Internal Clearance in Angular Contact Ball Bearings 2

[JJPellin]

We have experience a number of unexplained failures of angular contact ball bearings. We are attempting to understand the issues involved. I am hoping some bearing experts and pump users can help me separate the wheat from the chaff.

A number of years ago, we found we were experiencing high failure rates in the thrust bearings in some hot service pumps. The typical example would be our coker charge pumps. The bearings were MRC 7309 PDU 40 degree angular contact ball bearings mounted in a back-to-back arrangement with an ISO k5 fit to the shaft. These bearings are ABEC3 and ground for flush mount with no internal clearance. Installed with an ISO k5 fit (interference fit) to the shaft results in a light preload. A few years before, we had redesigned the pump to improve shaft deflection by moving the bearing span in and making the shaft a larger diameter. This resulted in more heat conducted down the shaft to the bearings. The pump service is approximately 725 °F. The bearings are lubricated with pure lube oil mist with directional reclassifiers.

Inspection of failed bearings found evidence of overload. The normal thrust load for the pumps is very low compared to the design load limit of the bearings. We concluded that the shaft thermal growth, in combination with the PDU bearings and the ISO k5 fit was resulting in excessive preload. We converted to a bearing with slight internal clearance (SKF 7309 BECBM). Our failures immediately ceased.

We needed to decide how to apply these bearings with increased internal clearance. We didn’t like the idea of stocking all of our 7000 series thrust bearings in both designs (PDU and BECBM). We did some research. We found that some major API pump manufacturers actually recommend BECBM bearings for all of the pumps they sell. We decided that the increased internal clearance could help up in pumps where hot service might result in excessive pre-load. We believed that the BECBM bearing would not do any harm in the cold service pumps. So, we converted all of our warehouse stock for these bearings over to BECBM and started installing them in every API pump that came into the shop.

Recently, we have had a couple of pumps where the BECBM bearings seem to be causing failures. We pulled out some pumps where a PDU bearing had run for 6 or 8 years. We converted to the BECBM and had immediate vibration problems and repeated bearing failures. Examination of the failed bearings showed signs of skidding. We are speculating that the increased internal clearance in the BECBM bearing may be resulting in skidding because of inadequate load. So, we are being pushed to convert all of our stock back to PDU.

We have a number of options. We could divide the population into cold service and hot service and use the PDU bearings in the cold applications only. We could convert everything back to PDU other than the specific hot service pumps that led us to make the change. We could leave the BECBM bearings in all of the pumps and only convert back for pumps that have a demonstrated problem with skidding. Or, we could apply alternate technology such as “PumpPac” bearings that use a 40 degree contact angle for the active bearing and a 15 degree contact angle for the inactive bearing. I suspect that I may be advised that we should analyze each pump individually to determine the likely pre-load, normal thrust load and apply the best bearing for that pump. This sounds very good, but we have about 2000 pumps and the investment in engineering time to analyze them individually would be a large burden.

I would like to hear about others who have experience with this issue. I would like to hear from bearing experts about the proper application of these different designs. Any help will be greatly appreciated.

Johnny Pellin

 

[spciesla]

Even though you have 2000 pumps, can you try a divide and conquer strategy. Group the pumps in families to reduce the amount of analysis needed to determine the "best" bearing for the pumps. Hopefully you will find that the majority of families clearly fit in either the PDU and BECBM design and you will only have to analyze the remaining families/pumps.

 

[electricpete]

I have very little experience with angle contact bearings. Your experience sounds pretty close to exactly what the literature would tell us to expect:

Too tight (including effects of interference and heating) -> just a little expansion below zero internal clearance can result in rapid failure

Too loose -> reduced life due to skidding, but more room for error (better to be too loose than too tight).

Reducing the contact angle to 15 degrees on unloaded bearing while keeping the 40 degree steep angle on loaded bearing allows you to keep the looser clearance with less problems from skidding on the unloaded bearing. On the surface (based on cursory knowledge from literature), it sounds like that's exactly what you need... something like MRC 8309. I have no experience with it though.

Skidding depends in part on D*N due to ball gyroscopic effects. Sounds like you have pretty small bearings 45mm bore which should work in your favor. What is the speed?

=====================================
(2B)+(2B)' ?

 

[electricpete]

Also, I was assuming you have lower thrust requirements in one direction.

=====================================
(2B)+(2B)' ?

 

[Tmoose]

I guess normally I'd try real hard to follow the OEM's recommendation. But if I read your post right you have strayed away from the original design a bit.
How well are you controlling shoulder, nut, and spacer face runout and perpendicularity?
What type of oil are you using? In some applications synthetic oils can broaden the survivable operating temperature range substantially.

My spindle oil mist tech info is buried someplace, but I sort of recall that like air-oil (some say oil-air) there is some latitude in volume allowable for helping with cooling, even though unlike air-oil there may not be nozzles that need to be positioned precisely.

