[Julian121]

How can I calculate the frequency of a flow phenomen?
Assuming I have a unknown frequency at 480 Hz based on FFT analysis, does it mean a unkown flow phenomenon is reapted with 1/480 period?

[ghorrocks]

I don't understand your question. If you have done an FFT and got a peak at 480Hz all that means is that it is picking up a 480Hz signal. Pretty obvious really. What that 480Hz signal is and whether it is known or unknown is up to you - you have not provided any details so we cannot help you further.

[Julian121]

Yes, there is a peak at 480 Hz. I understand that!
I’m simulating a compressor near to stall, so I’m expecting to see the frequency of stall cells that rotate around the annulus.
Stall cells are defined as low axial momentum or reversed flow regions.
The attached images show FFT result and axial velocity fields upstream of the rotor leading edge.
As seen, low axial momentum is accumulated at the tip.
I wonder if you could please tell me how to approximate the frequency of these low axial momentum regions and whether it is 480Hz?

[ghorrocks]

I don't know what parameter you have taken an FFT for, or which point it applies to. I don't know what the low axial momentum regions are in your image and I don't know what your geometry is. I also don't know how the stall cells propagate around the circumference.

So I am doing a lot of guess work here. But don't you simply work out from your simulation how fast you anticipate the stall cell to propagate around the circumference?

[Julian121]

Originally Posted by ghorrocks

I don't know what parameter you have taken an FFT for, or which point it applies to. I don't know what the low axial momentum regions are in your image and I don't know what your geometry is. I also don't know how the stall cells propagate around the circumference.

I should have provided more details about my work.
According to the literature, a stall cell propagates between 20% to 70% of the rotational speed of the rotor.
The regoins of low axial momemtum have been marked in the image.
The pressure signal was recorderd at IGV/Rotor interface and then the FFT analysis was made.

Originally Posted by ghorrocks

So I am doing a lot of guess work here. But don't you simply work out from your simulation how fast you anticipate the stall cell to propagate around the circumference?

I guess it is trivial but I do not know how! Could you explain please?
Are you using circumferential velocity and radius to approxiamte the rotational velocity of stall cells?

[ghorrocks]

I still don't know much about what you are asking.

Are you saying the literature says the recirculation cell, when just on the stall point, travels around the circumference at 20%-70% of the rotor speed? And that the rest of the flow remains attached with no significant recirculation?

If you have modelled this using only two blades I would not think you would be able to pick up this sort of phenomena. I would think you would need a full rotor model (or at least a lot of blades, maybe half the rotor).

[Julian121]

Originally Posted by ghorrocks

I still don't know much about what you are asking.

I would like to find the source of that 480 Hz. Is that really the frequency of the low axial momentum region (stall cell) or it is because of something else?

Originally Posted by ghorrocks

Are you saying the literature says the recirculation cell, when just on the stall point, travels around the circumference at 20%-70% of the rotor speed? And that the rest of the flow remains attached with no significant recirculation?

Yes, exactly but it may not occur exactly at stall point. The exact point of stall cell formation is not known. Nevertheless, it starts as part-span stall and then evolves into a full-span stall cell.

Originally Posted by ghorrocks

If you have modelled this using only two blades I would not think you would be able to pick up this sort of phenomena. I would think you would need a full rotor model (or at least a lot of blades, maybe half the rotor).

I have modelled just one passage. According to the literature, at least two or three passages are required to find the frequency of stall cells propagation around the annulus. Having said that, what I need to know is a way to approximate the rotational speed of the low axial momentum assuming it is a stall cell.

[ghorrocks]

To see what is causing the 480Hz peak you have to save lots of transient results files and look at them trying to pick some disturbance which is 480Hz. It may or may not be the stall point motion you suggest, you should keep an open mind until you find the source. But you should definitely look to see if it is the stall point motion.

I would have thought that to get the speed of the stall point progression you would need to model at least two passages (and preferably more), and then measure the time it takes to progress from one passage to the next. Then multiply by the number of total blade passages and you have the time to travel around the circumference.

[Julian121]

Originally Posted by ghorrocks

To see what is causing the 480Hz peak you have to save lots of transient results files and look at them trying to pick some disturbance which is 480Hz. It may or may not be the stall point motion you suggest, you should keep an open mind until you find the source. But you should definitely look to see if it is the stall point motion.

I’ve looked at the results at different time steps, but It seems that the low momentum region does not move around the annulus at all maybe because this plane is located insdie the passage or maybe because I have modelled just one passage. I can see the size of the low momentum region just increases and decreases at different time steps.
Assuming this low momentum region causes a disturbance that has some frequency, is it possible to approximate the frequency or rotational speed of this recirculation zone from velocity vectors?
Is it correct to approximate the rotational speed assuming the recirculation has a circular motion?


I’ve tried to create an animation showing the velocity vectors at the same turbo surface at different time steps but it was not successful.

[ghorrocks]

Have a think about what you are actually modelling here. When you model a single blade passage you make the implicit assumption that all passages are the same as the modelled one. This means by definition your single passage model cannot predict the separation travelling around the circumference as that would require the passages to be different (ie some with a recirculation and some without).

That is why I said a few posts ago that you will not be able to predict the motion of this recirculation with a single blade passage and you will need a full rotor model.

[Julian121]

Originally Posted by ghorrocks

Have a think about what you are actually modelling here. When you model a single blade passage you make the implicit assumption that all passages are the same as the modelled one. This means by definition your single passage model cannot predict the separation travelling around the circumference as that would require the passages to be different (ie some with a recirculation and some without).

That is why I said a few posts ago that you will not be able to predict the motion of this recirculation with a single blade passage and you will need a full rotor model.

Can't the low momentum region (the recirculation zone) induce a frequency on its own even though one passage cannot model the movement of stall cell around the circumference?


I've compared all frequencies at the operating points from choking to stall points. This frequency (480 Hz) just shows up at near stall so I think it is somehow connected to stall point.

[ghorrocks]

Can't the low momentum region (the recirculation zone) induce a frequency on its own even though one passage cannot model the movement of stall cell around the circumference?

Yes, it can. But it is likely to be wrong as modelling only a single passage artificially constrains the system. When the artificial constraints are removed then it is likely the simulation will result in a different frequency.

I've compared all frequencies at the operating points from choking to stall points. This frequency (480 Hz) just shows up at near stall so I think it is somehow connected to stall point.

Sure, it will be related. It shows in an artificially constrained system the frequency is 480Hz. I bet it won't be 480Hz if you model the same configuration but a whole rotor.

[Julian121]

Originally Posted by ghorrocks

Yes, it can. But it is likely to be wrong as modelling only a single passage artificially constrains the system. When the artificial constraints are removed then it is likely the simulation will result in a different frequency. Sure, it will be related. It shows in an artificially constrained system the frequency is 480Hz. I bet it won't be 480Hz if you model the same configuration but a whole rotor.

Totally agreed with the frequency idea that might change if the half or whole annulus is modelled. In fact, according to the literature rotating stall frequency is around 100 Hz.

Now I understand why the frequency of rotating stall cannot be predicted in this way.

Sorry to bother you again but now that I know the frequency at 480 Hz might be caused by the low momentum region (recirculation zone), how can I confirm it?

[ghorrocks]

It can be difficult to spot what is actually causing the spike in your FFT at 480Hz. All I can suggest is to record an animation of your results and try to pick a flow which is near that frequency.