We just put in service a CRD (boiler feed pump for non-nukes)pump with a speed increaser with a different set of gears. We have found a new peak in the vibration spectrum that we suspect is the gear natural frequency. Our concern is that we haven't seen this before and it shows up on both the motor and pump shaft. Were trying to get a handle on if this is a problem what is the severity.

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

Gearbox_Motor_Shaft_Bearing.doc (50 Kb, 29 downloads)

rgf,

It looks like 2616 HZ peak is 1xGMF of the input pinion gear mesh (1784*88/60 = 2616 HZ) and 1308 HZ is 1half of it, so probably there is a common factor in # of teeth and the above pattern is an indication of a problem.

The 2080 HZ peak could potentially be first harmonic of another GMF in this box, so I suggest cranking up Fmax to see the rest. Look also at the sidebands around these peaks and try to identify the offending shafts.

I have a problem understanding your comment "The peak at 1308 Hz is I suspect the gear natural frequency. The question is why would this dominate the spectrum?" It doesn't dominate the spectrum, its amplitude being approximately 0.01 ips (my estimate). Also in the spectrum is a 2080 Hz component (ghost frequency??) at 0.055 ips (again my estimate) and what someone else pointed out is mesh at 2613 Hz at 0.025 ips.

Certainly the 1308 Hx could be (as someone else pointed out) 1/2 of mesh if the pinion also has its number of teeth devisible by 2. Some Maag ground gears will also exhibit a 1/2X, usually very low as is the case here.

Post the exact tooth combination or ratio and perhaps we can offer some other suggestions.

I have revised the attachment. The frequency of interest is the peak @ 2080 Hz. The 2616 is the gear mesh of the input shaft. Sorry for the confusion. rgf

Gearbox_Motor_Shaft_Bearing.doc (50 Kb, 20 downloads)

HI rgf,
My first impression of the two suspected frequencies (356.77Hz = 12 x TS and 2080.95Hz = 70 X TS) are not gear resonance since they are both synchronous with the input shaft speed. The frequency at 1308.04 is ½ TS that could be normal to a certain extend. I would like to know the type of gear. Herringbone, single or double helical or spur. Most gear will show better in axial direction. I thing it would be interesting to get a synchronous time averaging waveform at the speed of the input shaft. The equivalent of four rotations (4 rotation X 60 sec / 1784 rpm = 135 millisecond.) at maximum resolution in acceleration would give you a great waveform to identify machining flaws. Of course, you will have to respect the sensor polarity.

Good luck, Marcel

This is most likely a "ghost frequency" and based on the info provided it is most likely on the gear, not the pinion.

Note that the frequency of 2080.95 is virtually an exact integer of gear rotational, 69.987X; it is also 29.427X of the pinion rotational. To identify possitively which rotor has the error do a time synchronous average of the casing signal using a keyphasor on the gear. Then repeat (don't move the transducer) using a keyphasor generated from the pinion. On the rotor that has the error the 2080 Hx component will remain since it is synchronous; it will average to "zero" and thus drop out of the spectrum on the rotor that doesn't have the error.

"Ghost frequencies" are introduced in the manufacturing process. It takes gears to make gears and if something in the manufacturing process had 70 teeth and an error then the ghost frequency of 70 x final operating speed can show up in the spectrum.

As to good or bad. 0.06 ips at 2100 Hz is about 2 g's. That in itself is OK except something like API says this type component should be less than mesh. However, my feeling is it is OK since where the energy comes from is somewhat less important than the amplitude. Having said that you need to be careful before answering the question good or bad. Your units are ips; many velocity coils are borderline useless at 2 KHz since they will suffer from a high degree of attentuation. You should be using a properly mounted accelerometer although something like a velomitor with respeonse up to that area of interest would be acceptable.

anticipated_frequencies_88_x_37.doc (27 Kb, 19 downloads) Anticipated frequencies

rgf

I got the impression that your tooth combination was 88/37. Is that correct? If so then the common factor is not a concern.

Other causes of fractional harmonics of gearmesh frequency, include pitchline eccentricity, uneven face width, and tooth transition problems. To more accurately define the problem is beyond the capability of a data collector/analyzer. (At least mine).

How big a change does this represent from the pre-rebuild condition? Did the machine shop do any bump tests during the rebuild? (yeah, right) Any other tests performed on the teeth?

I would definitely do like David says and look at the sidebands around all the frequencies of concern. That will point you to one or the other shaft.

I can't say how serious a problem you have, but there is definitely cause for concern there.

Good Luck,