This is a case study with a few twists and turns. The root cause is probably obvious (over-greasing), but the discussion leads to some some questions along the way.

Machine Description: This is a 3600 rpm, 100hp horizontal TEFC motor, driving overhung centrifugal pump through a Shim Pack Coupling.

Big Picture Overview of History:

• Motor Bearings last replaced 3/28/06.
  
• Motor Bearings regreased 9/06 and 8/07.
  
• Motor bearing vibration magnitudes and housing temperatures stable through January 2008.
  
• Machine Shutdown for about 4 weeks for plant outage during April 2008.
  
• No work was done on the machine during the outage
  
• Shortly after restarting the machine for plant startup, we noted on 4/28/08 that there was high noise, increased vibration magnitudes (inboard bearing), and high housing temperautres (inboard bearing).
  
• During bearing replacement on 5/1/08, it was noted that the cavity was completely full of grease. Grease was on the motor endwindings and also pushed out around the shaft (slide 9 shows grease coming out around the shaft).
  
• During bearing replacement, shaft and housing fits were checked - normal

I am guessing it is more likely the bearing was packed too full during bearing replacement in 2006, rather than that the bearing was greased until full during regrease (that would be a lot of pumping). Our guys are knowledeable on both activities, but regreasing is more of a routine activity, replacing bearings and repacking the cavity is not so frequent and more likely to be improperly done. The above time-line leads to the following question:

QUESTION 1 – Why did the problem show up on the first reading after machine down-period? Just a coincidence? Or is there something about sitting idle for a period of time that might make the lubrication situation worse. And why didn't the problem show up on temp and vib after relubrication?

Temperature Discussion: Temperature trend is shown on slide 8. Bizarre that the machine operates with normal temperature (compared to sister machines) with no change after relubrication, but dramatic change after plant outage. Thermal images shown on slide 21.

Vibration Discussion: The vibration pattern I would characterize as random non-syncronous soup. I could not identify any harmonic series in the spectrum. TWF on slide 3 is approx 40 g's pk/pk, but no periodicity evident. Spectra on slide 2,4,5. Demod Spectrum on slide 12 shows no frequencies anywhere near fault frequencies. Demod TWF on slide 11 has very low correlation. The only pattern I would identify looked something like 0.1 orders sidebands on each side of running speed and it's harmonics (slide 9). That particular pattern has in fact been there since the bearings replaced (slide 10). Waterfalls on slides 7 and 8 show the vibration pattern was very low and relatively stable through the period of relubrication and only jumped dramatically after the April 08 outage.

Bearing Inspection: As discussed above, the bearing housing was reported full of excess grease. I didn't get to see that myself. The bearing was turned over to me with grease already cleaned out. The bearing was cut apart for inspection. Observations from bearing inspection where as follows:

• There are some relatively hard specks of grease still adhering to the cage after cleaning.
  
• The races are stained light brown, typically associated with overheating of lubricant.
  
• A unusual pronounced wear track with raised edges along one isolated portion of the outer race, which I believe indicates ball skidding (do you agree?). Shown in slides 14-17.
  
• One ball has a large surface discolored and rough (slide 18). An adjacent ball has a stripe mark/discoloration, but not rough - apparent discoloration while stationary - either corrosion or heating during bearing removal. All other balls appear in reasonably good condition.
  
• The wear pattern on the inner race (offset from the center) appears to suggest axial loading on both bearings. The wear pattern on outer race offset in opposite direction (although not as much), again consistent with axial loading both bearings.

Question 2 - Do you agree the photo's look like skidding? Is skidding an expected result of overlubrication?

I have to go back and double-check whether the indicated loading would be opposite direction among both bearings (suggesting something like load created within the motor due to unusual lack of clearance for thermal expansion) or same direction on both bearings (suggesting perhaps external axial load transmitted from pump thru coupling... although even then that load should only be taken by one of the two motor beairngs.). I'll follow-up on that.

Question 3With the benefit of hindsight, do we assign any significance to non-sync sidebands of running speed such as 0.1 orders? Do we see anything in the vibration that we would associate with skidding?

I have seen lubrication problems before with similar non-syncronous soup - lots of non-sync stuff and no identifiable defect patterns. But I have never seen these types sidebands associated with lub problem before. I also usually find the lub in bad shape, but not a lot of defects on the bearing and never this skid-looking mark. So I have a suspicion the 0.1 sidebands might be related to skidding, although I don't particularly know why that would be so. Any comments on whether non-sync sidebands around running speed might be related to skidding or something else?

Slide 9 shows coupling shims were not cracked but all a little bit rusty. (also confirmed later when couplign was disassembled). Rusty shims appear pretty common for us, even on indoor motors such as this... do others see this often? I don't think the rust has much to do with it.
Question 4. Do you think rusty shims have any significance. Would you replace coupling shims with a little rust, but no cracks?

