35 replies to this topic

[manolis]

Hello.


Here is an unconventional cylindrical roller bearing:







Sets of auxiliary needle rollers (5), "trapped" among a raceway (3a) and two neighbor cylindrical rollers (4), prevent the cylindrical rollers from coming into contact with each other.


Here is an unconventional ball bearing:



and here is the “full complement version”:



The conventional cage has been substituted by subcages (6) and auxiliary balls (5).


And here is a “more conventional” PatRoller Ball Bearing:



More at http://www.pattakon....onPatRoller.htm


Thoughts?
Objections?

Thanks
Manolis Pattakos

[gruntguru]

Hi Manolis. I assume the intent in each case is reduced friction? If so, have you performed any analysis to verify the difference in friction?

 

For example in the first case (cylindrical roller) although you have eliminated the traditional cage and the associated sliding friction, your design has introduced 10 additional lines of rolling contact per main element. The viscous drag created by these extra contacts must be significant - even assuming zero load. One suggestion would be to modify all these additional rollers to a dogbone shape so that contact occurs only at the ends. Or perhaps modify elements 5a and 5d only - reducing the size of 6 contact areas.

[Greg Locock]

also the fatigue life is controlled by the parameter R/r (from memory), so your little drive rollers will Brinell the main rollers. Now, it should be that the forces involved with maintaining roller spacing are small compared with the radial forces so this doesn't matter (I suspect this is why cage friction is not /that/ significant). Is it feasible to build one of these from stock roller sizes?

Edited by Greg Locock, 21 September 2015 - 07:45.

[manolis]

Hello Gruntguru.

Here is a pair of conventional cylindrical roller bearings:



The drawing is based on data taken from a "FAG standard program catalogue".

With the same external dimensions (100x215x73), the red "full complement" version (which uses 34mm diameter rollers, while the blue "cage" version uses 32mm diameter rollers) has a load carrying capacity increased by 15%, in expense of a substantially lower speed limit.

For the same rpm, the sliding speed at the contact of two neighbor rollers of the full complement version is about twice the sliding speed at the contact between a roller and the cage of the "cage" version.

It is interesting that the speed limit of the full complement version (1,600rpm) is half than the speed limit of the cage version (3,200rpm) in case of oil lubrication.

The maximum speed of a rolling bearing is limited by the resulting temperature increase.

It seems that the temperature increases with the sliding speed: at 3,200rpm the sliding speed (between each roller and the cage) in the "cage" version is the same with the sliding speed (at the contact of neighbor rollers) at 1,600rpm in the full complement version.

It also seems that the rolling speed of the rollers onto the raceways doesn’t affect significantly the temperature (and the limiting speed) of the rolling bearing (at their speed limits, the rolling speed of the full complement version is about half than the rolling speed of the "cage" version).

If the above "theory" is correct (i.e. that the rolling friction is insignificant in comparison to the sliding friction), using sets of needle rollers (as those shown at the top of the drawing) between neighbor rollers of the full complement version, the +15% additional load carrying capacity remains attainable, with a rev limit substantially higher than the 3200rpm of the "cage" version.

However, it is not all about friction.

The PatRoller architecture enables new design possibilities like extended side fillets, more robust external bearing ring, protected internals, easy sealing, compactness, increased load carrying capacity, exploitation of the full length of the rollers / raceways, etc.

The conventional "cage" rolling bearings need space for the cage.
For instance, it is shown below the architecture of the LSL and ZSL low friction cylindrical rolling bearings:



The central "slim" cage (whereon the rollers slide) occupies a significant part of the length of the rollers. Without the cage, the same roller bearing would exploit the complete length of its rollers and its load carrying capacity would be proportionally increased.




Hello Greg Locock.

Yes it is feasible, and easy, to built a PatRoller "prototype" from parts of stock roller bearings.

For instance, with each set comprising one 2mm diameter needle roller, three 2.5mm diameter needle rollers and one 8mm diameter needle roller (the brown parts at the top of the drawing) the NJ2320VH turns to a PatRoller-NJ3220VH.
The clearance between neighbor rollers is 0.11mm.
The speed limit will shift from the 1,600rpm of the full complement NJ2320VH to, say, 3,500rpm (i.e. well above the 3,200rpm of the "cage" NJ2320E)

An arrangement / mechanism for the demonstration / measurement of the different friction, under various loads and speeds, of the PatRoller and of a similar conventional rolling bearing, both on the same "free" shaft, would be useful (something like a differential “scale” / dyno).

Thanks
Manolis Pattakos

[bigleagueslider]

Interesting concept. I'll have to look at it some more when I have time. But my first impression is that adding all those small diameter needle rollers just to eliminate the cage and/or allow a full roller complement seems to be more trouble than it's worth. As gruntguru pointed out, with medium to high dN roller bearings viscous losses predominate. The friction losses from a well designed cylindrical roller bearing cage/separator are usually quite small. The separator is really only working on the rollers that are passing throught the sector of the bearing where they are unloaded. The rollers that are passing through the sector of the bearing where they are radially loaded will maintain consistent circumferential spacing on their own. The number of cylindrical rollers supporting the radial load in a bearing will depend on the number and diameter of the rollers, the race diameters, and the bearing internal clearance.

