Have been watching a trend in vibration. This is a large vertical motor and pump. The motor sits about 6 foot off the floor on top of the upper pump casing.

three frequencies have been trending up. Running speed and two odd multiples (34.19X and 36.19x) Looks like bearing to me.

I wanted to verify that it was motor so I did an uncouple motor run last night. Motor speed went from 1188 to 1199 without any load as expected.

What was unexpected is that the odd multiples moved down in frequency instead of up with motor speed. The 34.19x went to 33.6x and the 36.19x went to 35.6x. So they stayed 2x apart.

Just wondering on everyones thought on why these frequencies went down instead of up. My only thought is that the reduced weight of not having the pump attached is the cause.

Hi Ron,

If there bearings are angular contact bearings, and they are not preloaded, the uncoupling could change the contact angle, I suppose (guess ). MBTBS! the bearings may also skid instead of roll with no load.

Any spectrums/waveforms?

Good luck!

This message has been edited. Last edited by: Bill Kilbey,

Here Bill. Not much in my normal spectrum. There is some more bearing noise out higher.

SCondPmp.doc (68 Kb, 64 downloads) Vibration trend

I'm certainly no motor expert but doing the math that comes out real close to 2400 cpm spacing. I think that might signal the loss of a phase. I would have to ask a motor guy for sure, though.

I left my glasses at the house so I hope I read everything right.

I figured 35.6021-33.6009=2.0012 * 1199=2399.4388

Can you zoom in on the high frequency area and see if there is really a sideband on the other side of the higher amplitude peak? Maybe in log scale? Is the resolution good enough to tell if it is really 2 x running speed and not line frequency related? It could be a natural frequency being rung at 2 x running speed. Got any twf's?

This message has been edited. Last edited by: Danny Harvey,

The two odd peaks are spaced 2 times running speed apart. That was true at 1188 RPM and with the motor uncoupled at 1199 RPM.

Do you have the bearing fault frequencies or the bearing numbers? What do the waveforms look like?

Ron,
A little help, please. How about the data from the other directions and locations and time waveform.
The higher harmonics have pretty much grown along with the 1x, so chicken or the egg, which is coming along for the ride?
They definitely are in the bearing region. Do you have the bearing numbers?
Uncoupled, the 1x is still higher than your baseline, so something in the motor has changed. The spectrum also shows that you are operating just below the reed frequency (natural frequency of the motor/pump), so your increase in 1x could also be tied to a decrease in stiffness that moved the natural frequency lower. Any history on where this natural frequency was several years ago? A plot from earlier in your history, with a log y scale will show it clearly.

Ron,

The reed frequency has always been at the same spot just popping up in amplitude a bit. That is why I wanted to check the motor uncoupled. Most times a pump problem will cause the reed frequency to increase in amplitude.

I do not have the bearing numbers for this motor yet. But I believe that the top bearing is a Kingbury with a sleeve and the bottom is a rolling element bearing. So this is most likely the bottom bearing.

My main question was if anyone had ever seen bearing frequencies shift to the left as motor speed goes up. Still has to be something like what Bill said about the contact angle changing with the lighter load.

SCondPmp1.doc (158 Kb, 29 downloads) More data

Ron,
It appears that the higher frequency phenomena is a harmonic of something in the 400 Hz range. They also appear to be related to the sixth harmonic of running speed. You time waveform doesn't have any indications of impacting, but it does look like energy from a VFD. Is this motor on a drive?
The only time I have had energy increase in frequency as the motor slowed or drop and the motor sped up involved an eddy current drive, for obvious readings. You don't mention what drives this motor.

Can you show the spectrum on a log scale?

It may show a larger family of peaks and make it easier to recognize the overall pattern which gives these two peask.

How could these two peaks be bearing fault related if they are spaced at EXACTLY 2xRPM interval. May be a better resolution will disprove that?

Could be upper and lower running speed sidebands around a bearing fault frequency harmonic... and for some reason the sidebands are much bigger than the fault frequency harmonic.

[More general discussion about change of frequency with change in contact angle:

• Contact angle is 0 for pure radial contact, 90 for pure axial contact.
  
• With decrease in axial load during uncoupled run, contact angle theta would be expected to decrease also (act more like a radial bearing).
  
• The fault frequency equations include cos (theta).
  
• When theta goes up, cos(theta) goes down.
  
• FTF, BPFO, BSF all have terms involving negative cos theta. Therefore, these frequencies should all go up during uncoupled condition (if the change were due to contact angle).
  
• BPFI has a positive cos theta and would be expected to decrease during uncoupled condition (if the change were due to contact angle)
  
• Skidding of rolling elements during uncoupled run might lower some of the fault frequencies.
  

If we assume your vibration is caused by bearing fault frequencies (a big assumption that remains to be seen), then since you observed a decrease in frequencies, it would appear either you are looking at inner race frequency or skidding causing fault frequencies to decrease.

I guess you don't have the bearing part number?

Honestly, I was going to ammend my previous reply to make another suggestion but found that Pete has already posted identical one regarding two 1xRPM sidebands around some small magnitude frequency. Truly independent opinion, .... must by correct .

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

If this motor has a Kingsbury thrust bearing, why would it have an angular contact guide bearing?

Coincidental that you asked about motor with Kingsbury bearing and angle contact bearing. I'm sitting in a hotel room near the repair shop for teardown of the motor with bad bearing discussed in the other thread. So I have studied the motor drawings (for my motor, not Ron's). It is a vertical power plant cw motor. On top it has Kingsbury thrust bearing and tilted pad radial bearing. On bottom it has an angle-contact bearing. The top thrust bearing is a single-acting Kingsbury and can't handle upthrust, so any momentary upthrust is taken by the bottom bearing. An angle contact bearing gives greater momentary upthrust capability than a deep groove which is probably why the manufacturer selected it. Note there are also a preload washer in the lower bearing housing to keep that angle contact bearing loaded when there is no upthrust (which is the vast majority of the time).

And as far as the discussion of contact angle, it applies equally to angle contact bearings as to simple deep groove / conrad bearings (deep groove bearings change contact angle when exposed to axial load also).

** BUT after saying all that, I have to acknowledge that you raise a very good point. If the motor had a Kingsbury thrust bearing, then rolling bearings should not be carrying any thrust at any time other than momentary upthrust. So from that standpoint, it seems there should be no change in contact angle of any rolling beairng due to uncoupling, if the motor does in fact have a Kingsbury thrust bearing.

Sorry it has taken a few days to get back. Doing some training off site.

It is not a Kingsbury. From the data I have, if it is correct, the upper bearing is a FAG 7334 and the bottom is a FAG 6222.

I have attached the hi-res spectrum in log scale. It does show a center peak. The two peaks that have popped up are running speed side bands.

From what you have calculated electricpete I guess this would point to an inner race fault. I will post the bearing numbers when I get them.

This message has been edited. Last edited by: Ron Stiemsma,

SCondPmp2.doc (58 Kb, 18 downloads)