Thermal survey for misalignment: Coupling was surveyed before bearing replacement as shown slide 20. It does show the heat conducted from the hot motor bearing on the left as expected. No evidence of heat generated within the coupling that I can tell. One note to explain the thermal image - the coupling guard covers the top of the picture and the out-of-focus effect of the coupling guard makes the coupling appear to get cooler going toward the top of the image... not really the case.
Another question that might be logical for readers to ask is whether there was early warning. Certainly there were no clues in the temperature (which I find unusual.... shouldn't overlubrication be indicated almost immediately on temperautre... especially on a 3600 rpm machine with large 6316 bearings?). And in the vibration from the waterfall and spectra you can see there was some non-sync stuff, but again it doesn't scream bearing to me (no fault patterns). Also we trend "spike energy magnitude" as well as another high frequency parameter for bearing degradation ... didn't see anything unusual until after the outage. I'll post those trends later.

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

Post_small.ppt (1,052 KB, 114 downloads)

Trends of spike energy for the inboard bearing is shown on slide 1. Trend for the outboard bearing slide 2 (note the outboard bearing was also full, but never got hot or had high vibrations). Trend for some sister motors is shown in slides 3-8. The inboard bearing was a little higher than outboard and sister motors, but no clear increasing trend until after the recent outage.

I did review the loading pattern implied by the ball path on the races. The pattern suggests an inboard thrust on the the inboard bearing and outboard on the outboard bearing. As if the rotor underwent thermal expansion and neither bearing could slide within it's housing to accomodate it. See slide 9.

Note the fits were normal. (3.1502" shaft diameter and 6.6930" housing bore recorded - 6316 nearing). Wavey washer is used at the outboard end.

The O.D. of the outer rings look almost brand new (slide 10). A little unusual imo. I'm used to seeing a little evidence of fretting on the outer ring.

So how do we view this evidence of axial loading on the bearings? Unrelated to the main failure? Contributor to the main failure? Result of the main failure?

It has been pointed out before (Arie Mol) that a bearing running hot reduces the clearance between the outer ring and housing.. possibly preventing the bearing from sliding axially. Do you think that might be the case here? Since bearing replacement in 2006, we had 120 - 150F bearing housing temperatures which only jumped up around 200F recently. Bearing might be 20F or more hotter than the housing.

TrendPostAndBearingLoadPathSmall.ppt (740 KB, 55 downloads)

Some of you may have noticed that I attempted to investigate whether evidence of misalignment was present by thermal scan of the couplilng. After further review, it appears that this technique (thermography of the coupling to investigate misalignment) is not applicable to shim pack couplings. This is based on Piatrowski's Shaft Alignment Handbook - Figure 2.41 - 2.46 shows experimental plot of coupling temperature vs misalignment for the following couplign types: jaw, tire, gear, ribbon, chain. For all those types of couplings, a noticeable temperature difference is created at 0.010 mils offset. The discussion indicates that heat is generated for the above type couplings which are elastomeric (jaw, tire) or mechanically flexible (gear, ribbon, chain). But no significant heat is expected for disk type or diagrhram type couplings since they get their flexibility from bending within their elastic range. The only thermal evidence of misalignment on these machines would likely be indirect (heating caused by loading of the bearings).

After thinking some more, the loading pattern described above (inboard thrust at inboard bearing and outboard thrust at outboard bearing) makes perfect sense given the preload washer configuration. Inboard bearing is fixed and outboard bearing is floating with preload washer pushing inboard on it's outer race. This puts the rotor in compression and causes the loading patterns described.

So with clear evidence of preload present on these bearings, then what caused the skidding? I assume it must be poor lubrication conditions caused by overlubrication. I have read that poor lubrication conditions cause skidding but I'm not exactly sure the mechanism. I assumed before that friction between the cage and balls prevented balls from rolling to maintain contact (hence the connection between poor lubrication and skidding). But I'm not exactly sure why overlubrication would increase friction between balls and cage.

Addressing just the skidding issue I think it has nothing to do with lubricaation. Proper bearing loading IMO is the main factor affecting skidding. In this particulaar case the bearing could be unloaded, say, due to mislignment.

With the damage in the outer race shown it is strange that it did not produce BPFO, at least in the demodulated data. Could it be due to wrong filtering?

The term I remember from way back is "churning". Overgreasing will result in higher temperatures due to the frictional heat developed by the lubrication (grease) colliding with itself. Increased pressure in the bearing itself will "restrict" not stop the rolling elements from rolling thereby causing wear patterns in the raceways due to apparent "skidding".
I have made an effort here where I work to question old practices such as pre packing a bearing with grease. The assumption is that you fill it to the brim and what isn't needed will flow out. Well, that's OK if there's room for such flow but in the case of a motor bearing, where does the grease go? Unless you have a drain port of some kind. My 2 cents.

