Basic Dynamic Load Rating values for same size bearing 1

[geesamand]

Fill me in - for the "C" value used in the classic unadjusted L10 calculation, why do top tier bearing manufacturers claim different values? Do the carburized construction actually give higher ratings than through-hardened construction?

This case is Timken style inch-size taper rollers.

 

[BearingDave]

History:

There is an equation to determine the dynamic load rating "C" of a bearing. It takes into consideration only the geometry of the rolling elements in a bearing. Things like the number of balls or rollers, the diameter of the balls or rollers, the length of the rollers, the contact angle, etc. are used. By this standard, if two manufacturers make a deep groove ball bearing with the exact same number of balls and the same diameter of balls, they should have the exact same dynamic load rating.

Here is something I found online that shows the equation to be used:

http://image.slidesharecdn.com/1-14...1/95/12-bearing-life-48-638.jpg?cb=1419674102

This is where it gets complicated.

Based on Hertzian contact theory, the C value, based on the geometry and equation described above, is the load at which the classic L10 life will reach 1 million revolutions with 90% reliability (which is to say if we apply to a bearing an equivalent load equal to the C value, then out of 100 bearings tested at this load, 90% of them will surpass 1 million revolutions before the first sign of fatigue (spalling) is visible on the surface of the raceways. 10% of them will not.).

The L10 life test specification is dictated by ISO. Criteria such as the lubricant viscosity (oil), speed and specific cleanliness parameters, are used.

However, L10 life is not only affected by the geometry of the rolling elements, oil and cleanliness. Other factors such as surface roughness of the raceways and balls, steel grade, steel heat treatment, osculation, etc. will also affect the life of the bearing.

Bearing suppliers therefore felt that the improvements they were doing to the design of the bearings were not captured in the pure calculated geometry calculation method of the C value based on the equation. They started to add "compensation factors" to increase the C values for their bearings. In turn, the L10 life would increase as well (since L10 = (C/P)^3 for ball bearings and L10 = (C/3)^(10/3) for roller bearings).

Additionally, the bearing suppliers know that customers, when comparing bearing "quality", simply look at the load rating and pick the highest one. After all, this will result in a greater calculated L10 life based on the equation. Often in machine design, the customer will ask "minimum life of 100 000 hours based on L10", and therefore the designer can simply select the bearing with the highest load rating to reach that life requirement.

Lower tier manufacturer will tend to inflate their numbers by greater values in an attempt to get customers to buy their product under the pretense that theirs will last longer. Based on the L10 equation, it will last longer, but in reality, they rarely do (theoretical life vs. service life).

And so the war began. One manufacturer boosts their load ratings. Others boost accordingly because customers would get the perception their bearings were of lower quality. Once everyone changed their load ratings, others increased their values even more so to distinguish themselves from the others.

To justify the increases in load ratings, bearing manufacturers started to add additional compensation factors to the "C" value to take into account other parameters. In the fine print of their publications, they say "our L10 life is calculated considering a perfect cleanliness and perfect lubrication scenario". This means that they can greatly increase the "C" value as they please, as long as they stipulate the conditions in which the "C" value is valid. They don't come right out to say this however. Often times it is written very small in the catalogue, or it simply says to contact their engineering department for more information on the validity of the C values and the result of the L10 life calculation based on this C value.

Only two suppliers that I know of have had independent testing to validate their increased load ratings, and those are SKF and FAG. A third party has verified the values in the catalogue with tests and independently certified that the values are correct and not inflated beyong their means. This was only done for spherical roller bearings however.
Most suppliers consider the "C" value to be used in the L10 life calculation considering a 1.0 kappa ratio (oil film safety factor). Some low tier manufacturers, from China typically, offer a C value in the catalogue based on a perfect lubrication scenario (4.0 kappa ratio -- which is full elastohydrodynamic film lubrication).

Overall, do not be fooled by the number in the catalogue. The dynamic load rating values are simply numbers that are to be used in the L10 life equation, and that result is theoretical. Different manufacturers have changed their C values based on different testing conditions. We cannot compare apples to apples in this case.

Certain bearing manufacturers have gone to ISO to try and standardize the methods for C value calculation and L10 life calculation. However, standardizing this would have many lower tier suppliers not being able to compete with the high end tier suppliers on load ratings. This would result in lower tier suppliers going bankrupt, others not following ISO and therefore doing their own thing. The jury is still out on what to do on this matter to make everyone happy.

