Sign up Calendar Latest Topics Donate
 
 
 


Reply
  Author   Comment   Page 2 of 2      Prev   1   2
electricpete

Sr. Member
Registered:
Posts: 627
Reply with quote  #16 

I believe I might have found evidence of ball skidding at the outer race from my vibration data based on change in the fault frequencies.

See attached file:

  • Slide 1 – no noise (10:13)
  • Slide 2 – with noise (10:00)
  • Slide 3 – my tables to develop estimates of BPOR and BPIR.  There are four quadrants based on top/bottom split and left/right split.   The top corresponds to no noise condition (inputs from slide 1) and the bottom corresponds to with-noise condition (inputs from slide 2). The left corresponds to BPOR info, The right corresponds to BPIR info.  Within each quadrant are four columns: First column is what I think the peak represents (example 2*BPOR); second column is the frequency of the peak in orders;  third column is what I would estimate BPOF (or BPIR) to be based on column 2 value.  Fourth column are my comments.  The bolded values are where I think the best estimates lie.

The bin width of my spectra were 11.5 cpm (0.006 orders).  I used peak labeling which gives a much better estimate of the peak frequency than the nearest bin center (assuming there are no other peaks nearby).   

The big-picture conclusion I think we should draw from it is that the BPOR went up and the BPIR went down when the noise was present.  I claim this could be an expected result when skidding occurs between the balls and the outer race.   It corresponds to a condition when the cage is travelling FASTER than it’s predicted frequency in pure rolling contact. 

Maybe your BS detector started going off when you read “FASTER”.  If  the balls are pushing the cage, and if there is sliding so those balls don’t push as well, how can the cage possibly go faster when skidding occurs?  It’s easy to imagine cage going slower.  A little thought reveals that the cage slows down (FTF decreases) when skidding occurs between the balls and the inner ring. That result is intuitive in the reference frame that we are used to thinking in… our normal stationary reference frame.   To visualize the other case where skidding occurs between the balls and outer ring resulting in cage speeding up, we have to put ourselves in the reference frame of the shaft.  In that reference frame, it is the outer race moving and creating the cage motion… if skidding occurs between outer race and balls then the cage slows in our rotating reference frame, which means it speeds up in the normal stationary reference frame where we measure vibration.  I think it’s a little more complicated than that, including the fact that ball can slide at both surfaces simultaneously, but it seems reasonable to me.

You still may not believe me that the cage can go faster durign skidding (I wouldn't believe me either). So I’ll show you the reference that got me thinking that way. “Solving Tribology Problems in Rotating Machines”, Chapter 3 on skidding.  

The relevant starting point is equation 3.7 which defines Percentage Cage Slip as deltaFc = (Fc-Fclav)/Fc  * 100%.   From section 3.12 Fc is the theoretical cage frequency and Fclav is the (average) measured cage frequency.   That definition of cage slip implicitly defines positive cage slip to be the normal intuitive kind where cage goes slower than it’s theoretical rolling frequency and negative cage slip to the the non-intuitive kind where the cage goes faster than it’s theoretical rolling frequency.   And in fact their measurements are showing both positive and negative cage slip present, depending on the conditions.   On the very first page of the chapter, you can see that negative cage slip is predominant under moderate speed operation and positive cage slip is predominant under high speed operation.  Why would that be? High speed creates centrifugal loading which keeps better traction between the balls and the outer race, so less likely to skid at our race and more likely to skid at the inner race and slows the cage below theoretical rolling speed. I conclude that our 1800rpm machine with 6313 bearing is presumably what the authors would call low speed and slips at the outer race which speeds the cage above theoretical rolling speed.   (aside they also talk about loading - low loading can facilitate either positive or negative slip, which makes some sense because the load is transmitted through both interfaces inner-to-ball and ball-to-outer). 

The authors also talk about ball slip or roller slip. That’s not relevant to my spectrum since I have no BSF patterns… I only have BPOR and BPIR which depend on cage slip. By the way, I tried to make sense of their equations 3.1 through 3.6 and my head almost exploded.  I trust they know what they’re doing, I’m just taking their results at face value.

