I am evaluating a strange case for a customer involving a rotor with different phase angles at resonance depending on whether the rotor is starting up or coasting down. Please bear with me as I don't want to provide too many details to protect the client's confidentiality.

All startup and coastdown data was taken with proximity probes mounted through the bearing housing, which is holding down the bearing upper half. Shaft stick and bearing housing measurements confirm that the bearing and shaft are moving in phase at operating speed.

The subject rotor has dramatically different phase angles at a resonant frequency during startups and coastdowns. I am unsure as to whether I am seeing a balance resonance (rotor first bending mode) or a structural resonance (bearing pedestal). Regardless of which it is, the phase angle at the resonance is at the same speed whether starting up or coasting down, but the phase angle is significantly different. For example, on the way up the phase at the resonance is usually around 200 degrees, and on the way down the phase at resonance is usually around 340 degrees. The difference is consistently 100 to 180 degrees and the startup vs. coastdown angles are fairly consistent.

The rotor is solid for the most part, but there are a couple of shrunk on parts which could be shifting. This was what I initially suspected, but now I'm not so sure. I also considered a coupling problem, but the coupling is rigid with minimal possibilities of looseness due to fitted bolts.

The rotor does not experience any appreciable thermal transients. Even when the rotor is started up and coasted down with no load at all, the phase angle shifts in the same manner.

One signficant difference between a startup and a coastdown is the time it takes for the rotor to pass through the resonance. On startup the time through the resonance is about 30 seconds, and on a coastdown the time is about 3 minutes.

We are planning some bump tests to attempt to determine whether the resonance is structural or rotor bending. We are also attempting to obtain information from the designer as to where the rotor criticals are supposed to be.

Has anyone had similar experiences? Any ideas as to what could be causing the phase to be different?

Michael Titone

I have seen this effect before, but I was more concerned with vibration amplitude at critical speed and not phase angle deviations between startup and coastdown.

Is the peak 1xSS amplitiude the same for both startup and coast down? If not, then this would also be an indication that damping is different.

I'll assume that the "rotor" is either a steam or gas turbine. In either case there would be steam/gas in the interstage gland seals that could provide system damping during startup, but would not be present during coast down. I assume that a change in damping could change phase angle without affecting natural frequency.

Michael,

My wild guess would be that this has to do with the way data is acquired on start up vs. coast down. Without going into much details, speed changes faster on start up then that on coast down and therefore some speed/phase points will be missing. Setting up more frequent data acquisition may lower the error. Consistency in the data does not suggest looseness.

This message has been edited. Last edited by: David_G,

Is this happening at every bearing? or at just one bearing location? Are you slow roll compensating the data? Does it still do if you don't compensate? Is the machine on turning gear between starts? Couple possibilities:
Slow roll runout level is changing between startup and shut down. Maybe due to a bow developing either while the machine is down, or due to a rub when the machine is running. Another possiblity is that the rotor thrust position is signifcantly different between startup and shutdown, thus the probes are looking at a different part of the shaft and a different runout pattern. There may also be some torsional wrap up on startup that is not present on shutdown, if this is like a long 10 bearing main power station turbine generator. If it is a gas turbine, then all sorts of funky stuff happens on startup including rabbit fits adjusting, thermal growth of disks and blades, bearing elevations changing, etc. Finally, cold starts are usually a bit funky because the bearing elevation offsets artifically impose a preload that disappears when the machine is up and has some heat in it.

Finally Dave is right, could be some data acq issues as you quickly go through resonance, is the vibration really high at the critical? Can you go through a bit slower?

In general I prefer the shutdown data for analysis purposes.

The conditions during start-up and coast-down will be different. For example, while there is torsion from the driver-to-driven during start-up, this is not so during coast-down. The process also reverses for oil pressure (main oil pump vs. aux. oil pump), rotor heating/cooling and process fluid conditions (steam/gas pressure). If the source of your resosnance problem is linked to one of these then it can further affect the force vector(phase angle).

There are only oil seals on the rotor with the problem. The amplitude of the peak is about the same if the rotor is just run up and then down. However, if the machine is allowed to reach thermal equilibrium, then the amplitude is much higher on the coastdown. This implies to me that the vibes on the adjacent rotor (which do change due to load) are exciting the resonance, not the vibes on the rotor where we are measuring the highest vibes. In other words, we might have a structural resonance on bearing pedestal X being excited by vibration from rotor Y (or Z or Q for that matter).

Good point Dave. Data acquistion is based on sampling based on changes in speed. So the number of samples is the same on startup and coastdown. However, we should have collected more samples to get better "resolution" on the plots. Had we collected more samples, we might have a clearer picture on what excactly was happening.

Note that the resonance in question is not well damped, indicating that it might be structural. One question I still have is whether structural resonances in general have consistent phase angles at their peaks regardless of whether they are seen on a startup or a coastdown. Has anyone tested a structural resonance with an exciter (as opposed to a hammer) and evaluated the phase angle at the peak based on increasing or decreasing the exciting frequency?

I don't have much worthwhile actual experience on this stuff, but here's a "fundamentals" point that might help a bit.
Don't forget that both the startup and rundown are transients (ie the system doesn't get to the steady state condition). So if the transients are different, the whole response, including phase angles, could be different.A lightly damped system will require more cycles, and thus more time, to reach a steady state. The rundown has more time (ie cycles) to get closer to steady state than the runup.

Sounds like you got some good practical input from people who know about turbines. I'm not one of them, but I'll throw my wild-guess in for fun.

You mention possible structural resonance, influence of another rotor, Steve mentions torque... put them all together and I can invent a theory:

The theory is that structural resonance combined with anything that changes rotor mode shape would give the behavior you describe. Because the structural resonance frequency isn't going to change. But change in rotor mode shape would change the phase. (I suspect that's the reason you're looking at structural resonance... so it's not really a new theory)

All that's left is to come up with stuff that changes rotor mode shape. I think Steve mentioned a few.