In your case of the Timken inch-sized taper roller bearings, it is definitely possible that they have changed the dynamic load rating value "C" to accommodate the perceived life increased of the case carburized heat treatment. Some suppliers believe case carburized results in longer service life. Others believe a through hardened steel will provide longer service life. Both arguments are valid, as which heat treatment will be best depends on the operating conditions in which the bearing is used.

Timken does not make a bad bearing. Not at all. In North America you would be hard pressed to find a better priced/quality ratio for taper roller bearings.

In conclusion, the thing to consider is stay with one of the major manufacturers (SKF, FAG, NSK, TIMKEN) and do not base your decision solely on a number in the catalogue.

If it was me, I would pick the bearing based on service level of the manufacturer, pricing, availability, and also which company is best in different types of bearings.

For example:
SKF is great at spherical roller bearings.
FAG is great at cylindrical roller bearings.
NSK is great at deep groove ball bearings.
TIMKEN is great at taper roller bearings.

Every supplier has their strong suit.

 

[geesamand]

Well our situation is we build gearboxes for a specific kind of application. We have a range of them, each with a range of ratios. Customers ask for basic service factor and L10 life for any application. We use a commercial design software to calculate these values, and publish them widely within the company as we size each application. As it turns out, we have a line with Timken style bearings and all of our data are based on Timken C ratings. Timken pushed their C ratings up many years ago - but nobody else making inch size taper rollers did. We'd like to use NTN and NSK inch-size taper rollers as an option, but we can't do it without re-running all of the calculations and reducing the L10's.

Do you happen to know how Timken has put together elevated C ratings? I know they declare the C value to be the load where life is "1,000,000 revolutions" - surely in optimistic conditions - but they haven't disclosed their test conditions relative to ABMA or ISO 281 rating methods.

David

 

[BearingDave]

It is reasonable to believe that all manufacturers have increased their load ratings by varying amounts, and only the manufacturer knows the true basic amount (calculated with the geometry of the rollers) and the increase they have applied.

One thing you could do is ask TIMKEN if they can give you an indication of the increase % they have applied to the C value. I doubt they will give you this information. They keep that under wraps pretty tightly. Depending on the person you talk to they may not even be aware of how the load ratings are calculated. Still, worth a try.

The L10 life calculation method is not precise at all. We call it the Basic Life Rating for a reason. L10 does not take into consideration changes in misalignment, operating temperature, housing rigidity, surface roughness, cage design, contamination, and the list goes on. We are talking here about comparing basic load ratings between manufacturers who do not calculate the load ratings the same way, and even if they did, L10 is a statistical calculation of when a bearing will start to show spalling at the surface of the raceway due to sub-surface initiated fatigue. The quality of the steel for all major manufacturers today is so good that bearings rarely fail in sub-surface initiated fatigue, but rather other causes (installation, contamination, lubrication, etc.).

We can get a more realistic bearing life by calculating the stress/deformation at the raceway/roller interface. This is done via the finite element method. Advanced bearing modeling software enables a person to do this. Examples of advanced software like this are the BEARINX software from FAG, and the BEACON software for SKF. Those softwares are only valid for the bearings of the company who developed it however because only they have the data for the true design of the bearings (hardness of the rings, variation of the hardness through the ring, surface roughness, contact angle, etc.).

So, where do we go from here?

You are in a tricky situation. Your customer is asking a minimum of 100 000 hours based on L10. The main (only) variable you can play with in L10 is the C rating of the bearing.
You go with the TIMKEN ones, and it passes 100 000 hours.
You go with another brand, it does not pass 100 000 hours.
On paper the TIMKEN looks better, but in reality the other brand may end up lasting equal or longer.

I suggest you discuss with your customers to educate them that:
- Load ratings are not the same per manufacturer due to compensation factors
- Yes, some bearings are better than others, but the compensation factors applied have not all been verified to substantiate the increases
- Basic Life Rating L10 is a statistical calculation based on sub-surface initiated fatigue life of the bearing. 90% of the bearing will surpass this value by an average of 5 times.
- Bearings do not typically fail in sub-surface initiated fatigue conditions
- Invite other bearing manufacturers (you mentioned NTN, NSK) to answer the question "your load rating is lower than TIMKENS, therefore your bearing will not last as long. Why should I buy your bearing?"

