Critical Speed Indications

[jacquelinev]

Does a drastic phase angle change always indicate a critical speed? (but vibration amplitude stays relatively constant during a decel)

 

[BobM3]

Are you sure your plot isn't just "rapping around", say from +180 degrees to -180 degrees? My experience is that a rapid phase change would indicate a resonance and there would be a noticeable amplitude change.

 

[TPL]

'Wrap around' can also take place between 0 deg and 360 deg.

Phase angle is the timing between two events, either between a trigger and a filtered waveform (absolute phase), or two filtered waveforms (relative phase). If one of the waveforms is of very low amplitude, then the measuring instrument might not be able to measure phase reliably, giving rise to varying phase angles - to overcome this, change the amplitude input range of the measuring device to soemthing more suitable (usually lower input range).

How long does it take for your machine to shutdown? If it is really fast(e.g. motor driven pump where the pump acts as an effective brake), then another possiblity is that your measuring device cannot sample data fast enough to produce reliable results.

A critical speed is characterised by a peak in 1X (usually) amplitude and a phase angle change of 90 degrees

If you are taking data from proximity probes, then excessive runout or glitch can sometimesmask the amplitude changes that accompany a critical speed - you should use compensated data in this case.

 

[electricpete]

The faster you pass through resonance, the lower the peak vibration magnitude. That is not just a measurement error - it is real. There is not enough time to add enough energy to build up to the high steady state vibration level if you pass through very quickly.

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[jacquelinev]

Okay, I mispoke. Here is the plot in question. We suspect this is the 2nd critical, which is near our operating range. The lateral report predicted 2nd critical to be in the 12,000 rpm range. What other explanations woulc produce this result?

 

[electricpete]

First thought from linear systems theory is that you passed by a "zero" response instead of a "pole" response. That causes drop in magnitude and phase shift opposite of that seen in resonance. Seems like that is what you have.

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[electricpete]

My comments referred to the area around 9,000 rpm. I ASSume that is the area you are talking about.

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[electricpete]

I doubt it corresponds exactly to your machine, but the behavior you describe exactly matches a a simple 2DOF system

Ground === k1 === m1 ===k2 ===m2 <==F

If you solve the system you will find it has two poles (resonant frequencies) and two zero's. The two zero's always lie in between the two poles (for 2DOF linear system). One zero is a frequency where m1 stops moving. (magnitude passes through zero and phase flips). The other zero is where m2 stops moving. Regardless of which mass you are monitoring, if you do a frequency sweep you will see a zero response in between two poles.

Den Hartog has a good graphical representation of poles and zero's for 2DOF system similar to Mohr's circle which shows the items discussed above.

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[TPL]

Its always hard to draw conclusions from just one plot; it would be helpful to see this data plotted in (compensated)polar format, and to compare the data from the other end of the shaft.

It would be useful to compare phase relationships at each end of the shaft to estimate mode shape

It looks as if the information is obtained from proximity probes: these measure relative displacement between the probe tip and the shaft surface - fine if the probe tip is fixed but, if there is a localised casing resonance around the probe location, you can get the curve of the shape that you are showing - the probe vibration and the shaft vibration act to 'cancel each other out' giving a probe output that goes down

 

[electricpete]

TPL is probably correct this is a relative motion. The same comments as above still apply. i.e. if the system is a linear system, the monitored variable will have poles and zero's.

A lightly damped system can be approximated as an undamped system. In an undamped system the vibration vector can only have two possible orientations, 180 degrees apart. As we sweep frequency past a 0, the magnitude of vibration passes through zero and the phase flips by 180 degrees.

None of my comments will help you visualize the mode shape. I am just pointing out that this is exactly what would be expected from a linear lightly-damped system as the frequency is swept past a "zero".


=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[zekeman]

The plot makes no sense between 4Khz and 8 khz for a linear system.
If you believe the phase data , it looks like a double pole at 4Khz ( resonance) with light damping going to 180 almost to 8Khz but the problem is you cannot simultaneously show no gain change when in fact there should be a drop of 40 db/ decade.
At 9 Khz there is a definite double zero ( certainly not resonance) and the behaviour of the gain is in the right direction but strangely high.
Looks like a made up problem to me like some professor asking a PhD candidate what it means. (just kidding)

 

[GregLocock]

That looks like plausible data for a sensor location that is 'remote' from the excitation.

It looks to me as though the sesnor location/direction is possibly nodal for the second mode. You need more sensors.


