Solving Bearing Cage Failure for 5+ Years

[Chronos]

I've worked on this problem for over 5 years without success. A machine I am familiar with has a chronic bearing failure problem. Specifically, the cage disintegrates after about 6 months. The bearing is used in a vibratory application. An eccentrically weight shaft impels a total mass of about 2000 kilos with a force of 12 g's max at a frequency that does not exceed 600 cps [usually closer to 300 cps]. The bearing cage invariably fails after about 1200 hours. The cage rivets break, apparently due to fatigue. We even tried a one piece cast bronze cage [very expensive], and it too broke. The bearing manufacturers' engineers are stumped, as am I. Any ideas? ps - no evidence of lubrication failure.

 

[wolram]

Chronos.
My company is presently researching ways to increase bearing life on a
multitude of applications in extreme environments
I cam across this site
http://www.reliability.com/articles/article30.htm
it may be of interest to you.

 

[FredGarvin]

Chronos,
Can you provide a bit more info? What is the type of bearing you are using? Is this strictly a radial load situation or do you have thrust loads as well?

Has there been any indication of ball pathway wear on the inner race? Do you monitor bearing temps? What about vibration data?

I'm wondering also if you're exciting a mode in the cage.

 

[Astronuc]

1200 hrs * 3600 s/hr = 4.32 E6 s.

4.32 E6 s * 300 cps = 1.296 E9 cycles. I impressed that you got 1E9 cycles

Has anyone done an S-N curve (Langer curve) for the bearing material? Does anyone have an idea as the magnitude of stress intensity (and $\Delta$K) in bearing box/cage?

Does anyone bother to monitor the acoustic signature. Low frequency noise amplitude should increase dramatically just before failure.

2000 kg / 12 g's ! Depeding on the geometry (and stress intensity factor), that could produce a substantial load.

As FredGarvin suggested, it would be useful to have some schematic of the bearing and loading.

 

[Chronos]

All good questions and much appreciated. The basic design uses two eccentrically weighted shafts. The weights are 'timed' to produce a horizontal impulse to a rotating shaft [the application is centrifugal drying of minerals]. The drive housing rests on leaf springs. Using a vibrometer, the vertical moment is nearly zero [<.10 g's]. If the weights get out of time... bad things happen. The machine self-destructs in about 5 minutes. But that rarely happens. A motor driven cog belt is used to power the vibratory mechanism. I shimmed the vibrator shafts to remove all the endplay on the vibratory shafts. That did not help [frankly, I think it made the problem worse]. I redesigned the bearing housing. Went from a wide outer race to a narrow outer [about 3"] and used the extra space to create a gravity fed lubrication puddle for the bearings to run in [only half way on the rollers]. The original design relied on spray lubrication. I never liked that design because the spray nozzles broke. My 'improved' design lasted about 200 service hours longer than the old design. I am obviously missing something.

 

[FredGarvin]

Without knowing what kind of bearing you are dealing with and an exact layout of the system, here's my first stab at things:

1. You said you shimmed the rotors to remove endplay. By doing so you may have created an excess in thrust loads due to thermals.

2. If you went to a gravity feed system on the lube, what is your scavenge scheme like? Is that gravity as well? You may be pooling too much oil in the bearing.

3. You said you redesigned the bearing housing. What did the runout of the bearing housing look like with respect to the shaft axis?

This is good stuff Chronos. Keep the info coming. Some bearing info and even a sketch would be a homerun.

 

[brewnog]

I second the request for a sketch...!

Ta

 

[FredGarvin]

Well Chronos...I couldn't wait, so I sent out a feeler to our bearing guru here at work. I didn't get three words past "high radial load" and he said "the cages are breaking, aren't they?" So...needless to say I got an education on what is going on. He also said it would be relatively easy to fix.

The big assumption is that you are using angular contact bearings. I think that is a safe assumption to make.

In a nutshell a few things have to happen. 1) Get the endplay back into the shaft. 2) Go back to the rivited cages. 3) Get your bearing folks to get you on to a less severe (close to 0°) angular contact bearing.

Without getting into too much detail, the load sharing between bearings is not happening the proper way. Because you have angular contacts, this results in ball speeds fluctuating around the perifery during operation. At an instantaneous moment in time, the balls are wanting to travel at varying speeds, depending on where they are in relation to the contact angle at that moment. This results in a wide range of ball speeds. The cage will essentially travel at a speed which will be the average of those speeds. So now the cage is either holding back a ball going faster or trying to speed up a slower one. That is all happening through the cage.