 

I think you might see problems with loads on the small rollers under operating conditions, where there is misalignment/displacement between the inner/outer races, dimensional changes due to thermal and dynamic load effects,  and how the positions of each roller in the system changes relative to the adjacent ones over the course of a single rotation. Also, as GregLocock noted, most rolling element bearings are sized based on fatigue life. With fixed outer race, a fixed radial load direction WRT the outer race, and a rotating inner race, one sector of the outer race is subjected to lots of fatigue cycles. And while the rotating inner race has even distribution of fatigue cycles, the contact stress levels are higher due to the much less favorable contact geometry. With your concept, the small diameter rollers will experience many times the number of fatigue cycles the primary rollers do. So fatigue life of these rollers is something you might want to look at further.

 

Good luck.

[Kelpiecross]

My initial impression is that some of the ramp angles etc.between the various diameter bearings are similar to those found in a "roller clutch"   and may be very inclined to jam - especially if the bearing is possibly overheating or running a bit short on lubricant. 

[manolis]

Hello Gruntguru

The dogbone idea is clever.
But the small diameter of the auxiliary needle rollers makes the dogbone difficult for normal size rolling bearings.
And the required manufacturing or modification is expensive.

It is also a matter of confidence: a potential customer would not want to risk with parts not made by well-known manufacturers.

Using only parts made by big manufacturers, some of the needle rollers can be made of two (or three) pieces in-line (common rotation axis).
The gap between the pieces prevents the “oilplaning” (question: would it be bad, or good, the oilplaning of the auxiliary needle rollers?).
This approach is cheap and directly applicable.
Besides, a potential customer would be way more confident knowing that all parts are made by, say, SKF or INA.



Here is another way to modify the conventional NJ2023E (single row cylindrical roller bearing with cage) to a PatRoller NJ2320E.



The cage is thrown away.
Two more cylindrical rollers are added (15:13=1.15, i.e. the load carrying capacity increases by 15%).
Each set of auxiliary needle rollers comprises one needle roller of 2.5mm diameter, three needle rollers of 3.5mm diameter each, and one needle roller of 6mm diameter (i.e. all auxiliary needle rollers are available in the market).

The resulting PatRoller NJ2320E combines the load carrying capacity of the full complement NJ2320VH (which has a speed limit of only 1,600rpm) with a speed limit higher than the 3,200rpm of the “cage” NJ2320E.



Hello BigLeagueSlider

The peripheral speed of all rollers (working rollers and auxiliary needle rollers) is the same.
The load taken by the auxiliary needle rollers is weak (it is similar with the load between the conventional cage and the cylindrical rollers).
Do these fit with fatigue and reliability issues?



When the rolling bearing undergoes, as a whole, significant linear accelerations (say as happens in the roller bearings at the big and small ends of a connecting rod of a motorcycle), the support of the heavy cage is a big issue; after a rev limit, the cage abuts and slides on the bearing rings and soon falls apart.
Think what happens in the case of the PatRoller (cylindrical roller bearing or ball bearings).



What about the simple PatRoller ball bearing :



The ball bearing at right has some 10% larger balls. How? It simply exploits the space occupied by the cage of the conventional ball bearing at left.

Thanks
Manolis Pattakos

Edited by manolis, 22 September 2015 - 06:43.

[rdyn]

There is not that much advantage for full complement cylindrical roller bearings.

But: if you make the roller diameter smaller the advantage increases. Full complement needle bearings are significantly stronger than caged needle bearings. Small auxiliary rollers would rotate at completely insane rpm - does not seem feasible.

 

Sometimes large g-forces are also a problem (planet gears, rotating planetary carrier)

Edited by rdyn, 22 September 2015 - 19:50.

[John Brundage]

What is the cost of manufacturing this bearing vs. a standard of the same size (using the bearing in your drawing as an example)?

[bigleagueslider]

manolis- Increasing the number of rollers by 500% presents a serious cost concern. These rollers are components that require expensive materials, careful heat treatment, and very precise machining, while the typical injection molded plastic retainer does not.

 

If you consider that most rolling element bearings are sized by fatigue life (ie. L10 or similar), then it becomes apparent that there are more effective and less costly ways to achieve a 15% increase in load capability. If you look at the factors used to calculate adjusted L10 bearing life, material quality has the largest impact. Simply using a slightly better quality material for the rollers and races without changing anything else would provide more improvement than any other single modification, and at lower cost.

 

One other issue I noted with your roller bearing arrangement is the apparent difficulty it presents for assembly. The model in your OP shows an outer race with extended flanges, which would require all of the rollers to be inserted radially in a specific sequence, and without any practical way to hold the small rollers in position during the assembly process.

[manolis]

Hello Kelpiecross.

You write:
“My initial impression is that some of the ramp angles etc.between the various diameter bearings are similar to those found in a "roller clutch" and may be very inclined to jam - especially if the bearing is possibly overheating or running a bit short on lubricant.”

