Rotating frequency vibration peaks on the bearing housing of a centrifugal compressor (Roots-Dresser, 12840 rpm, ~670 kW):

DE V 0,62 mm/s
DE H 0,62 mm/s
NDE V 0,49 mm/s
NDE H 0,22 mm/s
Ax 0,23 mm/s

Clear spectra, very low harmonics, higher overall (5-1000 Hz) is below 1 mm/s. GMF is above 1k low amplitude.

Can shaft vibration be high as to shutdown the machine through the Bently system installed with displacement sensors? Could it be a gap calibration problem?

Housing seems to be not very rigidly mounted to basement.

Yes, the shaft vibration is somewhat independant of the casing readings. It is possible to have high shaft vibration and "reasonable" casing readings.

As long as the transducer is gapped within the linear range, there should be no effect on vibration as a function of gap voltage. If the monitor is OK (green OK lights lit) this is not a problem.

Another possiblity is that there is scratches, nicks, magnetic spots or chrome on the shaft in the probe track areas causing an error in the reading. As a result the shaft vibration would read higher than it really is.

Depending on the age, your prox probe ckt calibration may be off and causing a bad reading. Do a quick cal check to verify.

I have found this on several occasions and have had to change out the driver.

Anything in the trip ckt should be calibrated or at least checked periodically.

If there was a suitable location, I would measure shaft 1X vibration and phase with a shaft stick and then subtract the casing motion from the prox probe readings (vectorially). This should match well with your prox probe readings. If the numbers don't match then you probably have something wrong with the prox probe reading, and if the prox probe reading is high it is probably something Steve mentioned, like scratches or chrome. I do this routinely on 3600 rpm turbines, and I have done it a few times on 5000+ rpm feed pumps, but I'm not sure how practical it is on a 12,000 rpm compressor.

I would expect the prox probes to read low if the system was out of calibration. Most probe/cable/driver failures result in low sensitivities (<190 mV/mil, pardon my units). I don't know if this is the only failure mode.

You could also have a monitor calibration problem. You could do a quick check on this by taking a reading directly from the driver or from BNC signal outs if available.

Michael Titone

Does the Bently give you any recorded data.

On some of our machines we have H and V magnitudes at all bearings plus a thrust reading and bearing temperatures. If we see several parameters trending together we tend to think it's real and only one parameter changing we tend to think it may be instrumetation.

Can shaft vibration be high as to shutdown the machine through the Bently system installed with displacement sensors?

Yes. Get some good data and have some one look at it.

Could it be a gap calibration problem?

Improper gap would generally cause lower than actual readings. Wrong calibration is possible for a number of reasons including the oscillator-demodulator and its match up with the probe and extension cable/connectors.

Other calibration type issues are also possible. If you come down check the instrumentation.

Noise other than shaft runout as well as runout is a possible cause.

Get the data and have it analyzed. Don’t discount real vibration and have a wreck.

Thanks for all replies!

So, this machine is just after maintenance and we was asked to balance it. The compressor has two disks on two planes of the shaft. We've balanced it with 5,3 g, which is 0,7 g more than ideal computed value (special balancing masses, there was no way to go closer). Highest bearing vibration went from 1,5 to 0,6. Periodical offline measurements of the plant (DLI watchman + expert system) shows better result than ever.

Online Bently system shuts down the machine. It would be very difficult to get any printed data.

How can chrome go to shaft? How can shaft be magnetized?

Is it a balance problem? Does your data support balancing to solve this?

How did you balance teh compressor? Low speed? Component stack balance? If an insitu balance was performed did you use the shaft relative probes in addition to the casing transducers?

If the shaft material can be magnetized, it may become so through machining/mechanical operations or by passing currents through the shaft. Chrome could be added during a repair; it would not be a contaminant.

You need to know the symptoms in more detail than just high vibration. You need to get the data from the shaft relative transducers. There are probably a few consultants on the board who would be willing to assist.

Your casing values are very low, yet your shaft probes apparently are high. Did you verify the casing measurements? For example, does the accelerometer sensitivity match the analyzer setting? A factor of 10 lower values is a common mistake, if a 10 mv/g acclerometer was used with a 100 mv/g analyzer setting.

Did you measure the AC output of the Bently system with your analyzer? This would cross-check the Bently electronics and provide spectrum and phase data for diagnostics. Did you measure the DC output from the Bently system with a volt meter? This would provide an independent measure of probe gap.

I am simply suggesting that all data be verified before trying to make a fault diagnosis. Beware of the Morton Effect for high speed compressors with high 1xSS vibrations!

It was in situ two plane balancing. There was no way to use shaft relative probes. 100 mv/g ICP piezo accelerometers were used. Measurements was taken with two independent analyzers with the same phase signal. Full balancing program was done with each analyzers, but not the same measurement directions. Phase data was very stable, and was the near the same in the two planes (force unbalance). During every trim the machine became better and better. There was no sensitivity mistake. Residual unbalance is less than 1 gram. Balancing radius is 4 cm! Very low unbalance, I think. Anyway the machine "answers" were very positive, so we could balance it more accurately but it needs to fabricate masses.

Thanks for useful advices! I've to get additional info to verify mentioned possibilities.

Libra,
Think the previous replies pretty much covered it.

Balancing the compressor using casing transducers dos not seem to me as a good way of dealing with trips on the shaft probes.

To understand the cause for the apparently high shaft vibrations, you'd need startup/shutdown data from the rack buffered outputs, preferably recorded with a multi-channel instrument so you can look at all probes. If such is not available you can use an analyzer which can do run-down. The overall & 1X amplitude/phase vs speed data should indicate if you have imbalance and/or excess runout/glitch.

If you can’t get rundown data you can still check if there is excess runout/glitch by looking at the vibrations at slow speed (e.g. 200-500 revs). If you have significant vibrations at slow speed it is most likely caused by runout/glitch (scratches, mechanical/electrical runout, etc as described in other posts) and this will affect the vibrations at full speed. Bently used to have pretty good info on runout/glitch on their website. API617 specifies the allowable runout – generally less than 7 microns.

If you are not equipped to deal with this kind of machinery/problem better get someone experienced to look at it.

A Bently Nevada application note on glitch:

http://www.bently.com/articles/apnotes/an002.asp