Sign up Calendar Latest Topics Donate
 
 
 


Reply
  Author   Comment   Page 1 of 2      1   2   Next
Curran919

Avatar / Picture

Sr. Member / Supporter
Registered:
Posts: 476
Reply with quote  #1 

I have a simple pumpset with 4 orbit planes on pump/motor DE/NDE. Pure sleeve bearings. We have good rundown data (3600-0rpm in ~80s). We get a dozen rotations at <60 rpm so have a good runout average. However, the runout from the two probes in one plane does not match, and I can't figure out how.

For the motor, they behave as expected. Very similar waveforms that are 90 degrees out of phase. However, the pump planes behave strangely. At first, they appear to be 180 degrees out of phase. However, looking at the waveform shape, they instead act as if one is polarity-inverted, which would then make it in-phase. Either way, it is not 90 degrees. Waveforms:

MotorRunout.png  PumpRunout.png  

If we filter a valid runout orbit (like the motor) to any frequency component and plot the orbit, we should get a perfect circle. However, we instead get a straight line (horizontal line for 1x component).

There are four pumpsets, but each is running independently, so there is no cross-wiring between pumps.
Assuming the two pump orbits are cross wired doesn't help.
This effect is visible on all 4 pumps, to slightly different degrees. One one of the four, you can kind of start to see the waveform shape repeated at 90 degrees, and the 1x orbit is more circular.

The only explanation I can come up with is that the probes line up on the bearing housing but are so cock-eyed that they have mutually exclusive probe tracks. The mounting hole wasn't too far from the shaft though, so I'm not sure if we can really get the probe tracks over a cm from eachother.

Any other ideas? The rest of the data is coming out kinda weird as well, so I'd like to chalk it all up to some measurement chain error and just retake some data...


George D

Member
Registered:
Posts: 36
Reply with quote  #2 
Different probe tracks make sense to me.  If I squint, I can even make sense that there are positive peaks roughly 90 deg apart... although not shaped alike at all?

Side interference may play into this as well?  We looked into the effect of side-interference a few months ago when we discovered a probe in one of our feed pumps was touching a stationary oil shield.  Below is a vendor plot describing the error if you have side-interference of as close as 0.5 inch for a 5 mm probe.  It appears that for a normally gapped probe, side-interference will make the probe read less than actual.  Now... if that side-interference is actually vibrating as well, that could throw the slow-roll indication into quite a difference state?

Prox Prob Side Interference.png 

As you discussed slow-roll compensated orbits, I wasn't sure if your discussion related to vector-compensation (only compensating the 1X vector against the slow-roll 1X vector), or waveform  compensation (compensating each waveform sample against the slow-roll waveform)?  The discusion of how a compensated orbit should produce a perfect circle was not clear to me?

Thanks for sharing...

Curran919

Avatar / Picture

Sr. Member / Supporter
Registered:
Posts: 476
Reply with quote  #3 
Yeah, if you squint, you can start to see some small artifacts that may be 90 degrees out of phase. I guess a lot of those small 'notches' in the waveform could be from axial 'assembly scratches' that you would see on both probe tracks even if they are axially offset.

For side interference to effect only one probe (not be rotating), yet have such a vibrational displacement as to have such a large periodic effect on the probe seems unlikely.

With the circular statement, I was talking about the runout itself, not a raw or compensated signal. If you plot the motor runout as an orbit, you get a symmetrical shape. If you narrowband filter that orbit shape to any freuqency, you will get a perfect circle. That can be 1x, 2x, 13x, always a circle.

The different probe tracks is the most likely culprit I think, but it doesn't address the fact that on both the NDE and DE planes, the signals look essentially inversed... If anything they should be independent signals, or show SOME indication of being 90 degree shifted.
John from PA

Sr. Member
Registered:
Posts: 815
Reply with quote  #4 
Certainly different probe tracks is the common cause, but I have to ask has the pump bearing journal ever been repaired by plating with chrome? This is a common repair method that unfortunately sometimes extends into the probe track area causing peculiar runout patterns. In addition, a relief around the probe tip is necessary. I have seen cases where a probe hole is missing relief on one of a probe pair causing differences. Lastly, the probe sees combined electrical and mechanical runout so the anticipated 90 deg relationships may not hold true in the instance of something like a work hardened area as a result of a 3-jaw chuck used for machining.
electricpete

Sr. Member
Registered:
Posts: 528
Reply with quote  #5 
I'm not much of a prox probe guy.  My thought/ question:


