liquid-gas either as slip-wall or as pressure outlet. I set my reference pressure (101325 Pa) on this surface (0, 0.437113, 0) m. After initializing the simualtion with static (hydraulic) pressure, the solver converges with reasonable velocities but I am getting inside the fluid domain huge negative values (~ -12 bar). I also limited the pressure in the set up, but it seems it does not help to much. I have also reduce the dt, change the pressure models (presto, second order) but it continues with negative values.

Does anyone a possible reason?

I appreciate in advance your help!

Anuhar

[Rob]

What are you mixing? With 100rpm and that viscosity you could see some very odd effects. 

[Anuhar.Nesme]

Hi Rob,

the mixture consists of water and powder (colour) and these are the resulting density and viscosity after mixing them (final product). 

[NickFL]

It could be the presence of cavitation. Physically the pressure itself does nothing to the flow field, it is the derivative of the pressure which drives the flow. In this way it does not matter to the solver what the pressure is, it can be adjusted up and down and also go negative. This is what we mean when we say Fluent lets the pressure float. The actual value really only contributes when looking at the material properties. In nature if the pressure sinks below the vapor pressure it would cavitate and we would have vapor bubbles in our tank. But if we are using constant property fluid, we don't allow this in our model. So, what the solver does is it goes into negative values.

Without knowing more about your problem, we cannot say if it is appropriate. Potential problems could be that model is scaled too small or too big and the rotations are not physically realistic for this size tank. Or it could be that there is nothing wrong and there is cavitation occurring in the tank.

[Rob]

2,000 Pa s is a bit on the thick side, water is 0.001 Pa s and some of the highly viscous gums I've come across are 10-20 Pa s but nonNewtonian so the flowing part is significantly less viscous. 

[Anuhar.Nesme]

Thank you NFLynn. Actually the model has the real dimensions and boundary conditions except for the interface liquid-gas, which as I wrote is assumed as slip wall instead of free surface to avoid a VOF-Simulation. In our laboratory one of the mixers (the one rotating at 100 rpm) undergoes a rupture at the shaft when it rotates under these considitons. For these reasons we conducted first a CFD-Simulation. BTW, based on the Reynolds number calculated at the fastest mixer (diam= 0,5 m) the flow should be laminar (~ 1,5).

 

[NickFL]

The free-slip boundary condition will be good enough. You mention the shaft breaks, so you have access to the mixer. I expect the smallest pressures (highest negative) occur very near the blades. Do you see any damage to the physical blades themselves?

And I second Rob’s comment below. What do quantities like torque look like when running the simulation? You should have these set up as report definitions that you are monitoring during the simulation. Are they comparable to what is measured in the laboratory? Maybe start off at a lower speed to validate the model so they don’t go breaking any more shafts.

[Rob]

What does the torque report return? Remember in CFD the mixer will rotate at the speed set regardless of whether it's possible to get a motor big enough to spin it.