Here is a different PatRoller cylindrical:





and here is the version without auxiliary needle rollers:






Quote from http://www.pattakon....onPatRoller.htm :

“The spacer 8 may seem like a wedge between two neighbor rollers; but the two neighbor rollers 4 rotate at the same direction (they both roll on the same raceway), which means the linear speeds of the two rollers 4 at their contact with the spacer 8 (at the two sides of the spacer) are opposite, which means that the spacer is receiving not a combined force that would push it along the radial direction, but a torque (pair of forces) that tends to rotate the spacer about its center.”

Thanks
Manolis Pattakos

[manolis]

Hello Rdyn.

You write:
“Full complement needle bearings are significantly stronger than caged needle bearings.”

Yes, but their speed (rpm) is substantially lower and their friction higher.
In the small end of the connecting rods of motorcycles they are used “cage” needle roller bearings (and a typical failure comes from the touching / sliding of the case on the piston pin).


You also write:
“Small auxiliary rollers would rotate at completely insane rpm”

No.
With equal peripheral speeds for the small diameter needle rollers and for the big diameter cylindrical rollers none rotates at “insane” rpm for its size.

The speed limit is at, more or less, 10m/sec peripheral speed, either you talk for big cylindrical rollers (say 30mm diameter) or for small diameter needle rollers (say 1.5mm diameter).

At 10m/sec the rpm of a 30mm diameter roller is 20 times smaller than the rpm of a 1.5mm diameter needle roller.
For instance, if the one rotates reliably at 5,000 rpm, the other rotates reliably at 100,000 rpm.

The 60,000rpm speed limit of the HK0306TN (6.5mm external diameter, 3mm shaft diameter, 6mm width) is no more “insane” than the 3,200rpm of the NJ2320E (215mm external diameter, 100mm inner diameter, 32mm cylindrical rollers diameter).

I.e. more correct than “insane rpm” is the “insane peripheral speed”.

Thanks
Manolis Pattakos

[manolis]

Hello John Brundage.

Regarding the cost of the PatRoller versus the standard rolling bearing:

Take the simplest – and most common – roller bearing (at left):



and think what is required to modify it to PatRoller (at right).

Seven spacers (more correctly: as many spacers as the working balls) replace the conventional cage (which comprises two “wave form” steel strips and seven pins / nails connecting them).

The first “proof of concept”, zero cost, prototype has already been made:



The original ball bearing is a double row SKF 4307 (80mm external diameter, 35mm inner diameter, 31mm wide, 13.5mm diameter balls, 51kN dynamic load, 32.5kN static load, 8,000rpm rev limit when oil lubrication).

A biro cut in pieces was used for the spacers.
The roller bearing runs nicely / quietly / smoothly.
It can carry as much load (radial and axial) as the original ball bearing.
It can run at high speeds.
The only limitation: if its temperature increases, the biro-spacers will melt / collapse.

From the sides you can see / inspect the balls (previously hidden by the cage).


The required spacers are lightweight, simple in form and cheap to make.

The assembly needs not special machinery (as the conventional ball bearings).

The cost of the PatRoller ball bearing is lower than the original ball bearing.


On the other hand,
if the space now occupied by the cage of the conventional ball bearing is exploited for bigger balls, in the same envelope it will fit a PatRoller ball bearing having increased load carrying capacity and higher speed limit.
And I can’t see a reason for not having lower cost than the conventional.
Can you?

Thanks
Manolis Pattakos

Edited by manolis, 24 September 2015 - 05:57.

[manolis]

Hello BigLeagueSlider

You write:
“These rollers are components that require expensive materials, careful heat treatment, and very precise machining, while the typical injection molded plastic retainer does not.”

These parts are available in the market at high quality (SKF made, INA-Schaeffler made, NSK made, NTN, made, Timken made etc) and low price.

For instance you can buy 2.25Kg / 5lb (it is 1,000 pieces) of 3.5mm diameter needle rollers of 29.8mm length each.



You also write:
“Simply using a slightly better quality material for the rollers and races without changing anything else would provide more improvement than any other single modification, and at lower cost.”

And why not to apply the PatRoller modification to the impoved rolling bearing made of the better quality material? A 15% increase of the load carrying capacity due to the high quality material, plus another 15% increase of the load carrying capacity due to the PatRoller design makes a total of 32% increased load carrying capacity.


In this pdf :

http://www.schaeffle...i_193_de_en.pdf

INA / Schaeffler presents its ball – roller bearing; the sides of each ball are sliced



to allow more balls, in expense of lower axial load carrying capacity and of lower speed.

How do you like INA’s ball-roller solution?

Thanks
Manolis Pattakos

Edited by manolis, 24 September 2015 - 07:38.

[John Brundage]

Hello John Brundage.

Regarding the cost of the PatRoller versus the standard rolling bearing:

Take the simplest – and most common – roller bearing (at left):




On the other hand,
if the space now occupied by the cage of the conventional ball bearing is exploited for bigger balls, in the same envelope it will fit a PatRoller ball bearing having increased load carrying capacity and higher speed limit.
And I can’t see a reason for not having lower cost than the conventional.
Can you?

Thanks
Manolis Pattakos

 

Manolis,

If you remove a cage and add rollers as is said below, the cost has to be higher. Due to the increased amount of parts, the cost of materials and labour time for manufacturing must be greater.

