Thrust bearing overheating--shaft sleeve restricting heat transfer?

[jimhokie]

I'm working with a pump manufacturer on a new design that is experiencing excessive thrust bearing temperatures. We know part of the problem is higher than designed-for thrust load, but the magnitude of the thrust doesn't seem to completely account for the temperature rise. This design employs a shaft sleeve that slides on the shaft with a close clearance, and the duplex pair roller bearing inner races are shrunk onto the sleeve (rather than being shrunk directly to the shaft). We are wondering if the lack of material continuity between the shaft and shaft sleeve, and the resulting decrease in heat conduction away from the inner races could be contributing significantly to the overheating, i.e., more of the heat goes into the oil rather than being conducted away through the shaft. The bearings are ring lubricated from a bearing housing oil sump. Does anyone have experience with this type of design or problem? Thanks for any thoughts or suggestions.

 

[JJPellin]

We have a large number of pumps that have bearings mounted on sleeves as you describe. Most of them use ball bearings as opposed to roller bearings. I would not expect the heat transfer issue to be the most important. I would tend to look elsewhere. But a number of possibilities come to mind.

Most of these installations in our plant do not have good specifications for the proper fit between the sleeve and the shaft. With the bearing mounted directly to the shaft, we would expect a slight interference fit. But the sleeves tend to be a slip fit in order to make assembly and disassembly easier. You could be putting too much interference in the inner ring of the bearing if you have a tight fit between the bearing and sleeve and a tight fit between the sleeve and shaft.

The outer race of the bearing could be too tight in the housing. This should be a loose fit as called out in the standard bearing fit tables. If it was tight, this could take up the internal clearance in the bearing and cause overheated.

Axial spacing is important. In most of these installations, the inner races of the bearings are clamped together but the outer races have axial clearance. If the bearing is being crushed axially, it could overheat.

Any rolling element bearing will run hot if it is over-lubricated. A high oil level flooding up into the rollers will cause it to run hot.

It is possible to overheat from inadequate lubrication. If it is a back-to-back configuration, the loaded bearing may be the one farthest from the oil ring. It could be that the loaded bearing is not getting enough oil. A second oil ring on the opposite side can provide oil to this loaded bearings.

Since you state this is a new application, I assume that you are cooling the oil rather than the housing. The old housings that had cast in cooling jackets would constrict the bearings and cause them to run hot.

If it is a hot process, you could have the opposite problem than what you proposed. If there is heat conducting down the shaft from the process, it can cause a bearing housing to run hot. In some cases a deflector fan clamped to the shaft can dissipate the heat so it doesn't make it to the bearing housing.

Addition details about the pump configuration could help us understand the potential problems.


Johnny Pellin

 

[jimhokie]

Thanks, Johnny...I misspoke about roller bearings. They are ball bearings. The sleeve does make assembly and disassembly easier, but the main reason for it in this design is because a larger bearing was required for life purposes and the sleeve was needed to fit the inner race ID.

I will take a close look at the fits between the shaft and sleeve, the sleeve and the inner race, and the outer race and the housing.

You're right--the inner races are clamped together. The manufacturer is working with bearing manufacturer recommendations for the amount of preload, so I'm thinking this may not be the problem.

They are confident the oil level is good. It is just slightly above the bottom of the balls, which will prevent the balls from churning the oil. Viewing through the oil fill hole shows the oil ring is doing its job well. I'll keep in mind the possibility that the inboard bearing (furthest from the oil ring) may not be getting sufficient lubrication, but the outboard bearing should be the one carrying the greater load.

There is no external cooling provision, by specification. Cooling of the oil sump by convection and conduction is supposed to do the job, thus there is not a lot of margin for unplanned heat generation. One theory is that given the large size of the bearing in relation to the oil sump volume, heat generation by shearing of the oil film is exceeding the passive cooling capacity. Any thoughts on this? Mechanically, this pump is based on prior proven design in this respect (except in the prior design, the bearing is mounted directly on the shaft), but obviously something is different for this design--thus the concern that the shaft sleeve may be impeding heat transfer away from the bearing. Has this ever been identified as a problem before, and if so, is there some kind of heat conductive "filler" that could be used between the shaft and sleeve to improve heat removal?