Is it possible that (even at this slow speed) the shaft is moving in a line (rather than an orbit) at a direction halfway between the two probes?  For example x and y probes both 45 from vertical, maybe the motion is almost pure vertical for some unknown reason.  Obviously that's not expected for runout, maybe something besides runout is influencing the motion even at that low speed.  Maybe vertical resonance?  
George D

Member
Registered:
Posts: 36
Reply with quote  #6 
OK... I’m tuned up with what Pete says...
“Maybe something besides runout is influencing the motion at that low speed.”
While many causes of vibration, such as that related to imbalance or resonance, are probably nullified at slow roll speed; shaft displacement due to other causes such as rubbing in a wear ring or throttle bushing, or a locked flexible coupling, may not be? Be curious to consider slow roll traces from previous cycles?
OLi

Sr. Member
Registered:
Posts: 1,519
Reply with quote  #7 
Maybe the sensor mounting? It is a relative sensor system. 
__________________
Good Vibrations since early 1950's, first patented vibrometer 1956 in the US.
http://www.vtab.se
Curran919

Avatar / Picture

Sr. Member / Supporter
Registered:
Posts: 476
Reply with quote  #8 
Thanks for the interest gents,

I have some more information. Below is the actual orbit of the 'runout' where the waveforms are transformed into vertical/horizontal components. It is 'unfiltered' but it is 1x-dominant, so we mostly see a 1x shuttling in the horizontal direction (sorry, image/orbits do not upload square for some reason).

In addition, you can also see that the runout does decrease in amplitude significantly between 90rpm and 30rpm. Considering first critical is far above this value, I didn't think rotordynamics would be in play even at 90rpm. Or am I wrong?

Quote:
shaft displacement due to other causes such as rubbing in a wear ring or throttle bushing, or a locked flexible coupling, may not be? Be curious to consider slow roll traces from previous cycles? 


If this 'runout' is somehow representing a real shaft signal... I'm not sure if a radial rub would cause a shape like this? There are some signs of rub in the full speed signal. The locked flexible coupling is an interesting idea. What happens with that exactly? On one of the other pumps (that is behaving 'well'), the pump-side hub actually had a clearance fit and could be moved axially by hand. It wasn't apparent in the other pumps, but is also likely somehow. We've already reordered hubs with the correct interference for all the pumps. That could very well fix all of these problems.

PumpRunoutOrbit.png 

Quote:
has the pump bearing journal ever been repaired by plating with chrome? 


The pumps are only 1 year past commissioning. I do not think they have made any journal repair (my colleague was on site). If so, you would still expect this to affect the runout like any other common type of probe track trauma, right?

Quote:
the anticipated 90 deg relationships may not hold true in the instance of something like a work hardened area as a result of a 3-jaw chuck used for machining.


If your runout was indeed dominated by a 3x signal from this, then I suppose your +90 degree lag would also appear as a -30 degree lag. But then it would also be obvious that it is a 3x signal, not the 1x that we see.

John from PA

Sr. Member
Registered:
Posts: 815
Reply with quote  #9 
https://www.maintenance.org/fileSendAction/fcType/0/fcOid/399590942964628183/filePointer/399590942964867906/fodoid/399590942964867904/3q2005_runout.pdf

OP, very well written article and definitely worth the time to read.
electricpete

Sr. Member
Registered:
Posts: 528
Reply with quote  #10 

Some rambling thoughts fwiw:

It’s not hard to imagine that at low speeds, the oil wedge loses it’s strength and gravity becomes a bigger factor.

Then maybe the shaft acts more like a marble sliding back and forth in the bottom of a greased bowl.  If the shaft were to slide back and forth along the circumference of the bearing (like a marble sliding in the bottom of a bowl), we might expect the orbit to look something like a smile.  Maybe since the motion of maybe 0.001” – 0.002” is along the bottom of a full circular orbit that might have a diameter of 0.008" (?), it's not enough for us to see the radius of curvature and it looks like a straight horizontal line.

So it seems easy to imagine the shaft oscillating in this particular orbit shape. What is not easy for me to imagine is what would excite it to oscillate at 1x in very slow speed ranges.

 

John from PA

Sr. Member
Registered:
Posts: 815
Reply with quote  #11 
What are the mounting angles of the two transducers?  Also, although it may not matter much, what instrumentation is being used to acquire/display the signals?
Curran919

Avatar / Picture

Sr. Member / Supporter
Registered:
Posts: 476
Reply with quote  #12 
Quote:
Originally Posted by John from PA
What are the mounting angles of the two transducers?  Also, although it may not matter much, what instrumentation is being used to acquire/display the signals?