BLS in essence asked the same question,  "manolis- Increasing the number of rollers by 500% presents a serious cost concern. These rollers are components that require expensive materials, careful heat treatment, and very precise machining, while the typical injection molded plastic retainer does not"

 

John

Here is another way to modify the conventional NJ2023E (single row cylindrical roller bearing with cage) to a PatRoller NJ2320E.



The cage is thrown away.
Two more cylindrical rollers are added (15:13=1.15, i.e. the load carrying capacity increases by 15%).
Each set of auxiliary needle rollers comprises one needle roller of 2.5mm diameter, three needle rollers of 3.5mm diameter each, and one needle roller of 6mm diameter (i.e. all auxiliary needle rollers are available in the market).

 

Edited by John Brundage, 24 September 2015 - 22:11.

[manolis]

Hello John Brundage.

Here is the drawing of an NJ2310E cylindrical roller bearing modified to PatRoller NJ2310E:



It uses 1.5mm diameter needle rollers and thin spacers (note: the diameter of each cylindrical roller is 16mm, i.e. it is more than 100 times heavier than a needle roller, and way more if the needle roller is short as in the animation).


Here is the drawing of an NJ2320E cylindrical roller bearing modified to PatRoller NJ2320E:



It uses 2mm diameter needle rollers and thin spacers (note: the diameter of each cylindrical roller is 32mm, i.e. it is more than 250 rimes heavier than a dull length needle roller).

In both cases the set of the working cylindrical rollers is about a hundred times heavier than the set of the added auxiliary needle rollers and spacers.


I hear you saying: “Yes, but what about the cost of these tiny auxiliary parts?”


The Kx20x24x10 is a “needle roller and cage assembly” like:



and comprises the auxiliary needle rollers required for the modification of the NJ2320E to PatRoller.

The price of the Koyo K20x24x10 :



is several hundred times lower than the price of the SKF NJ2320E:




Besides, the cost (material cost and manufacturing cost) for the single piece cage of the NJ2320E is way higher than the cost of the auxiliary parts that replace the single piece conventional cage (worse even, if the cage is made of bronze).


It is not the number of pieces that counts.
An elephant with 99 flies on its skin is still one elephant.

Thanks
Manolis Pattakos

[bigleagueslider]

manolis- To give you a clear illustration of what I noted about the huge cost/benefit provided by the latest developments in coatings and materials for rolling element bearings, take a look at this article from SKF. It shows an improvement in service life of at least 2 orders just by applying an amorphous carbon coating to the rollers of a common steel roller bearing. Applying an amorphous carbon coating to the existing rollers would be much less costly than increasing the number of rollers by 2-5 times. You need to evaluate your concept in terms of cost/benefit vs fatigue life.

[manolis]

Hello BigLeagueSlider.

Thanks for the SKF – DLC link.


As already explained, the number of parts is not what defines the cost.

The cost and size of the auxiliary parts are in another “order of magnitude” than the parts of the original rolling bearing (just like the elephant is in another “order of magnitude” than the flies at its back).

The cost saving from the elimination of the conventional cage may cover the cost for the two additional cylindrical rollers (32mm diameter each) that increase substantially the load carrying capacity of the PatRoller NJ2320E.



As for the DLC technology in the rolling bearings, I can’t see the problem.

Suppose that tomorrow SKF starts selling NJ2320E cylindrical bearings (215mmx100mmx73mm) with DLC coating.

These DLC rolling bearings will also be characterized by a speed limit, by a load carrying capacity limit and by an expected life limit (these limits would be superior than those of the conventional non-DLC rolling bearing, however they are still limits).

If you modify such a high-tech DLC NJ2320E cylindrical roller bearing to PatRoller, the relative improvement will be the same (for instance, its load carrying capacity will still increase by 15%: the elimination of the cage enables the use of 15 cylindrical rollers instead of the 13 of the original roller bearing; and it doesn’t matter whether these cylindrical rollers are conventional or DLC coated: the relative increase will be the same 15:13=1.15 i.e. 15%).

I hope it is now understood that the DLC technology (or the use of improved materials as mentioned in another post) is complementary to the PatRoller architecture. They can both combine to give top characteristics.



In this discussion they were repeatedly mentioned the advantages of using the minimum possible number of parts.

They were also mentioned the advantages of the DLC (diamond-like carbon coating) technology.

Think how many parts the PatRoVa rotary valve comprises and compare their number with that of a conventional valve train (poppet valves).

Think also the advantages of the DLC technology in the PatRoVa rotary valve (DLC on the two “front” surfaces of the rotary valve and on the two external “lips” of the combustion chamber).



(more at http://www.pattakon....akonPatRoVa.htm )

Then think of the reduction of the maximum local temperatures into the combustion chamber of the PatRoVa relative to the conventional.

Without hot spots, like the red-hot exhaust poppet valves of the conventional, the formation of NOx gets more difficult to happen (isn’t the recent "VW scandal" about NOx ?).

Think also of the new design possibilities the shape of the combustion chamber of the PatRoVa rotary valve can offer to the direct injection spark ignition and Diesel engines.