The pump is currently in a test loop, and the bearing temperature rise is apparent long before the loop water temperature rises, so the heat isn't coming from the process liquid, at least not initially. In it's design service the water will be typically around 120F.

The pump is a horizontal, multistage, high-energy pump running at 3600 srpm. Let me know if any additional information may be helpful.

 

[JJPellin]

Given this additional information, I would have a few more concerns. I still believe that the problem is not related to the limitation on head conducted into the shaft.

Since it is ball bearings clamped at the inner race it is very important that the bearings are correct. They either need to be a matched set that are ground for back-to-back mounting or two individual bearings ground for flush mount. These bearings can be ground for a pre-load. I would not recommend it. If you are concerned about ball skidding in the unloaded bearing, you can use a 40/15 "diamond pack" set with a 15% load angle on the unloaded bearing.

You stated that the oil level is at the bottom of the balls. Did you mean that the oil is actually up into the balls? I would normally not recommend this. If the oil ring is set up properly, it should provide as much oil as the bearing needs.

You mentioned that this is a larger bearing installed for load reasons. At 3600 rpm, over-lubrication is a great concern and the larger the bearing the more likely it is to be a problem. But this is easy to check. Drop the oil level and monitor the affect on bearing temperature.

You did not mention axial clearance on the outer races. These cannot be clamped hard together or the pre-load on the balls could be too great.


Johnny Pellin

 

[JJPellin]

I had to deal with some urgent issues and missed a few points with my last post. First, I should have asked what you were considering as excessive thrust bearing temperatures? How hot are they getting and how are you measuring the temperature?

I am surprised that the inner bearing is the loaded one. I am used to rotors thrusting toward the coupling. With a back-to-back mounting and the thrust bearing on the outboard end of the pump, the outermost bearing would see the normal thrust load.

For a relatively cold water service, no additional cooling should be needed. The final temperature of the housing will depend on the configuration of the housing (external fins or not) and the volume of oil in the sump. But these items should be the difference between a housing that runs at 150 °F and a housing that runs at 180 °F, perhaps. If you are much hotter than that, then there is probably something else wrong.

I keep coming back to over-lubrication as the most likely cause once you confirm that the mounting of the outer races is correct (not crushed in place).

You indicated that there was an expected high thrust load. I guess the last issue to confirm is the size of the bearing and the expected thrust load. If you can provide these, I could do a quick double-check on the bearing sizing.


Johnny Pellin

 

[jimhokie]

Thank you for the further information. Sorry for the delay in responding as I was traveling Wednesday and Thursday last week and played catch-up all day Friday.

I'm inclined to agree that the problem isn't heat conduction through the sleeve/shaft interface. I can see this adding a few degrees to a stable bearing temperature, but not like we are seeing.

The bearings are a matched pair clamped at the inner race by a shaft nut. At the bearing manufacturer's recommendation, the inner races are ground for a 500 lb preload. (The design thrust load is about 300 lbs.) They intend to decrease the preload once they get the thrust balancing mechanism working properly. Skidding of the unloaded bearing is a concern. Can you explain more about the 40/15 diamond pack set? I'm not familiar with what this is or does.

The original design oil level was well up into the balls, but they lowered it to where it is right at the bottom of the balls. Are you saying the level should be completely out of the low point of the inner race so the only oil ingress into the bearing is via the oil ring? I will suggest further lowering the oil level to check for an improvement in heat-up rate.

The bearing size was dictated more for life reasons than for load, I believe. The specification requires a minimum B-10 life of 50,000 hours. You say at 3600 rpm, over-lubrication is a great concern. Earlier in the test program, the problem was attributed to excess oil accumulating in the bearings because of a spring washer against the inboard bearing that was blocking the flow of oil out of the bearing. This problem was remedied, however the overheating continues.

