Bearing Reliability - 4 Key Learnings

From Imrtwiki

Return to Medium or High Speed Rolling Element Bearings

Contents 1 Bearing Reliability – My 4 Key Learning’s 2 First Key Learning – Bearing Natural Failure Pattern 3 Second Key Learning – Bearing Rarely Fail from Natural Causes 4 Third Key Learning – Bearing Load and Bearing Life Relationship 5 Forth Key Learning – Bearing Re-Lubrication 6 Conclusion

Bearing Reliability – My 4 Key Learning’s

My first job after I finished my initial engineering training was in the maintenance section of the Port Kembla Steelworks Foundries Department and two things from that short experience convinced me that I should make equipment maintenance my career. The first was doing the maintenance supervisor role and really enjoying working directly with a great group of guys including some very good tradesmen. The second resulted from two bearing failures that occurred in the main sand recovery plant. As these failures had caused some operational pain, I decided that I would determine the expected life of the bearings so that they could be replaced before they failed the next time. I grabbed the SKF bearing manual from the maintenance office shelf and started reading. What I learn from the manual and some other reading, started me on a lifelong interest in maintenance technology, bearings, condition monitoring and lubrication.

At this early stage of my career I had a very typical attitude, “Its only equipment maintenance. How hard can it be?”. You sometimes find this attitude in management, operational and other people who have not had first hand experience of trying to control the reliability of equipment. They tend to associate equipment failures with ‘maintenance incompetence’. People with this attitude tend to show little interest in maintenance when things are going well and sometimes anger when failures occur.

First Key Learning – Bearing Natural Failure Pattern

As mentioned, my first key learning about bearings was on trying to determine the expected bearing life of the failed sand plant bearings and this came from the shock of finding that my assumptions of how bearings Wear Out was completely wrong. This experience opened my mind to the complexities of what it takes to manage equipment reliability. What I found was that bearings don’t Wear Out. By Wear Out I mean that after a set amount of usage (Millions of Revolutions) the likelihood of a bearing failing will start to increase (see above figure). Bearing manufacturers have run to failure 1000’s of bearings in test rigs and found that bearing failed randomly through rolling surface fatigue. This means that there is not a point in time where the likelihood of bearing failure increases but it stays constant over time (see figure). When bearings are installed by experts and are well looked after, the failure pattern is random (like tossing a coin) and so asking how long a particular bearing will last has no sensible answer. Bearing are typically given an L10 Life in Millions of Revolutions. L10 is the life at which 10% of a group of bearings will have failed, where they have identical service conditions. If you had a group of bearings that had been in service for 40 years and the equipment item came out for an overhaul, most people would replace the bearings. The reality is that if a bearing has no faults it can be treated as if it is a new bearing and can be reinstalled. If a bearing is worth less than $100, it is unlikely to be worth reusing it but as bearings often cost many hundreds or even thousands of dollars, it should be strongly considered.

I vividly remember a large steel scrap chopper that had been in service for a many years and my maintenance planner for the area wanted to do a ‘strip and inspect’ in the overhaul shop. My expectation for the cost of the job was $5,000 to $8,000 but the cost estimate that came through after the strip down was about $25,000. I immediately rang the shop to find out why. It seems the shop tradesmen had assumed the bearings would be replaced with new ones and had oxy cut the old bearings off the shafts. The bearings were large and of an unusual type and were around $4,000 each. Previous Condition Monitoring had shown that the bearings were OK and they did not even have to be removed from the shafts.

The huge opportunity from bearing random life patterns comes from resisting the desire to overhaul equipment just because its bearings have seen good service. You should rely on condition monitoring (CM) to give you the confidence to keep equipment operating. If an equipment item is to be removed from service for an overhaul, get a vibration analysis done on the bearings before they are removed. This should increase your confidence in being able to reuse the bearings after visual inspections have also confirmed there are no other faults. Where in-service CM is not effective or practical and bearings replacement costs for a particular type of equipment are high, you could consider a simple bearing test rig in the overhaul shop and to do vibration analysis on the bearings so they can be reused.