Cheers

Greg Locock

SIGlease see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[jacquelinev]

zekeman, this is actual plot from a centrifugal compressor on test about 2 weeks ago. this was the first bundle tested. the second bundle, exactly the same in theory, does not show this at all.

 

[zekeman]

jacquelinev,
I'm still puzzled.
Exactly what are you testing?
Is this a transient with the compressor starting at high speed and decelerating or just starting from rest and accelerating?
You say that your second test doesn't confirm this finding, so how valid can the data be?
Can you be very specific about this test including where the sensors are located or you will get more wrong answers here and come away more confused than when you started.

 

[jacquelinev]

API 617 mechanical run test. probe is located on the drive end. compressor runs at max. continuous for 4 hours then up trip speed, then ramps down. so the plot represents vibration during its decel. the explanation we received from supplier was that this was "test conditions", but we would require more information, or explanation, since the second rotor did not experience this.

 

[electricpete]

If you believe the phase data , it looks like a double pole at 4Khz ( resonance) with light damping going to 180 almost to 8Khz but the problem is you cannot simultaneously show no gain change when in fact there should be a drop of 40 db/ decade.
At 9 Khz there is a definite double zero ( certainly not resonance)

I see phase change of 180 degrees at 4khz corresponding to single pole and phase change in opposite direction of 180 degrees at 9khz corresponding to single zero. Can you explain why you think it resembles double pole or double zero?

and the behaviour of the gain is in the right direction but strangely high.

Could be that a speed squared forcing function contributes toward this? (as frequency increases towards infinity, the response should grow without bound, unless there are two more zero's than poles in the transfer function).

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[zekeman]

Quote" I see phase change of 180 degrees at 4khz corresponding to single pole and phase change in opposite direction of 180 degrees at 9khz corresponding to single zero. Can you explain why you think it resembles double pole or double zero?"

Sorry about my double pole , double zero comment.
What I meant was a transfer function that looks like
F(s)*(s^2+ds+ a^2)/(s^2+cs+ b^2)
or one lightly damped complex pole pair near 4Khz and possibly one complex zero pair near 9Khz.
A single pole or zeros cannot produce 180 degree phase changes of the type you see here.
No mechanical system I know of can exhibit increasing gain without bound as you increase frequency toward infinity.
A possible explanation of the rapid gain rise after 9Khz might be yet another very lightly damped resonance near 10 or 11Khz which is not shown.
shown.

 

[zekeman]

"Does a drastic phase angle change always indicate a critical speed? (but vibration amplitude stays relatively constant during a decel)"
jacqueline,
Yes, the 180 phase changes coincide with resonance and antiresonance speeds and they do show correspondence with the gains.
The gain is far from constant since you are showing the log of the amplitude.
As I mentioned in my previous post,it looks like a resonance occurs a little after your plot ends. Could you shed some light on that?
Your mention that the second bundle does not exhibit these characteristics is probably due to some bearing, misalignment or other assembly problem ,which should be corrected by the mfr.

 

[TPL]

First of all, this is a lin-lin scale, with units of krpm on the x-axis and microsn pp on the y-axis - no logs or kHz are present.

The bode plots suggests that this is a reasonably well balanced rotor (13 microns at 10,000rpm) so the 1X forcing frequency excitation is likely to be quite low and, depending on the damping you might not see a classical resonance at the 2nd critical speed.

probe is located on the drive end - to drive this compressor on a test bed, some sort of slave coupling will have been employed - whilst this should be well balanced, it will not be perfect. Using the same coupling on apparently identical rotors will result in the differences seen in the two sets of results from each machine - its probably too late, but you could ask if the coupling could be rotated 180 degrees and the run repeated to see its effect on the data.

How does data from the non-drive ends of each machine compare?

The lateral report predicted 2nd critical to be in the 12,000 rpm range - that's a bit vague but supposing that the 2nd critical speed is at 12,000rpm - it is possible (again depending on the damping) that the resonant response envelope starts at around 9-10,000rpm (consider if this machine were only tested to a speed of 3000rpm - at 3000 rpm the machine vibration is clearly starting its approach to the 1st crtical speed at 3800rpm) - if this is the case then whilst you appear to be in the region of the 2nd crtical speed, I would suggest that you are far enough away for this to be no real problem.

I am guessing that this test carried out somewhere sunny on the West coast? If so, you could always try to ask for a simplifed unbalance response test by adding one small balancing bolt to the NDE balancing disc to see if this produces better exposure of any critical speeds.