Finally, you can potentially look at slightly modified L10 life calculations which does take into considerations more variables. Something like the L10m life rating takes into consideration the contamination level and lubrication oil film thickness. You can get an increase of up to 50 times the regular L10 result by tweaking those variables. SKF has a lot of knowledge on this as they helped develop it with ISO.
By using this method, you could boost all the bearings above 100 000 and then select purely on perceive quality, level of service, availability, and of course price.

This website explains the equations of the L10m as per ISO.

http://www.cwbearing.com/mod_ratinglife.html

Hope this helps

 

[geesamand]

BD, thanks again. I was able to obtain a couple of answers to my own questions.

Do you happen to know how Timken has put together elevated C ratings?

They re-rated inch-class taper rollers first in 1982. Details are in SAE paper 841121 and is academically rigorous in nature.
They re-rated inch-class taper rollers again in 2012. While it contains no technical explanation, it's discussed in this promo video:

https://www.youtube.com/watch?v=rVZHVjLRpYI

For example, a 46720 cone has a C90 rating (L10 of 3000h at 500rpm) of 17700lb today. From 1986-2012 it was 16400. Back in 1963 it was 15100.
Unfortunately for my company, we just took the higher ratings and updated all of our bearing rating tables. No sooner did the ink dry on that major effort, and the competitors came visiting and treating us well enough to consider them as an alternative. But the catalog ratings aren't the same.

In a nutshell, Timken does this by their own modification to the basic rating calculation - they added a material factor to the ABMA/ISO equation. This makes the increase appear in every rating and calculation. Reading the SAE paper, it appears they did diligence showing the improved material offers a life increase that relates to the 10/3 exponent.

It is reasonable to believe that all manufacturers have increased their load ratings by varying amounts, and only the manufacturer knows the true basic amount (calculated with the geometry of the rollers) and the increase they have applied.

I would agree, however it is not the case. NSK for example adheres to the ISO approach, which keeps material adjustments in the L10m column. They claim it's not as simple as just up-rating the basic load rating. NSK and NTN are strongly encouraging us to trust their proprietary calculation methods that are presented in a webpage, or performed by their engineers. We use a commercial geartrain design software that is based on standard L10 and L10m methods. If no manufacturer wants to provide data that conforms to the standard this makes our job difficult.

I suppose it's technically possible that both Timken and the others are correct, if the Timken method is most appropriate for case carburized bearings but not thru-hardened material used by the others.

I suggest you discuss with your customers to educate them that:

I have a large number of customers who understand machine reliability and would appreciate that we not only consider basic rating life, but advanced bearing life calculations. However my most particular customers are the least informed - engineering companies who design systems and want to both shop out equipment using specifications but also require full calculations with the order, often with a PE stamp on the bearing and gear calcs. These specs typically reference the old ABMA methods, and the engineers in control of each contract simply don't care enough or have time to listen to arguments to deviate from the spec after the contract is awarded. Or those engineers are too inexperienced to follow the argument and make a decision on behalf of their company that deviates from those specs. I think our only way to move past these antiquated specs is to declare our calculation method up-front, so if for example we choose to always include material adjustment to basic rating calcs, they agreed to it in writing.

The simplest solution is to qualify the alternate bearings on a per-bearing basis - if I can perform a rating calculation that is identical in all ways and representative of its usage - and show they are equal - then it would be reasonable to allow the alternates without adopting their individual calculation methods.

 

[BearingDave]

Hi Gee,

Tough situation you are in.

In a nutshell, Timken does this by their own modification to the basic rating calculation - they added a material factor to the ABMA/ISO equation. This makes the increase appear in every rating and calculation. Reading the SAE paper, it appears they did diligence showing the improved material offers a life increase that relates to the 10/3 exponent.

So, TIMKEN has increased the C value AND modified the L10m life calculation accordingly.
Other manufacturers have kept the C value the same, and modified the L10m calculation accordingly.
The modifications are different.
This means that if you take the C values from two suppliers as an input into your software, and calculate life based on a single L10 or L10m calculation, the results cannot be compared.

If no manufacturer wants to provide data that conforms to the standard this makes our job difficult.

