[Gert-Jan]

Hi All,

I have a steady state setup with rotating frame of reference. In this frame several objects move on a conveyor belt counter clockwise, through multiple chambers. These chambers are standing still in a stationary frame. In one chamber, the air is heated.

My problem is that I have two velocity fields: Velocity in Stationary frame and simple Velocity. The latter is the relative velocity, relative to either stationary or moving frame.

Now I expect my hot air to mix up in the velocity in stationary frame. Meaning that in the moving frame, the hot air will move counterclock wise together with the obtsacles. But this is not the case. The hot air is moving clockwise.

Bottomline, the problem is that the hot air is mixed in the Variable Velocity. This is correct in the Stationary domain with the chambers, but not in the rotating domain with the obstacles. Does anyone know a magic button to get realistic results? To let CFX perform mixing in Variable Velocity in Stationary Frame instead of the Variable Velocity?

I know the answer is transient. But a steady state, i.e. photo of the process, shouldn't that also be possible? Moreover, it would be a very good method to obtain an intiial guess for a transient analysis. If I would take my current steady state result as initial guess for a transient anaylsis, I would be far off since the hot air accumulates on the clockwise side, where it should go the counter-clockwise-side.

FYI: in Fluent, I notice the same problem.
FYI: The Alternate rotation model does not fix the problem.

Forgive me the coarse mesh/setup etc. It is just a simple test to find the correct settings.

Regs, Gert-Jan

[Opaque]

Since you are talking about an expected "hot air flow direction", can I assume you activated the buoyancy model?

If you did, keep in mind that gravity must be aligned along the axis of rotation; otherwise, the problem is transient since gravity changes direction in the rotating frame.

[Gert-Jan]

Gravity is not on. In my final application it won't be relevant either. Here I just have a small test setup to see how CFX handles my temperature field.

Since air pressure on both sides is zero, I see and expect air to be dragged along with the obstacles in counterclock-wise direction. So, I expect the hot air to also move counterclock wise, with the obstacles. It is not.

I have seen this before in other applications, so I am not really surpised, but am hoping for someone to know a magic button. Certainly I asked ANSYS, but no response yet.

[Gert-Jan]

There is a sort of workaround. In Post it is possible to overwrite variables with expressions. I managed to overwrite the variable Velocity with variable Velocity in Stn Frame. So I have Velocity in Stn Frame everywhere.

Then I used this modified results file in the solver manager and mixed temperature using solve fluids=f and solve turbulence=f. But results were not realistic either. It was not more than "a nice try".

[ghorrocks]

I do not completely understand your problem, but on the face of it I would suspect that the rotation is not set up correctly - are you sure the right bits are rotating and the right bits are stationary? I cannot think of any other way how you can get what you are seeing.

[Gert-Jan]

It is fairy simple. Just take an existing case with Frozen rotor, calculate temperature or any other scalar, and you will see that both mix in variable Velocity and not in Velocity in Stn Frame. In other words, the scalar will not be transported with the obstacle in the rotating frame, but in opposite direction.

[ghorrocks]

Thanks for confirming that.

I have not fully considered this, but my initial thoughts are that the Frozen Rotor approach is not suitable in this case. If you run it transient rotor-stator (and a full transient model) does it work correctly?

[Gert-Jan]

Yes, then it will work correctly, but this is not cost effective for my real case which is a very long conveyor belt.

I don't see why the scalar can't mix in Velocity in Stn Frame when Frozen Rotor is used and in Velocity when Transient Rotor-Stator interactions is used. That would be much more logical and in line with intuition. Also, a Frozen rotor would then provide a suitable intial guess for a transient case. Now, the scalars and temperature end up on locations that are opposite from what is expected.

Besides rotation, Fluent has an option for translation, as is used for e.g. trains in tunnels. A bit like my obstacles on a conveyor belt. But then I have exactly the same problem in steady state. I was advised to go for deforming mesh. I consider that as overdone for my case.

Any other suggestion is welcome.

[Gert-Jan]

Originally Posted by ghorrocks

Thanks for confirming that.

I have not fully considered this, but my initial thoughts are that the Frozen Rotor approach is not suitable in this case. If you run it transient rotor-stator (and a full transient model) does it work correctly?

As mentioned, I tried fluent with translation, and found the same behaviour. I posted it on the fluent-forum: Temperature mixing up in wrong velocity field


There, I added 3 pictures at a horizontal cross section, making the problem even better visible.

[Opaque]

Originally Posted by Gert-Jan

Hi All,

I have a steady state setup with rotating frame of reference. In this frame several objects move on a conveyor belt counter clockwise, through multiple chambers. These chambers are standing still in a stationary frame. In one chamber, the air is heated.