A less sexy interpretation of the change in fault frequencies that I saw would be that there was change in axial loading of the bearing (increased axial loading increases the contact angle from 0, which increases BPOR and decreases BPIR).   I don’t think it’s particularly likely for a horizontal motor with cross locating arrangement and no wavey washers.   There would to have been thermal expansion and one or both bearings stuck in the housing at the first measurement 10:00 just after starting.  BUT I had an opportunity to look at the replacement motor before it was installed, and remembering the unique cross locating design which has an endplay (~0.050), I did push/pull on the shaft extension to verify there was some play.... so there was no indication of binding of the replacement motor at that time.  (Although the motor was run once in between that replacement motor play check and these 10:00 / 10:12 readings, which could have changed something).       Also, when we subsequently replaced these noisy bearings, there were no anomalies in the bearing outer-diameter appearance or housing or the fit measurements.   I also monitored/recorded the bearing housing temperature with a temperature gun looking into the vent port during the run and didn’t see any unusual changes.   I cannot 100% rule out that there was some kind of axial loading due to thermal changes and bearing binding during the 10:00 reading that I was completely unaware of, but personally I'm leaning more toward the skid theory as the explanation for change in frequencies.   EDIT - Based on George's comment, I'll add a note that the thermal / axial binding scenario seems less plausible when we also consider that the altered defect frequencies corresponded with a change in audible noise and a change in spectral floor.  

I’m interested to hear any comments.  If I’m lucky, maybe I will see something distinctive once the bearings are cut apart  (but I’m not holding my breath).

 
Attached Files
pdf FreqShift3.pdf (378.33 KB, 22 views)

George D

Member
Registered:
Posts: 39
Reply with quote  #17 
Pete... People are providing valuable insight here.  You're welcoming all comments... hope I can keep up?

The main thing I see between the noisy and non-noisy spectra is an elevated noise floor in the noisy spectra?  I suspect ball skidding, in a bearing which rotates the inner ring, would result in something related to an outer-race defect frequency... as the rolling elements sequentially move into the outer race load zone, resulting in a "skidding" at BPFO frequency.  My intuition is that this could also result in an elevated noise floor? 

I've often thought that BPFO frequency, if related to skidding, would not necessarily be exactly at BPFO or one of its harmonics.  Instead, it would show evidence of a natural bearing housing frequency which modulates at BPFO?  That is, skidding is generating broad-banded noise, which excites a bearing housing natural frequency... which pulses, or modulates, as each ball enters the load zone.  In this case, the spectra would not necessarily show evidence of whole order multiples of BPFO, but would instead show evidence of BPFO side-banding around the bearing natural frequency as a result of its modulation?  In my own experience, I believe I've witnessed this.

I tried to look for BPFO sidebanding in the haystacking on the noisy spectra?  I "think" I see it?  But I may be talking myself into seeing something?

As an aside... a bearing vendor tried to calibate me by specify "ball-skidding" as a destructive event that should be avoided.  I think the thing we're looking at here is a relatively benign condition, in which balls may go from sliding friction to rolling friction as they move into a load zone?  I've long thought that this is what creates an intermitant ringing sound that operators commonly get concerned about?  I don't believe it's a "good" thing... but not, necessarily, an imminently threatening condition.  Anyway... I've changed my description of this type of event from "skidding" to "sliding".  That way bearing guys are less likely to lecture me...
electricpete

Sr. Member
Registered:
Posts: 627
Reply with quote  #18 

Thanks everyone for the comments. 

Quote:
The main thing I see between the noisy and non-noisy spectra is an elevated noise floor in the noisy spectra?
I agree.  The spectral floor of the audibly-noise spectrum is higher across the whole range of my spectrum.  It seems like that points towards sliding.   (I’m going to make an effort to use the phrase "spectral floor" instead of "noise floor" from now on based on this thread)

Also the 1st four harmonics of BPOR and the 1st harmonic of BPIR went up substantially (by a factor between 2 and 5).   That in itself is ambiguous (could be sliding or could be change in axial loading due to thermal/bound bearing I postulated before), but putting that together with the increased spectral floor and the increased audible noise seems to steer more towards the sliding scenario. 

Quote:
I've often thought that BPFO frequency, if related to skidding, would not necessarily be exactly at BPFO or one of its harmonics.  Instead, it would show evidence of a natural bearing housing frequency which modulates at BPFO?  That is, skidding is generating broad-banded noise, which excites a bearing housing natural frequency... which pulses, or modulates, as each ball enters the load zone.  In this case, the spectra would not necessarily show evidence of whole order multiples of BPFO, but would instead show evidence of BPFO side-banding around the bearing natural frequency as a result of its modulation?  In my own experience, I believe I've witnessed this.
When you say “bearing housing frequency which modulates at BPFO”, I think you are referring to a typical bearing defect pattern with a series of fault frequency harmonics (and their sidebands) that are highest in the neighborhood of the bearing housing natural frequency.  The only data I have is up to around 40 orders or 72,000cpm, which could easily be below that bearing natural frequency. So I can’t say if that pattern exists in the higher frequencies.
Quote:
I tried to look for BPFO sidebanding in the haystacking on the noisy spectra?
I don’t see anything like that.
Quote:
I think the thing we're looking at here is a relatively benign condition, in which balls may go from sliding friction to rolling friction as they move into a load zone? 
I agree it's relatively benign in the short term.  There were circumstances that drove me toward a conservative call on this one.
Quote:
As an aside... a bearing vendor tried to calibate me by specify "ball-skidding" ….I’ve changed my description of this type of event from "skidding" to "sliding".
I had noticed that the reference I linked above only refers to "skidding" in the context of "positive slip" (slip between inner ring and rollers, resulting in slowing of the cage).  The distinction between skidding and sliding is a little bit of a mystery to me. 