I'd expect the bearing manufacturer would be willing to review and evaluate your design for little or no expense.

Are you (or the OEM) cooling the housing? That can really exacerbate thermal preload issues.
Do these kinds of pumps ever use a shaft heat flinger/slinger?

Widely separated Rigid mounted Back-to-back bearings generally are more tolerant of thermal expansion than other arrangements because the axial growth is likely to greater than the radial growth, relieving some of the preload. There is a magic spacing that theoretically cancels nearly all thermal effects when fully warmed up. (But big differences in housing and shaft temp can render that dimension invalid. >>If<< the original design happened to have the magic dimension By moving the bearing span in you may have lost some temp tolerance.

A slightly esoteric vibration measurement (something based on gs, or one of the demodulated technologies) during test run in might be helpful in identifying loose or tight, hot or cold service units.

 

[JJPellin]

I will attempt to answer the questions you have asked and I will add some additional information to help clear up my question.

The decision to standardize on the BECBM was made by a team of engineers following research on the subject. At the time we had an alliance with a major world-wide manufacturer of API pumps. They recommended the BECBM. Every API pump that they deliver in the US comes with BECBM bearings installed. They were operating under the same rationale expressed by ElectricPete that it is better to be too loose rather than too tight. We talked to a major OEM of steam turbines. They make a line of small single stage turbines that use angular contact ball bearings. Every turbine they deliver uses BECBM bearings.

Plus, I should remind everyone we are talking about clearance differences in the range of microns. This was not a major design change. The decision to standardize was not made just for the relatively minor cost of double stocking. We currently stock 33 different stock items for angular contact ball bearings. We could have added BECBM bearings in all of the common sizes (7306 to 7316) for perhaps US $10,000. We were also concerned about mixed pairs of bearings. We have experience with the problems that can arise by installing two different bearings in a back-to-back pair. Stocking both types increases the chance of this error occurring. But, we were primarily motivated by a desire to avoid the failures that we were seeing. We knew for certain that we were experiencing failures in critical pumps because of inadequate clearance. We had a major API pump manufacturer telling us to standardize and we had some valid reasons to want to avoid double stocking all sizes.

Most of the pumps using these bearings are single stage overhung API process pumps. Most of them have back wear rings and balance holes which should result in a stable thrust in a known direction. But, there are some pumps with no balance holes. Others have back pump-out vanes. We also have a considerable number of pumps with double suction impellers in a between-bearings configuration. These have very little thrust and it is not always in the same direction. The vast majority of our API pumps run at 3600 rpm. Some run at 1800 rpm. Very few run slower than this.

It is not really a question of straying from the original design. Most of our older pumps would have been built with old-generation angular contact bearings using stamped steel cages. There is good data to prove that this is a poor choice. These are generally considered obsolete. Some of our pumps were originally supplied with double row bearings that could only be built with steel cages. Reliability expectations were much different in 1955 than they are today. If we insisted on sticking to the original design, we would have non-cartridge mechanical seals (or braided packing), lip seals, brass oil rings, greased gear couplings, jacket cooled housings and open-cup lubricators. It would be impossible to achieve modern expectations for reliability with these pumps without using some standardized upgrades to proven superior technology.

I like the suggestion by Spcielsa. I will probably end up dividing the population by pump arrangement and process temperature and trying again to get the best standard bearing for each group. At that point, we just have to look for the odd outlier that might require a detailed analysis.

We typically use mineral oils for ring oiled applications. Our oil mist is high grade synthetic. We have eliminated jacket cooled housings based on the published literature. Many of our pumps use oil mist lubrication. Most of the others use oil sumps with non-metallic oil rings. Some of these have oil coolers in the bottom of the sump. Most rely on air cooling from a housing with external fins. We have very good specs and repair practices with regards to shaft fit, perpendicularity of faces, etc. We torque our lock-nuts. Our precision repair practices have advanced to the point where I would no longer be concerned about our mechanics installing mis-matched pairs of bearings.

I was interested in hearing from other users of API pumps that might have experienced problems with either option (BECBM or PDU). Do other users simply put back whatever bearing was originally supplied? Do any of you have experience with the use of the 40/15 contact angle bearings? How do you determine where to apply them?


Johnny Pellin

 

[caoimhin1]

While it would probably involve a major redesign, a four point contact bearing, e.g. QJ306, would eliminate the skidding effect of an unloaded row.
Btw, when you say that the increased heat caused excessive preload, you mean radial preload?

Also, don't limit yourself to SKF, it would be worth talking to bearing OEMs such as Schaeffler and NSK, both of whom have specialised API angular contact bearings.

 

[JJPellin]

I have never heard of a four point contact bearing. I will do some research into that. Because of the contact angle in these bearings, I believe that any preload will be both radial and axial. But you are correct that thermal growth of the shaft creates a radial preload.