You asked about axial clearance on the outer races, saying they should not be clamped hard together. There is nothing in the design to provide any additional clamping of the outer races, however, in this back-to-back configuration, aren't the outer races essentially pulled together via the force transmitted through the balls by the preload when the inner races are clamped?

I also have some concerns about the assembly personnel. Can you think of any assembly related problem that could be causing the overheating in an otherwise correct bearing design? I'm just wondering if there is some assembly pitfall that an unknowing or sloppy mechanic could be encountering.

You asked what we consider excessive thrust bearing temperatures. The specification requires a stabilized bearing temperature below 194F. In the last test, the temperatures went from ambient to exceeding 194F in about 40 minutes, as measured by RTDs on the outer bearing races.

The inner bearing should not be the loaded one. You're correct that the thrust is toward the coupling end, which should be loading the outboard bearing instead. The thrust bearing is a duplex pair, size 7315D. By design, the net thrust load should be around 300 lbs, but is now running around 800-1100 lbs. Prior to a modification to the thrust collar design, the thrust was between 1500-2000 lbs. With the improved (but still over design) thrust load, there was little or no difference in the bearing heat up rate, which makes me think the problem is something other than the thrust load.

Any other thoughts?

Thanks again,
Jim

 

[JJPellin]

That is a very large bearing for the speed. If there is a properly designed oil ring that can deliver oil to the bearing, I would not want any oil level up into the balls or races. Since you have an instrumented bearing, it should be a simple test to drop the oil level and watch the temperature.

I do not like a pre-loaded bearing for this service. A flush ground matched pair should provide good service. If there is a concern about skidding, I would put in a "Pump-Pack" bearings set. I incorrectly referred to this as diamond pack above. The correct bearing would have a 40 degree contact angle on the loaded bearing and a 15 degree contact angle on the unloaded bearing. This makes the unloaded bearing more of a radial bearing. It decreased the ability of that bearing to take thrust load, but also decreased the incidence of skidding. I would consult SKF literature about this bearing design.

You are correct that the outer races are pulled together by the forces transmitted through the balls. But if you also crushed the outer races together with a cover that was machined incorrectly, it would distort the outer races and could lead to excessive heat and early failure. The outer races should have axial clearance. This is sometimes referred to as thrust float. At the very end of the assembly, a dial indicator on the end of the shaft should confirm that the shaft can be moved axially by at least 0.002" to confirm it has axial clearance and not crush.

The main assembly errors that could be contributors would have to do with incorrect fits between the bearing and shaft or bearing and housing. If these fits are good, it is key that the bearings are mounted using good practices. Specifically, the bearing should be heated and slipped onto the sleeve. It is possible to press them on, but they are much more likely to be damaged. The nut that clamps the races together needs to be tightened on the hot bearings and then retightened after the bearing has cooled. Ideally this nut would be torqued, but this is not vitally important. The bearing manufacturer should provide a recommended nut torque if you want to do this.

A large bearing at a relatively high speed with a heavy load is easy to over-lubricate. Based on SKF catalogs, this bearing should have a basic thrust load rating of 2,800 lbs for an L10 life of 25,000 hours. The thrust loads you are imposing on it are not too great. I would change to a bearing with no pre-load and drop the oil level out of the bearing. And, as already mentioned, I would confirm that shaft fit, the housing fit and the axial clearance. If skidding is a concern, I would change the unloaded bearing to a 15 degree contact angle.


Johnny Pellin

 

[jimhokie]

About your concern for the preload, since the thrust loads we are seeing are well within the capability of this bearing, how does the preload hurt?

Is there a way to determine if skidding is occuring by monitoring bearing vibration? Or, could monitoring bearing vibration provide any other potential clues to what is causing overheating?

There is supposed to be axial play in the shaft in order for the thrust disc axial gap to vary as needed. I believe they plan to measure this during the next test.

There should be no part clamping the outer races, unless maybe they somehow cock the bearing as they slide the housing in place. Is it possible this could be happening and not be obvious to the assembler?