Sledder2

Member
Registered:
Posts: 12
Reply with quote  #19 
Quote:
 If I’m lucky, maybe I will see something distinctive once the bearings are cut apart  (but I’m not holding my breath).

You will have to look at the ball pockets of the cage for some sort of abnormality... I'm still curious what the outer race looks like.

Good write up by the way.

Thanks
electricpete

Sr. Member
Registered:
Posts: 627
Reply with quote  #20 

ok, here are the initial bearing inspection results.  I have two bearings, call them bearing A and bearing B. During bearing removal, our guys did not keep track of which bearing was inboard/outboard nor which direction they were facing.

Slide 1 - Bearing A has 8 equally spaced distinct marks on the outer ring, clearly made while stationary (there is also some grease film present in this photo). Not very unusual.  Similar pattern on bearing B is much less distinct, less noticeable.

Slides 2 and 3 – Bearing A has an area of very minor markings spread out on the race.  It has the appearance of dried lubricant, but you can’t rub it away with your finger. Nor can you even feel it with your fingernail. This location is not at the location of one of the 8 equally spaced marks.  Bearing B has an even less noticeable pattern on the outer race.. more later slide 5. 

My thought - this pattern on slides 2 and 3 might represent an area where that sliding was occurring.    More later. 

Slide 4 – Bearing A (left) inner ring has a much more noticeable ball track than bearing B (right).  It is centered axially and uniform around the bearing.  Bearing A inner ring ball track is also much more noticeable than either outer ring (not shown).  I didn't particulary expect to see this if outer ring skidding were occurring, but I might possibly talk myself into believing that sliding at one outer race one location might somehow create more wear track all along the inner race... still thinking about that. 

Slide 5 - Bearing B outer race also has a surface marking pattern at one location on the outer race similar to what bearing A had.  But the bearing B markings are smaller (more localized) and are also offset axially from the center of the race, which I can't explain.  Like the bearing A pattern, I can't rub this off, nor can I feel it with my fingernail.   I tried just as hard to photograph this as I did bearing A, but it was a lot harder to capture this pattern for whatever reason.  I'm not sure what to make of this bearing B outer race pattern. 

These bearings were cleaned with rags, not solvent, so there are more stray grease films present in the photos than normal (bottom of slide 3, slide 4).  I haven't taken apart the cage to inspect it closely. I'm not expecting much there, not sure if it's worth the work. 

Conclusion: my personal opinion at this point is that the vibration changes that occurred when audible noise was present (defect frequency changes and increase in spectral floor) gave strong evidence of outer race sliding.  Even if I didn't see anything on the inspection, I would still lean toward that scenario. The inspection results were ambiguous.     That pattern on the outer race of bearing A in slides 2/3 might support the outer race sliding. The slide 4 and 5 observations, I'm not sure where or whether they fit in.  I'm interested in any thoughts or comments. 

 
Attached Files
pdf BrgDisassembledInpsectionR3.pdf (2.40 MB, 22 views)

electricpete

Sr. Member
Registered:
Posts: 627
Reply with quote  #21 
Another data point: The bearings are loaded below the SKF minimum recommendation:

So this bearing does not meet the minimum radial load, has no built in axial load (wavey washers) like most motors do, and I think what pushed it over the edge was the poor lub conditions.  Those poor lub conditions tend to retard ball rotation within the cage which increases tendency for sliding at the race.  Since it's a relatively low speed bearing, the sliding might be more expected to occur at the outer race vs inner race.  The direction of change in defect frequencies suggests the sliding did occur at the outer (vs inner) race.  The outer race is where we saw some unusual markings, after only 24 hours of operation. 

The strongest evidence of sliding was the noise and raised spectral floor, which came and went together.    The change in defect frequencies alone might indicate change in axial loading of bearing due to thermal expansion / sticking in housing or it might indicate skidding.  The fact that this change in defect frequencies occurred in conjunction with the change in noise and chance in spectral floor suggests the change in defect frequencies were due to sliding/skidding. 