Thanks
Manolis Pattakos

Edited by manolis, 26 September 2015 - 07:23.

[bigleagueslider]

Interesting discussion about your bearing concept Manolis.

 

During your research of IP prior art I imagine you saw this patent from 1979.

 

I'd like to discuss the issue of cost/benefit a bit more. Let's define exactly which situations your bearing concept might be the best option. First, as noted previously it would only be suitable for operation at modest dN conditions, similar to any other full complement cylindrical roller bearing. Second, if there is no limit on bearing OD then the slight advantage in radial load capacity provided by a full complement versus a caged bearing of similar OD goes away, since you can simply use a larger bearing. Third, if running friction/efficiency is an issue, then at modest dN a slightly larger caged bearing will be more efficient than a slightly smaller full complement bearing of similar dynamic radial capacity. So the only situation I can see where your bearing concept would be advantageous is where a slight increase in radial load capacity is required, there can be no change in cross section dimensions of the bearing (ID, OD, width), cost is not a primary concern, operating dN is modest, and efficiency is not a big concern.

 

Here is a good technical reference for running torque of cylindrical roller bearings.

[manolis]

Hello BigLeagueSlider

Thanks for your time searching the prior art of cylindrical PatRoller.
It helps.
I wish other forum members to do the same for the ball PatRoller.


The US4,174,141, and its two prior art references (US1,796,813 and US2,120,533) deal with the same problem.
The solutions they propose require additional raceways (or rings) at the sides of the working cylindrical rollers, they also require auxiliary (needle) rollers of the “dogbone” type (or of “non-constant diameter” type, in general).
Common characteristic of the above three patents: a significant part of the width of the rolling bearing is utilized for the auxiliary rollers and not for the working rollers that receive and carry the load (i.e. they necessarily have lower load carrying capacity).

In comparison, in the cylindrical PatRoller there is nothing extending outside the working cylindrical rollers. If necessary, the width of the rolling bearing can be equal to the width of the cylindrical rollers.


Regarding the friction in a rolling bearing, the estimation according SKF is like:



According this, in the full complement cylindrical roller bearings the contact between the neighbor cylindrical rollers doubles the friction (the constant coefficient of friction – table 1 – goes from 0.0011 to 0.002).
The relative speed at the contact between neighbor cylindrical rollers (case of full complement) is two times the relative speed between the cage and the rollers in case of cage-roller-bearing.

According the same table, the friction is proportional to the mean diameter of the rolling bearing. The smaller the diameter of the rolling bearing, the less the friction.

For instance, with 1 ton (10,000Nt) radial load on a cylindrical roller of 100mm mean diameter (it is the diameter of the circle wherein the centers of the roller bearings lie), the resulting frictional moment is 0.55Nm for the cage version and 1.0Nm for the full complement cylindrical roller bearing.
If the diameter of the "cage" roller bearing doubles, the frictional moment doubles (it goes to 1.1Nm).

So, a smaller diameter full complement is preferable, provided the cylindrical rollers are not allowed to contact each other.
This is what the PatRoller does. It allows “full complement” with less friction than the “cage” version.
If you look deeper, the spacers with the needle rollers:



receive the big part of the force between the neighbor cylindrical rollers (rolling friction), decreasing the sliding friction.

Worth mentioning: even if the “envelope” allows a bigger diameter rolling bearing to be used, besides its increased friction it is also its higher cost that must be taken under account.


From the commercial point of view, the common deep groove ball bearings are more interesting, because they comprise the big majority of the rolling bearings in service.

The spacers of a PatRoller ball bearing:





just bridge neighbor balls.
The overall length of the spacers is about half of the length (or periphery) of the conventional case.


From the technical point of view, the PatRoVa rotary valve with DLC coating (previous post) can bring a bigger change to the world.

Thanks
Manolis Pattakos

Edited by manolis, 29 September 2015 - 00:49.

[bigleagueslider]

You mention "friction" of cylindrical roller bearings. There are several sources of efficiency loss in rolling element bearings, viscous, sliding/skidding, rolling, hysteresis, etc. At medium to high dN, viscous losses are far and away the largest source. At low dN, sliding contact at the roller end faces and race shoulders, at the cage/roller contact, and contact between the roller/race surface due to skewing. The information from SKF you posted above regarding the use of a simple Mu value to estimate losses is not a good approach for most situations, since losses are significantly affected by velocity of the bearing components.

 

Also, one issue I saw with the Pat Roller concept posted above is that it does not appear that it can be assembled, at least as shown.

[gruntguru]

The Pat Roller appears to be a roller bearing with removable inner ring. Assembly should be easy for bearings with removable inner or outer ring. (The removable ring has only one axial shoulder.)

[manolis]

Hello BigLeagueSlider

In the SKF general catalogue, after the page “Friction / Estimating the frictional moment” (shown in my last post), there are 15 pages dedicated to the detailed frictional moment analysis.

But the general approach of SKF (a big and reputable rolling bearing manufacturer) is:
“The frictional moment is proportional to the working diameter of the rolling bearing and depends heavily on its architecture (constant coefficient of friction).”