Yes, I share your pain.

the engineers in control of each contract simply don't care enough or have time to listen to arguments to deviate from the spec after the contract is awarded. Or those engineers are too inexperienced to follow the argument and make a decision on behalf of their company that deviates from those specs.

What you say is the issue that is present in the bearing industry today.

Customers simply look for a life number to be reached regardless of the thought process behind the values used in the calculation
-- By this reasoning, are we to understand that we should simply select the cheapest bearing that has the highest load rating? Select a lower tier brand such as NACHI, KOYO, ZWZ?
Sigh. The above was a rhetorical question.

One could try to explain to the customer that significant cost savings could be obtained by listening your suggestions.
However, as you put it, many customers do not have time to do this. Bearings often make for a very small portion of a contract's cost and they can be seen as insignificant.
I've heard many explanations for why customers don't care and simply want us to "meet the spec".

The simplest solution is to qualify the alternate bearings on a per-bearing basis - if I can perform a rating calculation that is identical in all ways and representative of its usage - and show they are equal - then it would be reasonable to allow the alternates without

I don't believe you will be able to devise a rating calculation that is identical in all ways for all bearing suppliers. If you do, let me know, I would be interested in it!

You could however do the life calculation for each bearing based on that bearings' specific C value and that suppliers' L10m life calculation. In that manner you are calculating the fatigue life in the manner specified by the manufacturer, with the proper data. Doing this however we must take the leap of faith that the suppliers have not inflated the factors beyond the proper amount.

One way to facilitate this would be to use each customers' website calculation tool.

Overall, there is no way around it. You are doing a lot more work.

I think our only way to move past these antiquated specs is to declare our calculation method up-front, so if for example we choose to always include material adjustment to basic rating calcs, they agreed to it in writing.

I would agree that this is the only way in your situation for your particular customer base.

You need to have the flexibility to calculate life based on each suppliers' C and L10m method. By simply taking the C value and using the classic L10, you are not doing a true comparison between suppliers.

You need to add to the fine print that you calculate bearing fatigue life based on compensation factors for each supplier.

You could note in your fine print of your contract that if the L10m life calculations of two bearings are +/- 15% from one another, the choice of suppliers would be up to you, provided the suppliers are one from the following list:
and here you would list all major manufacturers you prefer to deal with.

On a deeper thought, there is too much faith put in the L10 or L10m calculations in my opinion. Much longer life improvements can be obtained by using proper shaft and housing fits, housing and frame rigidity, sealing, lubricant selection and re-lubrication quantities and intervals, as opposed to arguing about 95 000 hrs vs 100 000 hrs of "calculated life".

I'd be interested in your thoughts on the above comments.

 

[geesamand]

You need to have the flexibility to calculate life based on each suppliers' C and L10m method. By simply taking the C value and using the classic L10, you are not doing a true comparison between suppliers.

Right. If I compare a few top-tier suppliers bearings using their own full proprietary L10m method, and achieve, say, +/-15% of the life, it could be reasonable to argue these are equivalent bearings. There would be some trust placed in their calculation methodology but with a proper short list I see little risk.

You need to add to the fine print that you calculate bearing fatigue life based on compensation factors for each supplier.
You could note in your fine print of your contract that if the L10m life calculations of two bearings are +/- 15% from one another, the choice of suppliers would be up to you, provided the suppliers are one from the following list: and here you would list all major manufacturers you prefer to deal with.

This is an interesting approach. I have 3 suppliers in mind for my list. I reviewed a few typical specifications: a little more than half of them required ABMA L10 at full motor nameplate and design operating speed. This cuts out the use of compensation factors, as ABMA does not yet have an L10m method published. Therefore my fine print would have to declare the use of compensation factors employed (material) at the outset. I'm sure that Sales would rather go to market without any fine print, but if I can cut costs without losing quality, I suspect they would accept it if the fine print makes sense.

On a deeper thought, there is too much faith put in the L10 or L10m calculations in my opinion. Much longer life improvements can be obtained by using proper shaft and housing fits, housing and frame rigidity, sealing, lubricant selection and re-lubrication quantities and intervals, as opposed to arguing about 95 000 hrs vs 100 000 hrs of "calculated life".
I'd be interested in your thoughts on the above comments.