We didn't see any unusal rise in temperature on the bearing housing.  This fact doesn't particularly support the scenario of thermal expansion causing change in axial position causing change in defect frequencies, but it also doesn't particularly support a skidding scenario.   (If you asked me before this event, I would have suspected that skidding would typically show up as increased bearing temperature, but who knows.) 

If you asked me before this, I might have also expected to see wider spectral peaks or multiple spectral peaks as a result of cage speed changing during the measurement, but didn't see anything like that. 

If we had coupled the motor up and loaded it, there's a chance the sliding/skidding would've went away.  As you add load to the motor, everything tends to heat including the shaft.  Any temperature rise of inner ring above outer ring would tend to reduce the internal clearance which would reduce the tendency for skidding.   If skidding is going to occur, it's more likely with the motor unloaded (cool).

Skidding/Sliding is still my conclusion.  I really don't see any other explanation especially for the unusual intermittent change in audible noise and vib spectral floor which was  resolved by swapping the bearings / grease. 

For me it's a pretty interesting case study.  It's the first time I feel like I have been able to identify evidence of skidding with high confidence.   We were lucky to be able to grab vibration both while the audible noise was present and while it was not.  

Although no comments from the board lately.  Does that mean you guys disagree or maybe it's just too unorganized?  Let me know if some part of what I wrote doesnt sound reasonable to you.  I may try to pull together everything I wrote in this long thread into a more-coherent case study writeup for posterity.
OLi

Sr. Member
Registered:
Posts: 1,855
Reply with quote  #22 
If you compare with what I and others noticed with CARB bearings being very noisy so even the end user complained in several cases and it was repeatedly solved by increasing load. "Nothing" to be seen in vibration or bearing analysis by OEM. This bearing is a roller bearing and not in a motor but mostly fan bearings having problems but it is still skidding due to low load that give the noise so I agree to your conclusion and likely if connected and loaded it would have gone away but you never know. Roller problems above was loaded and operating in all cases. 
__________________
Good Vibrations since early 1950's, first patented vibrometer 1956 in the US.
http://www.vtab.se
electricpete

Sr. Member
Registered:
Posts: 627
Reply with quote  #23 
Thanks Oli.  I'm not that familiar with Carb bearings but it sounds like they were misapplied in some cases.

I do often hear a screech during direct on line start of a greased ball bearing motors... the screech goes away as they get up to speed. I tend to think that is also skidding.  I think skidding is more likely during start because the grease is cold and the acceleration forces may play a role also.
OLi

Sr. Member
Registered:
Posts: 1,855
Reply with quote  #24 
It is not the VBelt? :-) Yes it does take some time for the grease to extract the oil to do the lube. It is even more obvious in the permanent lubed "friction" bearings that is in all small fans that exist in everything nowadays like PC's, Cars, power supplies and you can get bearing instability in them too, very nice.
__________________
Good Vibrations since early 1950's, first patented vibrometer 1956 in the US.
http://www.vtab.se
tomcd3

Member
Registered:
Posts: 59
Reply with quote  #25 
SKF has empirically quantified when cold bearings will "screech" due to "cold" grease at start-up. We have also proven that the balls/cage will "hoot" under certain conditions. Two conditions to avoid in motors which are most susceptible to these issues. 
tomcd3

Member
Registered:
Posts: 59
Reply with quote  #26 
CARB is truly an amazing solution, but it can bite you if misapplied. It usu. is the ultimate solution for most SRB/SRB fans.

Some of the rules of fans and CARB that I have published over the years one must adhered to; some regardless of the brg type.

First rules of CARB®/SRB applications:

1.  CARB® is lubed from side not center like SRB                                                                                                                                             

2.  CARB® takes no thrust load, so be sure to include stab. rings in all housings (hence the new locating and non-locating terminology)

3.  when finishing assembly / rebuild of fan, be sure CARB® inner ring is within outer ring before startup (unless it is a fan moving hot air).

First rules of fan applications:

1.  surface roughness of brg foundation should be Ra < or = 12.5µm or better

2.  flatness (planicity) measured diagonally on housing foundation should be < or = ISO IT7 tolerance grade (see chart below)

3.  if necessary, any shim / spacer should support the entire brg housing base.

4.  tolerance of shaft under sleeve and seals should be h9 (see chart in SKF literature or webpage)

5.  out-of-round of shaft under sleeve and seals should be IT5/2 or better (see chart in SKF literature or webpage)

6.  surface roughness of shaft under sleeve and seals should be Ra < or = 3.2µm

I will provide so much more when you contact me regarding a fan you want to make more reliable, smoother and cooler running.

p.s. when moving hot air; it's all about the heat flinger (cooling disc).

Previous Topic | Next Topic
Print
Reply

Quick Navigation:

Easily create a Forum Website with Website Toolbox.