Thanks for the good comments David. I was beginning to feel like I was talking to myself. I guess that's what I get for making a long rambling post.

quote:

David wrote:
Addressing just the skidding issue I think it has nothing to do with lubrication. Proper bearing loading IMO is the main factor affecting skidding. In this particular case the bearing could be unloaded, say, due to misalignment.

I don't claim to have any great understanding of what skidding is all about (it's a little mysterious to me). Clearly lack of loading can be a contributor. But poor lubrication can be a contributor also. It shows up in many different references. Below are a few excerpts that I found from a quick look (there are a lot more out there that I could find with a little digging)

quote:

Excerpt from SKF Interactive Engineering Catalogue:
In order to guarantee satisfactory operation, deep groove ball bearings, like all ball and roller bearings, must always be subjected to a given minimum load, particularly if they are to operate at high speeds or are subjected to high accelerations or rapid changes in the direction of load. Under such conditions the inertia forces of the balls and cage, and the friction in the lubricant, can have a detrimental influence on the rolling conditions in the bearing arrangement and may cause damaging sliding movements to occur between the balls and raceways

quote:

Excerpt from "AVOIDING ELECTRIC MOTOR DRIVE END BEARING PROBLEMS" By Dan L. Leah, P. Eng, Engineering Manager, Western Region
NTN Bearing Corporation of Canada Ltd.

There are a number of factors, which dictate whether sliding or rolling will occur under low-load conditions. These include cage design and material, the amount and type of grease in the bearing and the ambient temperature. Some might say "I never experienced this problem with XYZ brand bearings . . ." Perhaps not, but it can happen with any brand of cylindrical roller bearing. NO design is immune. Bear in mind that skidding can occur without being detected.

To avoid this problem, it is suggested that the rotor be loaded. This can be done by adding weight to the rotor shaft by installing a sheave or heavy coupling half. The minimum recommended load to avoid skidding is 4% of the static load rating of the bearing (Cor), which can be found in the manufacturer's catalog.

As a further aid to prevent skidding, we suggest that the roller bearing be washed in CLEAN SOLVENT, dried off with a lint free towel and oiled on all internal surfaces with thin oil (SAE 5 - 10) prior to installation. The grease can then be applied as normal. The oil film will keep the drag of the cage pockets to a minimum allowing the roller train to turn freely when the motor is started.

Note the last reference is directed toward cylindrical roller bearings, but I believe the same principles apply.

Given that we have strong evidence of preload (the offset wear patterns on the race) and strong evidence of poor lubrication (housing was found packed full of grease and spilling out around the shaft in both directions), I am inclined to attribute the skidding to poor lubrication (rather than lack of preload), even though the exact mechanism by which poor lub is supposed to lead to skidding is a little mysterious to me (is it friction between ball and cage preventing rolling?). And does skidding require BOTH low load and poor lub or can it occur with just one... beats me.

quote:

Davide wrote:
With the damage in the outer race shown it is strange that it did not produce BPFO, at least in the demodulated data. Could it be due to wrong filtering?

Guess it could be. My hi-pass filter was 600hz; sample rate was 12,800hz; I have another envelope decay parameter whose value I can't recall. But I didn't see any series of BPFO harmonics in the regular spectrum, so I don't particularly expect to see it in the demod spectrum (at least in my implementation where the demod spectrum is based on same sample rate as the regular spectrum). I do have the acceleration twf data in xls file if anyone has the ability to process it to look for something I may have missed (let me know and I'll post it).

But now that you mention it, the vibration expected from skidding and from this damage is interesting to try to figure out. I have asked the forum what kind of vibration might be expected from skidding (?), but no response. I guess I can reconcile why there is no evidence of BPFO impacting - the pattern on the outer race is not exactly a defect, the sliding appears to have pushed a little ridge of material out on each side of the ball path... the balls don't roll over those ridges... just through the middle of those ridges. I should also mention that the ridges are not as deep as they appear on the picture. You can feel them on both sides, particularly toward the trailing edge of the skid mark, but they are not so deep as the picture makes them appear.

Interesting to see the comment in the last quote above "Bear in mind that skidding can occur without being detected.". Exactly what the author had in mind I can't be positive, but if the meaning was that skidding does not give an easily recognizable bearing defect pattern, then it certainly seems consistent with what I saw in this one case study. And there are still those strange 0.1 order sidebands....?

