[driu]

Hi Everyone,

I am modeling a case that seemed very simple at first. However, it appears to be more complex than what I was expecting.

It is a stirrer in a dish. The stirrer is off centred. I am interested in getting streamlines or particle trajectories from specific seed points over finite period of time. I am using CFX to do that.

I have started with a frozen rotor approach and found out that:

a) Streamlines enter the rotating domain but end up terminating at the stirrer (on the trailing side). This happens when the variable is Water Velocity in Stationary Frame.

b) Or, the streamlines can enter and exit the rotating domain ok but their path is broken (opposite direction on stationary and on rotating). This happens when the variable is Water Velocity (stationary frame for stationary domain and rotating frame for rotating domain).

Please see some pictures attached.

In addition to the frozen rotor simulation, I have done transient rotating runner simulation. Same problem. I suppose this is because streamlines only use the instantaneous flow field.

Can anyone think of a way to get past this problem by using streamlines? Or is particle tracking my only option?

If so, is it possible to model massless particles in CFX?

Looking forward to your comments!

Kind regards,
Audrius

[ghorrocks]

FAQ: https://www.cfd-online.com/Wiki/Ansy...f_reference.3F

[driu]

Hi Glenn,

Thanks for the link.

Apologies if I am missing the point but I knew the difference between the two variables and I knew that the Velocity in stationary frame should be used to have continuous streamlines. Problem is that by using Velocity in stationary frame I have an issue with streamlines terminating at the stirrer. They terminate at the 'trailing' edge/side of the stirrer if you will.

Please see the attached image of a 2D example. The way I understand it is that based on the velocity field, the streamline would continue where the stirrer is. Because streamlines are based on the instantaneous flow field, the stirrer does not move and is blocking the way for the streamline. So the streamline terminates.

My question is are streamlines fundamentally not suitable for what I am trying to do or is it something I do wrong?

Many thanks,
Audrius

[ghorrocks]

They terminate at the 'trailing' edge/side of the stirrer if you will.

Yes, that's right. On a wall with a moving normal component (eg your stirrer), then there will be a velocity normal to the wall. This means some streamlines will start or terminate at the moving wall.

A stationary wall has no velocity normal to the wall, so a streamline cannot terminate at a stationary wall.

[driu]

Thanks, so does this mean that particle tracking and a transient simulation is my only option to make sure the streamlines (or particle trajectories) do not terminate at the moving wall?

[ghorrocks]

From the definition of streamlines there must be some streamlines which terminate at a wall with motion in the normal direction in a steady state flow. Also, from the definition of streamlines, they will not terminate at a stationary wall. There is nothing to make sure of, CFX will not be solving the Navier Stokes equations if it infringes this.

[driu]

I'm sorry if it seams that I am repeating the same question, but my point is what should I do to track a point particle through the domain. If I use streamlines, by definition they will terminate at the moving wall. That's in steady state. Even if I do transient simulation, the streamlines actually represent steady state because they are based on instantaneous flow field. So this means that I shouldn't be using streamlines for this application, correct?

[ghorrocks]

No, this is a different question - you are now asking what path does a particle take through the domain.

If you draw streamlines in the stationary frame then that will be the best picture of a flowpath through the device. Some streamlines will terminate on the rotor, but most will not so put enough in so you get some going past. This is unavoidable when you are doing a multiple frame of reference simulation.

The alternative is to do a full transient simulation and model massless particles. Then you will get the true path lines, but the effort involved in doing this is far higher.

[driu]

Thanks Glenn, really appreciated. This now defines my problem very well. I am indeed interested in what path does a particle take through the domain.

I will try to use many streamlines to get past the rotating domain. The problem here is that if a streamline terminates at the stirrer, this means that I do not have the information for that particular seed point.

A quick thought, maybe if I see it terminating at the stirrer, I can do another frozen rotor simulation where the rotor is in a different position, say at 90° offset and hopefully the streamline originating from the same point will get past. Not a watertight analysis though...

There might be some follow up questions from me for the massless particle tracking but I admit I need to read a bit more on particle tracking. I had just a quick look and was struggling to specify them as massless.

[ghorrocks]

I should point out that if you do a full transient simulation with particles, then if you draw it with the rotor at a fixed position then some path lines will go through the rotor. The only way you will avoid path lines going through the rotor is to do an animation of the particles and rotor. In other words you need to include time, you cannot do it in a single image.

The problem here is that if a streamline terminates at the stirrer, this means that I do not have the information for that particular seed point.

At a different rotation angle that seed point might result in a streamline which goes through..... You mention this in your next sentence.

[driu]

I should point out that if you do a full transient simulation with particles, then if you draw it with the rotor at a fixed position then some path lines will go through the rotor. The only way you will avoid path lines going through the rotor is to do an animation of the particles and rotor. In other words you need to include time, you cannot do it in a single image.

Sure, actually my goal is not an image but a data set that contains [time], [x], [y], [z] for each path at each timestep.

[ghorrocks]

OK, in that case you need to be aware that the frozen rotor approach is an assumption which obviously only takes into account one position of the rotor. Under some circumstances this is still a reasonable approximation of the true transient flow, but under some circumstances it is very incorrect. So be aware of the assumptions inherent in your approach.

[driu]

Oh yes, ideally it should be a rotating runner simulation. However, it will most probably make the computational cost unfeasible. I will see how it goes.