My problem is that I have two velocity fields: Velocity in Stationary frame and simple Velocity. The latter is the relative velocity, relative to either stationary or moving frame.

Now I expect my hot air to mix up in the velocity in stationary frame. Meaning that in the moving frame, the hot air will move counterclock wise together with the obtsacles. But this is not the case. The hot air is moving clockwise.

Bottomline, the problem is that the hot air is mixed in the Variable Velocity. This is correct in the Stationary domain with the chambers, but not in the rotating domain with the obstacles. Does anyone know a magic button to get realistic results? To let CFX perform mixing in Variable Velocity in Stationary Frame instead of the Variable Velocity?

I know the answer is transient. But a steady state, i.e. photo of the process, shouldn't that also be possible? Moreover, it would be a very good method to obtain an intiial guess for a transient analysis. If I would take my current steady state result as initial guess for a transient anaylsis, I would be far off since the hot air accumulates on the clockwise side, where it should go the counter-clockwise-side.

FYI: in Fluent, I notice the same problem.
FYI: The Alternate rotation model does not fix the problem.

Forgive me the coarse mesh/setup etc. It is just a simple test to find the correct settings.

Regs, Gert-Jan

Gert-Jan, from the velocity and temperature plots something seems off in the setup.

If I understood the setup, there are separation plates creating cavities, correct? Those plates are uniformly spaced in 360, correct?

If the understanding is correct (I can be confused), I will be afraid of using periodicity boundary conditions at exactly the next occurrence of a plate. I will create the mesh such it cuts a "cavity" so the flow is periodic to the "next cavity" that completes the sector being modelled.

Still spinning my head around.. Sorry.

[Gert-Jan]

Please follow this link:

Temperature mixing up in wrong velocity field

These pictures, generated by Fluent, clearly show what is going on. The geometry is very similar, it is however a translation, not a rotation like in CFX. But the effect is exactly the same.
The obstacles in a moving frame go to the left, so is the air as shown using Velocity in Stn Frame. However, the hot air goes to the right, with the relative velocity field. Here it is relative to the moving frame, leading to a velocity to the right. Confusing indeed.

[Opaque]

Have you tried using the mixing plane model?

Using the frozen rotor approach, the cavities downstream will never see the effect from the cavities, flow upstream. However, the "time-averaged interface model", i.e. mixing plane, should connect all the cavities upstream and downstream.

Do not worry about the naming "mixing plane", it is effectively a circumferential averaging that attempts to predict the time averaging in the transient model.

[ghorrocks]

Have you confirmed that a Transient Rotor Stator model has the hot fluid going in the right direction? It would be good to confirm this because TRS also uses velocity in the rotating frame of reference, so if a TRS model works correctly we will have to think about why frozen rotor goes in the wrong direction but TRS goes in the correct direction. Or maybe TRS goes in the wrong direction as well?

[Gert-Jan]

In transient mode it works well. Then the temperature is shifting to the left. This is because.....

in transient mode, the rotating geometry is physically shifting to the left. This rotation speed is higher than the air velocity to the right, leading to a net transportation to the left. In fact, because of drag with the environment the air is lagging a bit behind the fixed rotation speed.

________________
Some background:
In rotating applications, the variable "Velocity" is the difference between the "Velocity in Stn Frame" and the Rotation speed:

Velocity = Velocity in Stn Frame - Rotation speed

In my rotating domain, the rotation speed > "Velocity in Stn Frame", so the variable "Velocity" is opposite to the rotation direction.

In the end...... with frozen rotor settings, the geometry is frozen, and the temperature is mixed in the Variable Velocity, which is opposite to the Rotation direction. That is the main problem.

Transient is too costly, so my escape would be to mix the temperature in the variable "Velocity in Stn Frame". How?

[Opaque]

How are the results from mixing plane/stage model?

[Gert-Jan]

A mixing plane does what it says, It mixes over the Interface where Frame change take place. Or tries to determine an average over the Interface. As a result, the temperature goes to the left and the right.

[Opaque]

Good. Then, you should be getting something close to the transient rotor-stator model.

Mixing plane solution ="In theory"= TimeAveraged(Transient Rotor-Stator solution)

That is better than any Frozen Rotor solution.

The Frozen Rotor solution has two issues:
1 - Because of the "frozen" relative position, it allows the "wake" to propagate "too far" downstream with respect to the true solution
2 - As you have seen, it does not "see" the upstream influence since the relative position is never updated.