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

I looked up a few more references to see if there was any more relevant info about the role of lubrication in skidding:

quote:

Handbook of Lubrication and Tribology, Volume 1, Chapter 13 Lubricating Industrial Electric Motors:
In rolling element bearings, excessive viscosity can result in ball/roller skidding and viscous drag.

quote:

Handbook of Lubrication and Tribology, Volume 2, Chapter on Roll\ing Element Bearings:
Selection of proper oil viscosity is essential and is based primarily on expected operating temperature, speed, and bearing geometry. Excessive oil viscosity many cause skidding of rolling elements and undue lubricant friction with severe overheating and raceway damage.

quote:

ASM Handbook, Volume 18, Friction Lubrication and Wear Technology ISBN 0-87170-380-7:
Also, when speeds are high, depending on the amount of lubricant residing in the bearing cavity (free space), viscous drag friction of the lubricant through which the rolling elements orbit can be significant. In fact, excessive lubricant in the bearing cavity during high-speed operation can cause skidding (that is, gross sliding of the rolling elements over the inner raceway). Skidding in the absence of adequately thick elastohydrodynamic lubricant films can result in substantial damage of the contact surfaces and rapid bearing failure.

The first quotes above two talk about the role of viscosity in causing skidding. It tells us only that lubrication parameters play some role related to skidding.

The last quote above specifically mentions excess lubrication as a cause of skidding. I'm not sure why they specifically mention skidding on the inner race instead of outer race though.

By the way - thanks for your comments also Ron. I missed them the first time around. I guess you are saying it is not so much poor lubrication as extra grease getting in the way? That is something to think about.

I have seen different recommendations on packing. Most say pack the cavity partially somewhere 25-75%. But for packing within the bearing itself, some references say pack it full, some say pack it roughly half of empty space full, some say just spread it around thoroughly around the rolling elements/cage. Seems like not a lot of agreement on how to pack the bearing itself.

I have always thought that for high speed and large high speed bearings we should stick on the low side for both the cavity and the bearing. But I think the practice for most of our technicians has been to pack the bearing full and cavity half full. Maybe something for me to follow up on.

I have seen many skidding effect on brg but most of them were on spherical and cone brg. Large amount of grease trapped in the brg around the rolling element may cause enough restriction to slow down the rotating element as they comes out of the loading zone. Once the rolling element re-enter the loading zone they have to accelerate rapidly to accommodate the linear speed of both inner and outer race. This effect may cause either sliding mark on the outer race at the beginning of the loading zone or a waved wear pattern similar to washboard on a gravel road on the outer race as well. Both phenomenons are dictated by lubricant, brg geometry (internal clearance) and load. Sliding mostly damages the rolling element causing 2 x Ball spin frequency and washboard wear induce strong BPFO.
Best regard, Marcel

Pete - I'm definitely interested in what's going on here, but I have been very busy. Still digesting the data and posts. I am interested in how you are doing the autocorrelation in Excel. Too bad Emonitor doesn't have the feature. Could you email me a description and/or spreadsheet.

Steve

I am still bothered by the fact that ball and race damage did not show up in vibration data. One possible reason could be because of low HP filter setting at 600 HZ. For a 2 pole motor it is too low letting in too much of non-problem frequencies.

On another hand there is a defiinitive problem when looking at the spectra and TWF. The data is very random. This fact may be another indicator of looseness which exhibited itself so profoundly due to unloaded bearing. Could something happened during the outage that caused that?

If the bearing outer ring is unable to slide axially then axial preload in ball bearing can become extreme and overheats the oil film and overloads the raceways.
However for same reason at other temperature level the preload may become zero, introducing positive ax / rad play in bearing causing skidding because contact angle is no longer defined.
Both phenomena could have taken place however at other time and other temperature condition.
Regards,
Arie Mol

Good comments.

Marcel – your comments seem to summarize nicely everything I have been reading about skidding in the last few days.

Arie – good point that the bearing is not always in the same condition. If the damage had occurred/begun during the start after the outage, everything would have been cooler.

David – could be looseness if it were for the period immediately after start. But it apparently did not remain during operation as shown by the ball tracks. Our records show no work on the machine during the outage (which is what makes it strange that it appeared immediately after the outage).


===================================

I noticed a feature on our "skid mark" that appears like a comet as labeled in slides 2 – 4 attached. It has a head at the beginning and tail at the end. In slide 2, the direction of rotation is labelled based on the known direction of rotation of the machine (not based on looking at the marks and what I think is going on). The head is at the leading part o the skid mark and tail at the trailing part.

I noticed in a textbook (slide 1) that their view of skidding has a similar comet-like appearance. I have labeled what I think would be the direction of rotation for their bearing IF it is behaving like ours.

So.... any ideas what causes that pattern at the head of the comet, which would presumably be where the rolling elements initially lose contact with the race while entering the unloaded zone? I would have thought there is not much damage created upon losing contact as making contact (and indeed the worst damage of raised edges is towards the end of the tail). Maybe it is just the case that all the rolling elements lose contact with the outer race at exactly the same location every time?

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

SkidMarkCometAppearance.ppt (532 KB, 25 downloads)