Johnny Pellin

 

[electricpete]

Most of my experience is vertical motors. When you factor in machine weight, most (not all) have heavier axial thrust in downward direction, with upthrust occuring only in smaller magnitude and brief transient periods. In that application it makes sense that the bearing like PumpPak should be used.

Since 4-point contact bearing was mentioned, I’ll talk about one machine we have that uses it. We have an 800hp 1800rpm vertical motor driving vertical pump (don’t know the pump type... sorry) where pump has no thrust bearing of its own and follows the pattern I said above (mostly downthrust). The motor lower bearing is a guide. The motor upper bearing is a pair / set called MRC 97320UPD

It includes two bearings:
MRC 9320UP = 100BZ03 = “4-point contact “ = “gothic arch” bearing.
Described here:

http://www.skf.com/binary/12-97842/...&WT.z_oss_boost=0&tabname=All&WT.z_oss_rank=3

If you took a cross section and looked at it in the orientation of first page upper left hand corner, and exagerated the curvature, the rings look like gothic arch rather than portion of a circle. Where the race contacts the ball is not 2 points 180-out, but 4 points as depicted.
The shape provides ability to have steep contact angle in both directions.
Unfortunately to assemble this beast, it requires split inner ring. That’s ok because balls don’t run on the split which is at the centerline.

7320P = 100BZ03 + 100BT03
angle contact bearing in tandem, for matched use when purchased as part of set 97320UPD

The outer rings are clamped tightly by the housing at both the top and bottom of the pair. So any movement must be endplay of the 9320UP bearing. We measure 0.030 – 0.040” repeatedly on several different motors, which surprised us. It is apparently the internal clearance of that bearing.

We have historical high / variable 1x vibration on these machines. I realize that can be a lot of things, but the most knowledgeable motor repair shop engineer I know said it is due to that bearing configuration in this application (vertical motor). He said he always sees this symptom on vertical motors with this bearing configuraiton, and he has had huge success installing the pump-pak bearing in place of this configuration, although we haven’t done it yet at our plant.

You have seen the adverse effects of too tight and of too loose. So you can try to match the tightness better to the application (hot or cold) as one approach. Also if you have applications that have significantly lower magnitude/duration of thrust in one direction (like most vertical machines), then you can gain a lot of margin for error/variability (example – variable temperature) using the shallower contact angle on the non-loaded bearing. At least that what logic tells me. If it's not easy to classify your pumps according to thrust, it may not be of much use.

=====================================
(2B)+(2B)' ?

 

[electricpete]

By the way, I'm still curious, how fast do your machines run?

=====================================
(2B)+(2B)' ?

 

[JJPellin]

It was burried in my last reply. Probably 75% of our pumps run at 3600 rpm. Most of the remaining 25% run at 1800 rpm. We have a few that run at 1200 rpm.

Johnny Pellin

 

[JJPellin]

I have received confirmation from two other plans owned by our parent corporation regarding this issue. One is a large oil refinery on the Texas Gulf Coast. The other is a large chemical plant in the Chicago area. Both of them use BECBM bearings as the standard for all API pumps. The refinery uses 40/15 contact angle bearings for applications that have demonstrated a skidding problem. The chemical plant uses a pre-loaded bearing similar to the PDU if skidding is seems as an issue. I am still waiting for a response from some bearing manufacturers.

Johnny Pellin

 

[BearingDave]

The k5 machining tolerance on the shaft is typically standard for centrifugal pump applications. What tolerance are you using in the housings?

I know you are thinking preload is leading to issues in the bearings, however, I have heard that BEGAM paire angular contact ball bearings (CB = normal axial clearance, GA = slight axial preload) helps to deal with the skidding of the unloaded bearing. Pumpac is indeed another choice at your disposal.

 

[JJPellin]

Our housing fits are H6.

Johnny Pellin

 

[BearingDave]

JJ,

H6 is standard in most pumps for the housing bore tolerance. So your shaft and housing fits appear to be good.

Please find the attached PDF from SKF detailing the different options (covers BECBM, BEGAM, PumPac, etc.)

https://www.dropbox.com/s/os0jfdxy6ehbxh8/SKF_Bearings_in_Centrifugal_Pumps_Application_Handbook.pdf

In my experience I have seen BEGAM slightly preloaded angular contact ball bearings pairs fix some problems with the skidding of the unloaded bearing, since the preload guarantees a certain amount of load on the unloaded bearing.

Let me know when you have downloaded it so that I can remove the download link.

 

[JJPellin]

I wanted to give an update on this problem.

We have continued to find pumps that experience high acceleration levels when converted from the PDU to BECBM type bearings. We have had a chance to convert some of them back to the PDU and confirm that this solved the high acceleration. We have developed a tool for evaluating the best bearing for each application and have some unexpected findings.