Second Key Learning – Bearing Rarely Fail from Natural Causes

My second major learning about bearing reliability took a while longer than the first. For seven years in the 1980’s I worked at the Power Station and Utilities Section in the Port Kembla steelworks and one of my roles was setting up and supervising the condition monitoring system for its rotating equipment. This CM program discovered many rolling element bearing faults and I was diligent in investigating the causes of these failures. I still have a 900mm wide draw in my office full of the damaged bearings I recovered. In all these bearing fault investigations there were always causes other than natural rolling surface fatigue driving the bearing deterioration. The observed failure pattern is an Early or Infant failure pattern (see figure).

The typical bearing failure causes from this Power Station and Utilities equipment were from incorrect equipment or bearing assembly, water washout of lubricant, no re-lubrication, excessive force and false Brinelling (external vibrations when bearing is stationary). Many of my reliability peers in other industries had similar experiences of Early Failure patterns dominating bearing reliability and it is also well documented in maintenance literature. Its not that the bearing rolling surface fatigue is not active, it is that other failure causes dominate the failure pattern. Bearings that have no failure causes other then normal rolling surface fatigue tend to be the ones that are still in service and running well. There is no shortage of causes for early bearing failures and they are generally related to bearing assembly errors, equipment installation errors, operational abuse, lubrication issues or contamination problems.

As all early Life failures for bearings are avoidable, they represent the biggest opportunity for reliability improvement and cost reduction.

The figure above indicates that if a bearing reaches a certain life then it will last forever. What I am suggesting is that where avoidable bearing failure causes have been eliminated, equipment will often be removed from service for other reasons before bearing rolling surface fatigue failure occurs. 100years or more service is not out of the question. The next question is, which of your bearings will have this very high level of reliability?

Third Key Learning – Bearing Load and Bearing Life Relationship

If a bearing has reached the end of its early failure period (this might be a number of years) without faults being initiated then there are two key drivers for reliability. The first is lubrication, which I will deal with as the last key learning. The other is a bearing in-service load in comparison to its designed load capacity.

The effect of bearing load on rolling element bearing life is well known. The L10 life of a bearing is calculated by L10 = (Dynamic Load Rating/ Service Load {dL/sL}) to the power of 3. In simple terms this means that if you double the load on a bearing you reduce the life by a factor of 8. A fact that is less well know is that doubling the load on a bearing will also reduce the warning time (PF Interval) given by condition monitoring by a factor of 8. Thus reducing the load on a bearing (eg. by improving coupling alignment) can have a huge positive impact on bearing reliability. This reliability improvement is shown by bearing theory but it has also been proven in practice and documented by many thousand of people around the world. The practices required to achieve this improvement are well understood by the condition monitoring and reliability community, with one of the most widely used bearing force reduction approaches being improved coupling alignment.

If you are interested in understanding the expected service life of a bearing you need to estimate its dynamic load to its service load (dL/sL). The dynamic load is a design capacity number that is available from the manufactures of a bearing and the service load is the actual normal load the bearing will experience in operation. The detailed method to do a bearing service life analysis is given by the bearing manufacturers but the following simple approach is a good starting point. If a bearing dL/sL is 6 or less then it is considered highly loaded and will be susceptible to failure and should be condition monitored rigorously. An example of this is with a number of modern types of gearbox designs, where if you install a 100kw motor onto a 100kw gearbox and then run the motor at full rated load, then the gearbox will likely to be prone to bearing failures. Issues like shaft alignment and lube cleanliness become especially critical to achieve reasonable levels of reliability. Contamination particles in the oil down to 3 microns in size need to be filtered out both before commissioning, during any oil top-ups and while in-service.