Probes are at 45s (10:30 and 1:30).

The DAQ is our same proprietary Labview/NI based DAQ that we've been using for 10 years. The displays are from a Matlab library that that my predecessors programmed in 2004. Both have been continuously validated.
vogel

Sr. Member
Registered:
Posts: 75
Reply with quote  #13 
It seems to me that it is real displacement and not an issue with the measurement chain or runout related. The orbits are too similar in both bearings, I don't think that would be a result of cock-eyed probes looking randomly at different tracks.

Assuming that it is actual motion ...
I think that it is too low speed/frequency to be a resonance. Anyway, if you filter the orbits at 2X, can you still see a straight line?

A locked coupling or electricpete's marble theory look more probable to me.

It seems that in Time=550s you already have flat orbits, at what point of the shutdown the orbit starts to flatten?
John from PA

Sr. Member
Registered:
Posts: 815
Reply with quote  #14 
Now that I'm aware the probes are 45° L/R of TDC, I'm inclined to go with...

Quote:
Originally Posted by electricpete

It’s not hard to imagine that at low speeds, the oil wedge loses it’s strength and gravity becomes a bigger factor. 

Then maybe the shaft acts more like a marble sliding back and forth in the bottom of a greased bowl.  If the shaft were to slide back and forth along the circumference of the bearing (like a marble sliding in the bottom of a bowl), we might expect the orbit to look something like a smile.  Maybe since the motion of maybe 0.001” – 0.002” is along the bottom of a full circular orbit that might have a diameter of 0.008" (?), it's not enough for us to see the radius of curvature and it looks like a straight horizontal line.

I have seen this behavior on large STG's, and I think, although can't say positively, it is more common with a LOP bearing.  At speeds less than 100 RPM, it would be anticipated that we have boundry lubrication conditions so the shaft would have a climb and fall behavior within the bearing clearance.

As to 

Quote:
Originally Posted by electricpete

So it seems easy to imagine the shaft oscillating in this particular orbit shape. What is not easy for me to imagine is what would excite it to oscillate at 1x in very slow speed ranges.



It would seem to me that at such low speeds, electrical and mechanical runout would become the dominant contributing factor to the signal.  These would be repetitious to the shaft turning speed, thus a 1X portrayal of the behavior. 

I asked about instrumentation because in the products I'm most familiar with we tend to discount the shaft centerline information below 60 or 100 RPM, depending on the product.  To a lesser degree we discount orbit information as well.  If however the peculiar data can be supported by other supporting data, then it may be acted upon.  If for example, we see a 60 RPM orbit outside the bearing clearance circle, and we know that on a previous run the bearing reached 130° C (265° F) then a bearing inspection might be justified.  As an aside, this points out why one might desire a shaft centerline plot with an overlay of a properly scaled orbit positioned appropriately. 

 

Curran919

Avatar / Picture

Sr. Member / Supporter
Registered:
Posts: 476
Reply with quote  #15 
Quote:
Originally Posted by vogel
Anyway, if you filter the orbits at 2X, can you still see a straight line?


At 550s (100rpm), the 2x component is less than 1 micron (compared to 30micron for 1x), but it is also horizontal and straight.

Quote:
Originally Posted by vogel
, at what point of the shutdown the orbit starts to flatten?


It is quite elliptical (4:1) during most of the coastdown, and slowly flattens from 400-100rpm. Changes to magnitude (see below)

Quote:
Now that I'm aware the probes are 45° L/R of TDC, I'm inclined to go with...


To be clear: the first waveforms are of the raw probe signals. The orbits I posted later (and associated waveforms) are corrected to be true vertical/horizontal.

Unfortunately, no centerline plots, as the modules we were pulling signals from were pre-filtered.

Since we are moving in the direction that this is a real signal, its probably pertinent to share some operating data. I should have done this before. Here is the spectrogram of the coast down. Lower auxiliary plot shows the tracked 1x component. Rightside auxiliary graph shows a peakhold. During operation, 1x (60Hz) is actually quite low with a 2.5Hz dominant component. That low frequency component alternates between full clearance whirling and similar 'marble' movement. During coastdown, the 1x component peaks at the same ~2.5Hz.

  PumpSpecgram.png   

Previous Topic | Next Topic
Print
Reply

Quick Navigation:

Easily create a Forum Website with Website Toolbox.