It is a “rule of thumb”, not a law.
There are conditions wherein the rolling friction prevails, others wherein the sliding friction is the big part, other wherein the viscous friction gets more important, and so on.
However the above rule of thumb is still a good approach.

For instance, doesn’t it seem reasonable that the friction is, more or less, proportional to the working diameter of the rolling bearing?
The various “friction” generating processes create friction forces; these friction forces acting on a bigger eccentricity (as happens in a bigger diameter roller bearings) cannot help creating heavier “frictional moment”.

In my last post I forgot to mention that with the PatRoller you can either increase the load carrying capacity of a specific rolling bearing, or you can achieve the same load carrying capacity by modifying a rolling bearing of smaller working diameter (i.e. of smaller friction according the above “rule of thumb”).
With an additional external ring, the smaller diameter PatRoller bearing can be mounted in the original envelope.

You write:
“The information from SKF you posted above regarding the use of a simple Mu value to estimate losses is not a good approach for most situations, since losses are significantly affected by velocity of the bearing components.”

Isn’t, more or less, the velocity of the bearing components proportional to the working diameter of the roller bearing?



You write:
“Also, one issue I saw with the Pat Roller concept posted above is that it does not appear that it can be assembled, at least as shown.”


Regarding the assembly of the “single-piece ring” PatRoller ball bearings (hello Gruntguru):



(sorry but I have to post the drawings again and again), things are quite simple:

The spacers have a small elasticity (this is why they are shown with cuts / slots).
All the balls and spacers, except one spacer, are inserted easily between the two bearing rings (i.e. between the outer and inner bearing rings, each one being a single piece).
To install the last spacer between its two neighbor balls, these balls are pressed, by a tool, away from each other; the already installed spacers are elastically deformed allowing the rest pairs of balls to slightly approach each other increasing substantially the distance of the balls at the ends of the last spacer.
After the insertion of the last spacer the balls are freed and the ball bearing is ready to work.
If, for some reason, the ball bearing is to be disassembled, all the difficulty is the removal of a spacer. The neighbor balls of the spacer are pressed away from each other, the rest spacers are elastically deformed allowing the rest pairs of balls to slightly approach each other etc.


But I think you mean the assembly of a cylindrical PatRoller bearing like the one in the following stereoscopic animation:



There are various ways to do the assembly.

A first way:
The last spacer is different (is the “master spacer”); it comprises two pieces secured / locked to each other (just like the master link of the chain of a bicycle).

A second way:
The small elasticity of the spacers is exploited.
The assembly is more than easy before the last spacer.
In order to insert / install the last spacer, its neighbor cylindrical rollers are pushed away (along the periphery of the external bearing ring) from each other; each pair of neighbor cylindrical rollers abuts on the needle rollers of the spacer between them, and tries to push away the needle rollers from each other; the needle rollers abutting on the spacer push the spacer to slightly expand. Due to the big number of spacers, the distance of the two cylindrical rollers wherein the last spacer is to be inserted increases substantially allowing the last spacer with its needle rollers to pass through the gap and get installed. Then the cylindrical rollers are released and the cylindrical rolling bearing is ready to work. The spacers are trapped between the cylindrical rollers.
For the disassembly, the reverse procedure can be followed.



The above drawing shows the moment the last spacer (shown by purple color) is inserted.
Compare the drawing with the previous posted (both for the NJ2320E).
Each of the rest spacer has to lengthen by 3.5% between its needle rollers during the assembly of the last spacer.

And here is the PatRoller NJ2310E during assembly:



In order the gap between the neighbor cylindrical rollers of the last spacer to open at 1.5mm (and so to allow the spacer with the 1.5mm diameter needle rollers to pass), each one of the rest spacers must expand a little so that the distance of its two neighbor cylindrical rollers to reduce by 0.05mm (14*0.05=0.7 and 0.7+16.8=17.5=16+1.5 )


A third way:
The external bearing ring is heated at, say, 225 degrees Celsius, the rolling elements are cooled at, say, –75 degrees Celsius (dry ice). The thermal expansion / contraction of the parts makes the gap between the last pair of cylindrical rollers adequate for the last spacer to pass (the calculations are easy; if there is interest, I can explain why 300 deg Celsius temp difference is adequate for the assembly of the PatRoller NJ2320E). As the temperature difference between the parts weakens / fades, the roller bearing cannot disassemble.
To disassemble the rolling nearing, a spacer needs to break / to be sacrificed.

A fourth way:
It combines the second and the third ways. With the external bearing ring hot, the required elastic expansion of the spacers reduces.

A fifth way:
Instead of heating the external ring, a high pressure can be applied on its inner space. You install all components - except the last spacer – at their positions into the external ring (the last spacer is slightly pushed between its neighbor cylindrical rollers but the needle rollers cannot pass through the gap).
Then you put two sealed covers at the sides of the external ring.
Then you fill the space into the external ring with oil and increase the oil pressure.
The external ring expands elastically, the gap at the last spacer increases and the last spacer “slides” between its neighbor rollers.

Thanks
Manolis Pattakos

Edited by manolis, 01 October 2015 - 04:32.

[bigleagueslider]

Manolis- Thanks for the responses.