One that I picked up on was torque present during run-up. I would think a coupling transmitting torque can be much stiffer than an unloaded coupling. (In fact Gerald D'ans published info on that for tiny couplings used on tiny machines... varied with coupling style.). Stiffer coupling changes the boundary condition for the rotor and changes the mode shape resulting in the phase change. But still same structural resonant frequency giving magnitude peak at same frequency on way up and way down.

I think the coupling stiffening under load applies to flex coupling and possibly gear couplings, but not rigid couplings.

This message has been edited. Last edited by: electricpete,

So have you tried running it up and not doing anything with the generator, and then run down.
Depending on the design, even just flashing the field could induce a rotor bow in the generator under the right conditions.

With an exciter test you get the same results on the way up in frequency versus the way back down. An exciter is really no different than using the machine rotor for excitation, so for a structural resonance you should get the same results both ways, unless something is changing in between.

Yes. The phase behavior (120-180 deg different up vs. down) is the same when the unit is run up and back down (warm) without flashing the field. The difference is that the amplitude is much larger when the machine is put on line and allowed to heat up. Since the problem bearing does not support a rotor that undergoes an appreciable thermal transient, this leads me to believe that it is being influenced by a rotor in the train that does, probably the next one over. If we do have a shifting loose part, it may be located on the adjacent rotor as well.

Runout is only about 0.5 mils or less on all points in question, and amplitudes at the resonance are anywhere from 5.0 to 15.0+ mils at resonance.

There is a significant 2X component at full speed, but I see that often on similar machines and haven't been too concerned about it, and both it and the 1X component are low (~2 mils).

A co-worker is performing bump and/or modal tests today to attempt to determine whether the pedestal and support structure or the rotor is resonant.

I like the sound of the idea about it requiring more samples to reach steady state. However, I really look at a lot of these machines, and I have never seen one that acted quite like this. A difference in amplitude and a small (25 to 45 degree) change in phase is common, but this one is the largest difference I have ever seen. I also see quite a few that rub thru a critical and have inconsistent phase and amplitudes, but they don't tend to be repeatable over multiple starts and stops. The machine is question has been run up and down 20+ times due to normal operation and some attempted balancing (not by me) and the behavior is very consistent.

This would make a good case study. Maybe after the dust settles, I will write it up (like I'm really going to take the time to do that).

Thanks to everyone for all the input so far. You have confirmed many of my ideas and given me some new ones to consider.

You wrote:
“The rotor is solid for the most part, but there are a couple of shrunk on parts which could be shifting.
This was what I initially suspected, but now I'm not so sure.”

I agree with this remark.

I had two cases where shrunk-fitted parts caused vibrations at resonances, which would differ between start-up and coast-downs.

One almost delayed the commissioning of a 1000 MW nuclear unit in the eighties. All extraction pumps for the condenser were each driven by a 4-poles 4.5 MW vertical induction motor. For these powers, at least one bearing must be insulated to prevent stray current to flow though the antifriction bearings and cause electro-corrosion. The manufacturer had placed the insulation between the inner ring of the upper conical bearing and the rotor shaft. Problem was that ageing (polymer curing) caused the shrink fit to get loose. The shaft would then run eccentrically with respect to the inner ring and cause a huge unbalance. The eccentricity would change when one crossed the critical speed. Amplitudes at the resonances would thus be different from start-ups to coast-downs. At nominal speed we could observe huge variation of vibrations when we bumped the motor with a heavy beams. The solution was to transfer the insulation to the stationary parts instead of sandwiching it between inner ring and shaft

A second example was with Westinghouse brushless exciters fitted with diode wheels on 300 MW units. When crossing the 1800 rpm critical speed from below we got one value for max vibrations, which differed from those, obtained during subsequent coast-downs. Again diode wheels were shrink-fitted with insulation material whose thickness may vary in time due to aging.

When crossing critical speed, the centrifugal force applied by the shrink-fitted parts to the shaft undergo a sharp phase variation and thus tend to push these parts to a new position causing a variation of the unbalance, in case of looseness.

Speed gradient are known to shift the observed resonance frequencies, the max vibration levels and their phase. If amplitudes are similar in both cases, the phase shift should be quite small. If damping were very small, amplitude during fast start-ups would be markedly smaller than during coast-downs

Is this a heavy mass rotor?. Is this a not enought stiff shaft? , maybe designed in this way due to high shaft speed. Those two conditions help to have a throught pass critical speed rotor.
Had you recorded start up?. If you observe amplitude back jump increase, you can think you have torsional whirl during start.
Answer to this question maybe help you to get a feeling on your thought.

Regard
Mike

Of course, the rate of acceleration (up or down) will affect the amplitude and phase of the vibration, but I don’t think that has much to do with this case. Most likely some of the forces change from the start to the shutdown.

I would review the polar plots from all bearings and each sensor. Lay the plots out together to see what the vector changes in the modes are. Changes to the modes have different effects at different locations, but this can often help to understand why some locations see apparently larger changes.

I would also compare slowroll data between starts and shutdowns over various runs. Changes in the slowroll generally reflect changes in mass centerline. Slowroll changes most often result from thermal effects.

Also, review shaft centerline plots. Unusual alignment changes can cause rubs or indicate rubs/significant preloads.

Are you considering a resonance to be at a fixed speed? View the polar plots to look for the telltale modal circles as the indication of the resonances. Resonance location can shift due to changes in alignment and the subsequent change in bearing loading in some machine trains (Someone mentioned that this had a generator. Bearing loading in large turbine generators can affect the resonances, even the stability of the system.).