Jim

 

[JJPellin]

My concern about pre-load is that it is not as exact at the bearing manufacturer might lead you to believe. They achieve pre-load by grinding the races in such a way that there is a slight interference between inner race to ball to outer race. They really aren't setting a pre-load in pounds; they are setting a certain amount of strain because of the interference. They calculate the amount of load that would correspond to that strain. But, if they are a little bit heavy on the strain, the pre-load goes up fast since we are talking about hard steel parts. And, when the bearing starts to heat up, the balls grow more than the races and the load on the balls goes up. That's why I prefer zero pre-load. Then you have a little bit of margin for error in case the pre-load ends up being a little bit more than expected because of dimensional tolerances or thermal growth.

The vast majority of our pumps use back-to-back flush ground angular contact thrust bearings of this type. They are all flush ground with no pre-load. Almost all of them have the oil level below the outer race with oil only delivered by the oil ring. And it is very, very rare for us to have a bearing fail from skidding of the unloaded bearing.

As far as skidding, I don't believe it would show up conclusively in the vibration. We usually find it by examining the failed bearing after the failure has occurred. But you are asking the right questions. You should analyze the vibration to see if there are signs of bearing fault frequencies that could indicate problems with balls, races or cages.


Johnny Pellin

 

[jimhokie]

Update: I've suggested they do the oil level test you recommended and questioned the preload. They are coming to the conclusion that the oil reservoir and surface area of the bearing housing are undersized given the heat input to the bearing, but we still want them to investigate any other potential cause or contributor. This is ongoing.

More details on the sleeve...I'm wondering if it is too thin and it's flexibility can be an issue. Here are the fits of the shaft, sleeve and bearing ID:
Shaft OD: 2.375 +0.0000/-0.0005
Sleeve ID: 2.375 +0.0005/-0.0000
Sleeve OD: 2.9531 +0.0000/-0.0002
Bearing ID: 2.9528 Max/2.9526 Min

Can the maximum 0.001 diametral clearance between the shaft and sleeve cause a problem with the 0.289 thick sleeve? Or could it be an issue of different thermal expansion coefficients that it creates a problem when heated? The shaft is 410 and the sleeve is 416, so I wouldn't expect much difference there.

Since the bearing housings for all units are already manufactured, we are looking at any easy-fixes to improve cooling of the housing. One thought I had was a coating to maximize the emmisivity, but I'm not sure how much a factor radiative cooling is or how much improvement this could offer. Any experience with this, or any other thoughts for additional cooling? They've proposed incorporating a fan onto the shaft flinger just inboard of the housing, but we have a number of issues against that and prefer a different solution.

 

[JJPellin]

The fit numbers you give are normal for the bearing to sleeve and quite tight for the sleeve to shaft. In general, the bearings we have in this arrangement have a much looser fit between sleeve and shaft. I don't believe fit is the issue. The thickness of this sleeve is not unusual at all. I don't believe it is too thin. We have bearings of this configuration running in vacuum tower bottoms service at 600 °F and higher with no cooling or unusual accommodations. We have bearings of this configuration running in coker charge service at 700 °F and higher with no cooling or unusual accommodations. The main difference is that our pumps have the oil level below the balls of the bearing with oil delivered only by the oil ring and bearings ground for zero pre-load.

As you describe it, you are in a test environment with bearing temperature monitored. Dropping the oil level is a zero risk test. This bearing, with these loads should be able to run for hours, or perhaps days with no oil delivered. Dropping the level for a few minutes will tell you what you need to know. I would strongly recommend dropping the oil level and trending the bearing temperatures. Flush ground bearings with zero pre-load might result in a slight drop in long-term reliability if you end up with some skidding. But it sounds to me like you need to get over this immediate short-term hurdle so that testing and development can continue. Get rid of the pre-load and drop the oil level below the bearing. I don't believe any coating could have a measurable affect.


Johnny Pellin

 

[abdul]

Jim,

Ball & Roller thrust bearings are not simple to troubleshoot, many factors can combine to cause unusual loading. However, you are correct in thinking that the high temperatures are due to some type of installation problem, and not merely due to increasesd axial hydraulic pump load. Here are several other major causes of high rolling element thrust bearing temperatures, so when studying these points, please look carefully at a x-section drawing of the pump:

1- Axial thrust load resulting from misalignment of the outboard bearing housing bracket to the pump casing, or due to runout of its flange face (non-parallel). Check brg housing on lathe, machine as necessary, and re-align to pump. Max tolerance = 0.001 inch runout.