 

Looking at your cylindrical roller bearing concept, one thing that stands out is the radial height of the outer race end flanges. One function of race end flanges in a cylindrical roller bearing is to keep the rollers aligned relative to the race surfaces. And to perform this function the inner flange faces need to have minimal operating clearance to the roller end faces. There can be a fair bit of sliding contact occuring at the end faces due to roller skewing moments. All cylindrical rollers skew to some degree due to misalignement between the races when loaded. The precise shape (flange layback angle, end crowning, etc) of the roller end face and flange can have a significant effect on friction losses at this contact. The much greater radial height of your bearing's flanges might effect the sliding frictions at the roller end face due to skewing. So this is something you might look into. Here is a good technical reference:

[manolis]

Hello BigLeagueSlider.

The long side ribs (flanges) of the cylindrical PatRoller are not a limitation because it is not obligatory to use the entire “much greater radial height” of the side flanges to support the cylindrical roller ends.

You can have “zones” (“circular track areas” between two radiuses) slightly extending from the rest surface of the side flanges wherein the end surfaces of the cylindrical rollers slide / abut.

If desirable, these zones can be exactly as they are now the side ribs of the conventional cylindrical roller bearing.

Optionally they can be more centrally located to make, with the same clearance, a better alignment (less skew) of the cylindrical rollers (it is easy to think how: the bigger the distance of the two “point” a cylindrical roller is abutting on the side flanges, the better the alignment of the cylindrical roller; and the taller the side flange (but not taller than the radius of the cylindrical roller) the bigger the distance of the two “contact” points).

It is a matter of balance between the sliding friction and the better alignment of the rolling elements.

On the other hand, the “much greater radial height” of the side flanges (even if only a part of them is used for the alignment of the cylindrical rollers) makes the outer bearing ring more robust, reduces the number of parts, protects the inner parts etc.

Thanks
Manolis Pattakos

[bigleagueslider]

I don't see the need for making the outer race "more robust" by using significantly longer flanges. I've never encountered a situation where a conventional roller bearing outer race lacked sufficient "robustness" that could be addressed simply by using longer end flanges. I can't see any practical reason for using them. Is there something I'm missing?

 

There is also a good reason for minimizing the radial height of the inner flange face that contacts the rollers ends. The process used to finish grind these faces requires a carefully shaped wheel that does not lend itself to the taller flanges. It is very easy to produce "grinding burn" (or detempering) on the flange faces due to a large contact area with the grinding wheel. Using the minimum flange height practical reduces the likelyhood for grinding burn.

[manolis]

Hello BigLeagueSlider and thanks for insisting on this thread.

A U-beam shaped external bearing ring saves width and enables a thinner external raceway (i.e. bigger diameter cylindrical rollers).

As you correctly write, a bigger diameter grinding wheel may be required.

But these are details.
The PatRoller architecture fits with both: tall and short side ribs.
A conventional rolling bearing cannot use tall side ribs because they make impossible the insertion of the conventional spacer.

Here:



they are used wire-frame spacers and hollow auxiliary rollers (a stereoscopic GIF animation is at http://www.pattakon....e_frame_STE.gif )

The sliding speed between an auxiliary hollow roller and its shaft-wire-frame is about half than the rolling speed of the same auxiliary roller on the working cylindrical roller wherein it abuts and rolls.


And here:



is a PatRoller ball bearing with inflexible spacers.

One spacer is the master spacer comprising two threaded parts and a clip key (the weight of the master spacer equals with the weight of any other spacer).

During assembly the two parts of the master spacer are bolted completely.

After inserting all balls and spacers between the two bearing rings, the two parts of the master spacer are unbolted as required (to give the desirable clearance between spacers and balls) and finally the clip-key is inserted to forbid the two parts of the master spacer to turn relative to each other.

The assembly / disassembly is more than easy.

The entire ball bearing can undergo severe accelerations without a problem.

The cost is not higher than a conventional ball bearing of similar size.

Thanks
Manolis Pattakos

[bigleagueslider]

The full complement cylindrical roller bearing and the deep groove conrad-type radial ball bearing examples you show brings up a good point regarding load capacity. One way to improve radial load capacity of either bearing type is to produce radial preload in the bearing so that there is better load distribution between the rolling elements. This is easy to achieve in a ball bearing simply by applying axial preload or by using a split race. However, achieving radial preload in a cylindrical roller bearing is not so easy.

[manolis]

Hello BigLeagueSlider

Take a conventional or PatRoller cylindrical roller bearing having D1 outer raceway diameter and d cylindrical roller diameter.

The diameter D2 of the inner raceway is slightly (say by 0.02mm) smaller than D1-2*d and the bearing is not-preloaded (i.e. D2=D1-2*d-0.02mm).

Suppose you replace the inner bearing ring with one having a raceway of diameter:
D2=D1-2*d+0.02mm.

The bearing becomes preloaded. If the 0.02 increases to 0.03, the preloading is heavier.

For its assembly all you have to do before inserting the inner bearing ring is to heat the assembly of the external bearing ring with the cylindrical rollers in it (or to cool the inner ring).
How much?
Quote from http://www.pattakon....onIdleValve.htm :
“An idea of what the 0.02mm is: by changing the temperature of the 102mm long intake valve for 17 degrees centigrade, its length changes by 0.02 mm”)

The same are applicable in the case of a conrad-type ball bearing.