I became interested in how reliable bearings can be after I set up some return oil line magnetic chip collectors for condition monitoring of five of the gearboxes in the tin plate rolling mill where I worked. This recirculating oil system lubricated 16 gearboxes in total. All the return line magnetic plugs showed huge amounts of collect steel debris (much more than shown in the picture). Chatting to my mechanical planner from the mill we determined that the metal was coming from rotation of the outer bearing races in the very large pinion gearboxes that drive the mill work rolls. The filtration on this recirculating oil system was by a cleanable fine wire mesh type filter with an additional magnetic capture system. As contamination such as this is well known to significantly reduce bearing reliability, upgrading the filtration system sounded like a good idea. These gearboxes had been in-service since the early 1960’s when the mill had been built, so I searched for previous gearbox bearing failures from both the maintenance computer records and the older card rack maintenance history system (The old manual card history was much easier to use). To my surprise I found no evidence of any previous bearing failures in these gearboxes, even with the serious levels of steel debris contamination entering the bearings for probably over 40 years. See the picture to the right of the typical bearing failure mode for solid contamination. With no bearing failure history I could not justify upgrading this filtration system as we had lots of other improvement priorities.

A reason why rolling element bearings are often very reliable is that actual bearing service loads are often significantly less than the bearings design capacity (dL/sL is high). The reasons typically are:-

• Design engineers are often conservative in their estimates of average service loads as it is difficult to estimate what they will be. This is becoming less so with some of the computing tools now available to design engineers.
• Many designers treat L10 bearing lives as the actual life for rolling element bearings.
• Many systems don’t actually run at full design load.
• In many situations larger bearings are selected for reasons other than bearing service loads.

A common example of the last point is where the designer chooses the bearing to match the applications required shaft size and this often gives a larger bearing size than is actually required by the load. Always pay attention if a shaft has a step down to the bearing size (as per the figure to the right) as this makes it more likely that the bearing has been selected on the bearing loading and so will be more sensitive to bearing failures. This is a legitimate design approach.

Going back to our bearing reliability story with the rolling mill. One fantastic characteristic of steel as an engineering material is that there is a stress level below which fatigue failure will not occur. So there is also a bearing load below which bearing rolling surface fatigue will not occur. This typically occurs when the dL/sL is more than 16, when some minor lube contamination levels are taken into account. As the mid 1900’s rolling mill designs were very conservative the pinion gearboxes were made very solidly. My estimates were that the pinion gearbox bearing dL/sL number was well above 16 and the bearing surface damage from the mild steel debris would be less severe as it was softer than the bearing metals. I concluded that this was why the bearing reliability was high. I later found a similar situation in the hoist gearboxes of a very large drag line I was doing a bearing inspection on during its overhaul. All the bearing had a very large number of galling lines (metal to metal pickup) across all the cylindrical roller bearing rolling surfaces caused by the gearbox shafts slipping sideways, probably caused by the slewing the dragline with load off the hoist ropes. As the galling lines would have caused significant stress raisers in bearing surface, I was surprised to find no spalling defects in the bearing rolling surfaces around these lines. My conclusion was that the dL/sL numbers were high and so surface fatigue did not occur even after extended service. Another example of achieving very high levels of bearing reliability is with pump bearings in petroleum plants that have sophisticated reliability practices. Once issues such as alignment, balance, bearing installation, lubrication, soft foot and pipe strain are being well managed, pump bearing failure cease to be a problem and pump life is determined purely by seal life.

When I examine a bearing reliability issue the first thing I do is look up its Dynamic load rating. I then think about what the actual in-service loads that might be occurring and ask myself is it closer to Dynamic Load/6 or Dynamic Load/16 (eg sensitive or safe). To estimate the in-service loads I always start with a quick calculation of the static loads on the bearing such as from the shaft and rotor weight, V belt loads, etc. Then I think about other bearing load from issues such as couplings forces, gear loads, rotor pressure differentials and vibration levels. Many bearings are failure sensitive to axial loads, so this has to be considered as well. If there is condition monitoring history of bearing deterioration, I look at the speed of the deterioration. Fast deterioration is a symptom of higher bearing loads, often from issues like bearing installation error or excessive misalignment. The extrapolated warning time (PF interval) can be up to 10% of the L10 bearing life. I suggest all Reliability Engineers should learn how to calculate L10 bearing lives and these detail are readily available through the bearing manufacturer’s literature.