Vibration Limits

This problem really came to a head because we implemented vibration limits for our pumps that include acceleration as well as velocity. With acceleration limits, we find that we are uncovering issues with chronic cavitation in some pumps and bearing skidding in others. If it was not for the acceleration limits, we would probably have not recognized the problem with the BECBM bearings until the failures started to stack up.

Analysis Tool

I had suggested before that it would be a burden to try and analyze 2000 pumps for the ideal bearing selection. But, we came up with a way to ease the pain. We developed a simple tool to calculate the net thrust on pumps of several standard configurations. The tool uses the geometry of the impeller, wear rings and mechanical seal to sum up the thrust forces. The tool is based on tools provided to us by our two major API pump manufacturers. We have not been attempting to use this tool to analyze all pumps. But, as part of the scoping process, we have been using it to calculate thrust on pumps as they fail.

Calculated Thrust Loads

For many years, I have been making decisions based on the common assumption that most pumps thrust toward the suction. This may not have been a good practice. We are finding that a great number of our pumps normally thrust away from the suction. Some thrust in reverse only under some operating conditions. Others thrust in reverse over the full range of the curve. The amplitudes of the thrust have also been surprising. Most of the single-stage, overhung, API process pumps we have analyzed have a very low net thrust. If you compare the expected thrust to the minimum recommended thrust for a 70 degree angular contact thrust bearing, we have a problem. These pumps do not even thrust enough to provide good ball tracking on the loaded bearing, much less the unloaded bearing.

Bearing Selection

Based on the direction of thrust and the expected net thrust loads, we have been selecting between bearings with slight internal clearance (BECBM), bearings with slight pre-load (PDU), 70/15 degree pairs (PumPac) and 15/15 degree pairs (DiamondPac). As I noted above, we have successfully solved some of the high acceleration issues that we had created with our standardization on the clearance bearings. We will be tracking these decisions to see the effect on vibration and bearing life over the next few years to fully evaluate the results of the changes.

Pump Manufacturers Recommendation

Ironically, our largest API pump manufacturer just came out with a technical bulletin recommending that their customers continue to use bearings with additional internal clearance (BECBM) in order to combat darkening oil from skidding. Our other major API pump manufacturer still supplies BECBM bearings in every API pump that they sell. I think that these Pump OEMs are steering their customers in the wrong direction in many cases.

I am considering the possibility that this whole effort could be worthy of publication. I am going to track the results and compile the data that I would need to build this into a technical paper for presentation at the Texas A&M International Pump Users Symposium in 2015. I am curious what level of interest might exist for this sort of paper. Any suggestions are welcome.


Johnny Pellin

 

[MikeHalloran]

With 2000 pumps in operation, you're certainly able to accumulate more significant data, faster, than any pump manufacturer, so you probably know more than any of them, at least about the specific pumps and specific operating points you've investigated.

So yes, you should be preparing a paper, if not for presentation then at least for the benefit of your successors at your location.


ISTR that single suction pumps are sometimes equipped with shallow impeller vanes on the back face, so as to develop enough pressure behind the impeller to balance, or partially balance, the pressure developed across the front face of the impeller.
I'm thinking that if you're trying to evaluate the bearing thrust situation within a given pump, you should also be including information about that hydrodynamic thrust balance.






Mike Halloran
Pembroke Pines, FL, USA

 

[JJPellin]

MikeHalloran - Thank you for the encouraging comments. None of the pumps we have evaluated so far had back pumpout vanes. The vast majority of our pumps have back wear rings and balance holes. Our thrust tool has a section for back vanes. I will need to allow for a less accurate result for those cases.

Johnny Pellin

 

[MikeHalloran]

As I said, you're learning stuff that the pump manufacturers (or at least their external interfaces that you can find) don't know. I would love to see a paper come out of that.

I would also fully understand if you decide to not reveal a damn thing hereafter, because you're getting traction against your problems, and that may give you a competitive advantage that is too valuable to give away.



Mike Halloran
Pembroke Pines, FL, USA

 

[BearingDave]

Hi JJ,

I would definately be interested in reading a paper on your findings, should you choose to publish them.

I myself diagnosed pump failures recently, and in doing so communicated with two major pump manufacturers. They both indicated that some pumps are designed so that ideally the net thrust on the shaft in negligeable (no thrust acting in either direction). However, they say that in practice and depending on hydraulic operating parameters, the net thrust is usually towards the suction. They also provided insight that at startup, you may be seeing a backthrust in the opposite direction before steady-state conditions are established (another reason why the ACBBs are mounted in pairs).

I imagine with over 2000 pumps, you have some that have hydraulic conditions which cause the thrust to act towards the suction, and some to act in the opposite direction. Have you yet made a correlation on which pumps seem to fail, and what hydraulic conditions they were subjected to?