I agree with your list if you allow me to add accurate understanding of the load spectrum and keeping water out of the oil. I'd rather own a machine with mid-range bearings in a top tier machine design and maintenance than vice-versa. It is rare to find a situation where the customer will notice the difference between a box with 100,000h bearings and 150,000h bearings of the same construction. Either it lasts for a couple of decades or it is not maintained and falls apart.

It's quite clear that the engineering companies who spec these machines don't know/care about real-world limitations of bearing life - they provide a paper illusion of reliability. The contracts are written in a way that encourages me to build a machine without enough metal or rigidity, junk seals, gearing with half-bogus ratings, oversized junk bearings, etc. Fortunately these contracts tend to have a short list of approved suppliers - the challenge is that newcomers with less robust products are constantly lobbying to get on the list.

The catch is that the end user must approve all exceptions to the contract, and there is nothing they enjoy more than to refuse a trivial exception in order to further defer payment and extend their warranty window just a bit longer. 99,000 hours with ample conservatism is fodder for that agenda.

 

[Tmoose]

Hi BearingDave,

"On a deeper thought, there is too much faith put in the L10 or L10m calculations in my opinion. Much longer life improvements can be obtained by using proper shaft and housing fits, housing and frame rigidity, sealing, lubricant selection and re-lubrication quantities and intervals, as opposed to arguing about 95 000 hrs vs 100 000 hrs of "calculated life"."

What is your take on "endurance limit" ratings published by FAG and ( with much more conservative values ) SKF?


They make sense to me, but of course require "utmost" cleanliness (which is not L10's expectation, making me think surface distress is more important than sub-surface initiated fatigue) , details to prevent skidding, etc.
Some big name industry experts claim fatigue limit rating is pretty much nonsense. First from questions about the concept of surviving 10^7 0r 10^8 cycles being the same as infinite life, and then even challenging the test methods and conclusions used by FAG and others.

 

[geesamand]

Tmoose,

I'm interested in that general question also.

I've read some ultra-high-cycle fatigue studies and came away with a lack of consensus. My experience in UHCF design assumes no such thing as an endurance limit. Therefore I remain skeptical that suddenly, a couple of researchers at one company observed runout points on the test grid and therefore declare a fatigue limit for every single bearing they make. Other researchers have failed to duplicate their results. And somehow this very new and controversial model becomes a standard. That smells funny to me.

I have a rainy-day project to play with these so-called fatigue limit numbers and compare them to L10 and L10a values for the same conditions. (Do they correlate to an obscenely bearing life, or are they under 100,000hr, where this could be commercially disruptive?) Are the lubrication conditions repeatable outside of a lab environment? Also, how much of a window exists between the minimum load and this fatigue load - is the range wide enough to be practical? I haven't made time to explore this yet.

If I was buying machines with bearings rated to ISO 281, I would expressly prohibit the use of fatigue limit calculations until more consensus is obtained.

David

 

[BearingDave]

Hi BearingDave,

What is your take on "endurance limit" ratings published by FAG and ( with much more conservative values ) SKF?
Some big name industry experts claim fatigue limit rating is pretty much nonsense.

Tmoose,

I'm interested in that general question also.

Hi Tmoose,

I believe you are talking about these two articles written by Edward V. Zaretsky of STLE.

Find below download links to the article (they are pretty good, worth a read):
Part 1
Part 2

In the articles the author goes on to say (from my understanding) that there is no sound scientific basis to claim that there is a fatigue load limit for rolling element bearings, and that eventually, even if the load is very low, the bearings will fail in subsurface initiated fatigue eventually.

My thoughts on this are:
- L10 and L10m are basic life ratings used to get an approximation of bearing life, not an exact value. We should not treat it as anything other than an estimation.
- The L10m method including the fatigue load limit does have its issues, but overall we can agree that if a bearing has a load applied below the fatigue load limit, it will not last "forever", but still last "a very long time" (if we consider the end of life as being subsurface initiated fatigue). "A very long time" typically surpasses the intended design life of the entire asset.
- If we want the true fatigue life, we need to calculate it with the Finite Element Method using advanced software which takes into consideration the true geometry and materials of the bearing. L10 or L10m is a basic method that gets us in a ballpark, while the FEM gets us in left or right field. Testing the bearings, over time and in service, will get us on the first or third base within the left of right field.