I like these kinds of problems because they are always a learning experience.

 

[Chronos]

Thanks again. Duh, I forgot to mention we are using cylindrical roller bearings. That was probably a bad idea to start with, but it was cheap and looked like it might work. Of course it don't look so cheap when you have to replace them a couple times a year. In the original design, end play on the vibratory shafts was not considered. They had about 1/32 slop built in due to tolerance stacks. After the first bearing failure, removing the endplay was the first attempted solution. That did not work. If anything, the bearing life actually decreased slightly. The gravity flow lubrication appears to work nicely. The bearing retainer acts as the dam and excess lubricant overflows and returns to the sump. The housing runout is no more than .0005".

 

[Astronuc]

It would help even if there was a simple diagram/sketch to show the geometry of the shaft and bearings, but I think I have an idea of what you are doing. An image (from digital camera) would be useful. Also, if you could provide a model number and manufacturer, it would be helpful.

Do you have pictures of the failed bearing cages? Do the fail and the same location(s)? Any galling in the race? If so, please characterize.

I suspect that you are using the bearings outside their intended design envelope. Most roller bearings wear out quickly if the vibration amplitude is significant, and that sounds (no pun intended) like what the machine is doing.

Each roller bearing would act as a local stress riser in the bearing race (a bit like shot peening, I imagine).

You might consider magnetic bearings, but I believe they are quite expensive.

 

[Chronos]

Since the vibratory moment is paralles to the roller direction, there is a recoil between the shaft and inner race. But lubrication viscosity is sufficient to prevent the roller surfaces from making metal to metal contact with the shaft. The outer race and inner races show normal wear, but the cage is clearly fatigued [especially noticeable in riveted cages]. Evidently the roller ends are prying the cage apart. I have no idea how or where the lateral force originates. NOTE: the leaf springs do twist slightly in the direction of the vibratory drive motor. I sometimes wonder if the drive motor, which is base mounted, should be spring mounted. Perhaps the slight twist over time introduces a torsional moment.

And I don't disagree with the notion the bearings don't match the application. What is really annoying is that the bearing manufacturer [who recommended the application] is not willing to discuss the problem unless we agree to pay them consulting fees.

 

[FredGarvin]

Ideally roller bearings are the proper choice when dealing with high radial loads only. The tough part is figuring out where the other force components are coming from. However you do need some way to take the axial loads you are undoubtedly creating. Usually we go with an angular contact bearing/roller bearing combo.

Are these bearings rigidly mounted or do they have outer race lube/damping?

 

[Astronuc]

What is really annoying is that the bearing manufacturer [who recommended the application] is not willing to discuss the problem unless we agree to pay them consulting fees.

This situation has become more common place in the last 20 years. As prices (and revenue) decrease on technological commodities, companies put their staff under more pressure to provide after-market sales. That means you buy a product, and then pay extra for O&M. It's more commonplace in low volume products like aerospace and nuclear industries. In high volume industries like consumer electronics and computers, the product is usually designed for about 4-5 years before it 'needs' replacement.

I have also encountered more than a few technical sales people who really don't know what they are talking about, particularly with a specialized application such as your bearing is a vibratory load.

As for the condition, the location where failure initiates may be experiencing alternating compression/tension (local stress concentration) or compression with high shear stresses. Ultimately that will lead to void/crack nucleation, which then propagates to failure.

I addition to magnetic bearings, a gas bearing might do the trick. I would still like to see a schematic or image of the bearing and the housing.

 

[Free Anarqy]

(just a little late) Did you tried SKF /VA405 bearings? it looks like fatigue cracks due to high accelerations. These bearings are intended for vibrating applications. I hope my answer doesn't arrive too late (I have just read your question)

 

[JaredJames]

Well this thread is 5+ years old. I would seriously hope it isn't still an issue (something tells me it isn't).

 

[AlephZero]

Well this thread is 5+ years old. I would seriously hope it isn't still an issue (something tells me it isn't).

Well, if the bearing cage fails every six months, the practical engineering fix is just replace it every 5 months. The technical term for this is "preventative maintenance"

 

[JaredJames]

Something I'm sure they'd have noted circa 5 years ago...