A side effect of preloading is the increase of the friction and the decrease of the maximum speed.
In the catalogue of the crossed-roller bearings (for high precision fitting) of Schaeffler, I see 8m/sec maximum peripheral speed without preloading and only 4m/sec in case of preloading (in both cases: oil lubrication).

Thanks
Manolis Pattakos

Edited by manolis, 08 October 2015 - 03:46.

[bigleagueslider]

"A side effect of preloading is the increase of the friction and the decrease of the maximum speed."

 

This is incorrect. Most rolling element bearings require a certain minimum level of preload that is a function of operating dN. The reason rolling element bearings require some amount of preload is to prevent skidding. A pair of angular contact ball bearings without preload operating at high speed and little load would quickly fail due to skidding. The carefully controlled preload adds a bit of prevailing friction, but the bearings cannot function properly without it.

 

There is a nice engineering tool on the SKF website that calculates the minimum amount of preload required for a specific bearing set and operating speed. Higher operating speeds typically require greater preload.

[manolis]

Hello BigLeagueSlider.

You write:
“
"A side effect of preloading is the increase of the friction and the decrease of the maximum speed."
This is incorrect. Most rolling element bearings require a certain minimum level of preload that is a function of operating dN. The reason rolling element bearings require some amount of preload is to prevent skidding. A pair of angular contact ball bearings without preload operating at high speed and little load would quickly fail due to skidding. The carefully controlled preload adds a bit of prevailing friction, but the bearings cannot function properly without it.”
”

The preloading cannot help increasing the friction and decreasing the maximum speed.
However, and no matter how bizarre it seems, the same preloading extends the expected life of the bearing and improves its reliability.

According their manufacturers, for high speed roller bearings increased clearance and higher accuracy are required.
The heat created by the friction inside the roller bearing is what limits its maximum speed.
For instance, the maximum speed of the preloaded “cross roller” bearings of SKF is half than without preloading (which means that at the speed limit of the preloaded “cross roller” bearing, the same, but non-preloaded, “cross roller” bearing has half friction under the same load.

The manufacturers of roller bearings propose a minimum radial load:
“In order to ensure operation without slippage, the bearing must be subjected to a minimum load (Fr min) in a radial direction.
This applies particularly in the case of high speeds and high accelerations.
In continuous operation, a minimum radial load of the order of P/Cr>0.02 is therefore necessary.”

With the preloading and the minimum radial load they are avoided critical (as regards the reliability) conditions of operation.


In a similar way, avoiding the “zero speed” of the piston rings at the Combustion Dead Center, the sleeve valve engines achieve a better reliability in expense of more friction and higher specific consumption.
More at the http://forums.autosp...pposed-pistons/ thread.

Thanks
Manolis Pattakos

[bigleagueslider]

Manolis-

 

You are confusing things a bit by using the unusual example of a crossed roller bearing with radial loads and high dN. Crossed roller bearings are primarily used at very low dN with combinations of axial and out-of-plane moment forces, such as in a slew ring. These bearings have a substantial amount of sliding contact at the outer roller end faces which results in high levels of friction when axial preload is applied. The situation is much different with a pair of angular contact ball bearings, a 4-point ball bearing, or even a pair of tapered roller bearings. The optimum amount of preload does increase the prevailing rolling friction in the bearing system, but it also reduces the friction losses from skidding and also increases fatigue life by providing a better load distribution between the rolling elements.

 

Very high dN ball and roller bearings that have a fixed outer race, rotating inner race, and a thin wall (hollow) shaft, often require additional internal radial clearance. This is because at high speed the dynamic forces cause the inner race to grow slightly. The rotating inner race also tends to run hotter than the fixed outer race since it has a less efficient pathway for conducting heat away.

[manolis]

Hello BigLeagueSlider

I can’t see where the disagreement is.

The preloading increases from slightly to significantly the friction (at least at some operational conditions); at the same time the preloading makes the roller bearing more reliable by avoiding the skidding and the local wear, and by distributing the load among more rolling elements:



By the way,
is there any difficulty in applying the desired / required preloading to the PatRoller rolling bearings?

Thanks
Manolis Pattakos

[bigleagueslider]

Which specific type of rolling element bearing do you have in mind?

[manolis]

Hello BigLeagueSlider

You write:
“Which specific type of rolling element bearing do you have in mind?”


Case A

Full complement, divided inner ring, ball bearing:




Case B

Cylindrical roller bearing:



Spot on the inner auxiliary needle-rollers (they have bigger diameter than the outer auxiliary needle-rollers); the inner auxiliary needle rollers take the centrifugal loads of the spacers.


Case C

Single piece bearing rings, inflexible spacers:



Thanks
Manolis Pattakos

[bigleagueslider]

Take a look at this ball bearing concept from RBC where they use slightly undersized balls to act as separators in applications where consistent friction and debis generation are critical.

 

http://img.directind...s-145177_1b.jpg

Edited by bigleagueslider, 18 October 2015 - 00:46.