Massed particles behavior in centrifugal separator

[PaulNielsen]

I am simulating basically a rotating cylinder of oil (see attached .png file). The purpose is to during post-processing release massed particles which have a lighter density than the fluid and see how they behave.

The results I should be seeing is that lighter particles move inwards and upwards.
The results I am getting are however showing the particles moving radially outwards. But still upwards at least so the gravitational pull seems to be working but not the buoyancy. 

For heavier particles it seems to work fine as intended.

Is there a way to achieve what I want to see?

I am running this in CFD2018 as the trace module is heavily bugged in 2019. I have also checked both with and without relative velocity.

[PaulNielsen]

I tried attaching .cfz file but it won't accept it for some reason.

[PaulNielsen]

I zipped the .cfz to try if it worked this way and it appears to work like this.

[Anonymous]

The basic accounting for mass density seems to be working OK.

[PaulNielsen]

As i mentioned in my initial post i also found that gravitational pull seems to be working fine. The problem is when rotating the lighter particles aren't behaving as intended (i.e. moving towards center of rotation). Heavier particles do work.

[PaulNielsen]

Here is a simple example of my problem lokking at flow around a bend.

Fluid has density of 900kg/m3 and 46cst viscosity.

Adding massed particles with 2000kg/m3 gives intuitively correct results:

Adding lighter massed particles with 1.2kg/m3 should yield opposite results. however this is what i get:

This is same as the resulting traces without mass.

Adding massed particles with 450kg/m3 (half of the fluid density) yields:

Which is opposite of how the particles should be moving. Unless i am largely mistaken in some way.

[Anonymous]

Those results seem plausible to me.  I see no way for a very light particle (small diameter and density), which is seeded on the outer diameter, to just cut across the flow field towards the inside diameter.  Particle traces are post-processing only so the flow field is already set.  The massless traces show this by basically following the curve around since there is no outside forces on them and they are slaved to the existing flow streams.

Perhaps you need to restrict the seeding to only the inside diameter.  That would show that light particles would hug the bend while heavier (or larger diameter) particles would have their momentum carry them across to the outer diameter.

[Anonymous]

"particles aren't behaving as intended (i.e. moving towards center of rotation)"

am wondering here if the problem comes from the flow solution instead of the traces tool. I mean isn't the physic you are trying to capture the result of 2 combined effects?  vorticity fluxes+secondary flows and inertial+drag effect.

Is the turbulent flow solution completely accurate? (turbulence model/analytical validation/vorticity/etc...)

The literature might contain useful reference on this bended oil pipe+mass particle problem.

[PaulNielsen]

But then why does it work for heavier particles of the same size?