If a bearing has a dL/sL number of more than 16, then the decision for re-use of the bearing during an overhaul can be made with only a detailed bearing visual inspection, as condition monitoring of early bearing surface fatigue faults should not be required.

Forth Key Learning – Bearing Re-Lubrication

I have discussed the potential for some rolling element bearing to outlast the equipment that they are installed in. In comparison the lubricant that enables the bearing to have such a long life does not tend to last forever and so many bearings requires re-lubrication. There are two types of lubrication strategies. The first is Lube Once bearings such as for sealed bearings (‘Lube and Forget’). Also in this group are all the other ‘Forgotten to Lube’ and other ‘Unable to Lube’ bearings. The second lubrication strategy is where re-lubrication can and is applied. Lube Once and especially ‘Forgotten to Lube’ bearings tend to reveal themselves in unexpected ways at between 3-5 years for higher speed 24hr service machines and at 8 to 15years for low speed equipment. These times are significantly less if the service temperatures are above 80 Deg C. The solution for ‘Forgotten to Lube’ bearing failures is obvious but requires lots of had work in finding all the forgotten bearings and should be carried out by front line lube techs with access to machinery drawings and given the time to do this work. ‘Lube and Forget’ sealed bearings are often an easy design solution but can in many situations be a real barrier to maximising equipment service life and achieving minimum Life Cycle Cost. As an example many smaller motors with sealed bearings have to be routinely replaced at about 3-5 year in 24 hr service situations and this produces all the consequential risks of early failures on change-out. If these bearings could be re-lubrication they would likely run 2 to 5 times this service period.

One requirement for a good Condition Monitoring program is the ability to detect symptoms of inadequate lubrication. Vibration analysis programs have a range of different parameters that can be used to detect bearing lubrication problems and some of these are better than others. If you have a good lube CM parameter, up to 70% of the faults detected by your program are often lubrication stress symptoms. My suggestion is if you experience unexpected bearing failures for bearings rotating faster than 600rpm and you observe some cage wear on disassembly of these bearings, you should look at adding some better lube monitoring parameters.

The second lubrication strategy is where re-lubrication can and is applied. The major question here is how often and how much lube should be added. I remember asking a very experienced lubrication engineer how much grease should we apply to a specific bearing application. My expectation was that he would go to a table and read off a number. The frustrating answer I received was a question “How much are you currently adding and are there any problems?”. There are re-lubrication charts available with variations for service temperatures but if you compare this to the quantities of grease in a sealed bearing and the service time they achieve, the lube chart quantities look excessive. Also if you look at many heavy equipment applications with automatic re-lubrication systems, the re-lubrication quantities suggested by the charts seem very low.

Whatever your re-lubrication practice is, you can probably reduce your lube usage and cost. To do this at low risk you should have some good lubrication CM parameters. Ultrasonic listening instruments are one technology that can be used. You should also take the opportunity to inspect any bearings on equipment that is being overhauled or repaired. Cage wear is often a sign that the re-lubrication interval or quantity may not be adequate when no significant levels of lube contamination is observed.

Conclusion

The four key learning’s for controlling bearing reliability are:-

• Understanding the bearing natural random failure pattern
• Understanding that bearing rarely fail from natural causes
• Understanding the bearing load and bearing life relationship
• Understanding bearing re-lubrication requirements

The key thing to remember is that all early Life failures for bearings are avoidable and they represent the biggest opportunity for bearing reliability improvement and cost reduction.

Article by Peter Todd - Industrial Maintenance Roundtable Facilitator NSW

Note - The dL/sL symbols for the bearing 'dynamic load/service load' used in this article is not standard. Bearing suppliers use C/P, which is a bit hard to interpret when it is used by itself. Also in actual bearing life calculations the service load parameter (P) is more complex than I have suggested. For example it is now common to allow for contamination factors when calculating bearing life.