2- Axial thrust load due to outboard bearing housing internal shoulders machined incorrectly, and the non-parallel face contacting the thrust bearings leads to high stresses on the balls during each rotation. Place housing on lathe and true up both active-nonactive faces in the housing. This includes the axial bearing housing cover, if this axial cover contacts the outer race of the bearing.Max tolerance = 0.001 inch runout.

3- Axial thrust load from a bearing-shaft sleeve, whose abutment face at inner race contact, is not machined/ground 90 degree to the sleeve bore. Max tolerance = 0.001 inch runout.

4- The inboard bearing outer race, if not allowed to float axially at least +/- 1 mm, will lead to a high axial load on the outboard bearing. Improper fitting can allow this inboard outer race to be locked in one direction by axial bearing housing covers. This bearing's OD should also have a 0.001 to 0.0025 inch diametral clearance to housing, otherwise thermal growth will cause it to lock in place.

5- As Mr. pellin has implied, your high rpm, and larger bearing OD, leads to high ball velocities. Therefore, using high viscosity oils leads to high friction, and high temperatures. I recommend using turbine oil ISO-VG-68, and not a higher viscosity for this application.

6- High axial thrust load from couplings; if its a gear coupling on the test stand, ensure that axial float of the spacer is 3/16 inch. If its a dry flexible element coupling, then look at checking for excessive thrust from motor; run the motor solo to establish its magnetic center to reduce axial force on pump. If its a dry coupling, and the motor has a thrust bearing, make sure that the coupling has a slight gap before bolting the flanges, such as 0.020 inch, to allow for thermal growth of both shafts into the coupling. This thermal growth leads to high thrust force on the pump thrust bearing.

A very useful and practical book: "Essential Concepts of Bearing Technology", Fifth Edition , by Tedric A. Harris :



Be sure to come back with your findings,


Abdul

 

[jimhokie]

Thanks for the additional input, Johnny. I will continue to pursue varying the oil level and eliminating (or at least reducing) the preload. They are worried about ball skidding, and long-term reliability is an important parameter. Currently, they do plan to run a test with zero preload on the bearings. I'll let you know the result.

And thanks to Abdul for all the additional points to investigate. I will attempt to have our tech library purchase that book.

I will be traveling again the next two days, and will follow up as I learn more (or have more questions!).

Jim

 

[jimhokie]

Update: When they tested with 0 preload vs. 500 lb preloaded bearings, the stabilization temperature improved, but still exceeded the specification rise above ambient. At the same time, they corrected a slight misalignment at the radial bearing housing end that was causing a rub, noticable by the shaft coming to an abrupt stop near the end of the coast-down.

Regarding the overlubrication possibility, after much resistance, they used a plexiglass end cover and strobe light to visually confirm excessive oil in the bearing, then followed up with lowering the oil level and further reducing the stabilization temperature. However, it is still not meeting the rise above ambient requirement. It is looking more like the passive cooling capacity of the housing is just insufficient given the heat input of this bearing, so they are looking at possibly increasing the housing size, including the addition of cooling fins. They are also looking deeper into any other possible alignment problems that may still be contributing.

Thanks greatly for all the expert tips to point us in the right direction!

Jim

 

[chief]

You can reduce the thrust on the bearing by fitting a balance tube between the suction and discharge nozzles. Also it may be worth looking at the lubrication qualities of your oil, a lighter viscosity oil will run cooler.

Offshore Engineering&Design

 

[jimhokie]

Thanks for the ideas, Chief. The pump does have a balance disk, which wasn't working properly at first, but they've solved that problem. But there is still some overheating due to the size of the bearing and operating speed, with too little cooling capacity. A lower viscosity oil may address that problem though, so I'll look into that possibility. Thanks again!