h a l f b a k e r y

meta:

account: browse anonymously, or get an account and write.

 user: pass:
register,

# Tri-Layer Bearings

Use more rollers or balls, instead of cages
 (+5, -1) [vote for, against]

Synopsis:
See 2nd, 3rd, and 4th links.

Details:
Let's start by considering a sledge (a super-heavy-duty sled). It is dragged across a flat surface, and there is often a lot of friction. To ease the effort of dragging, rollers might be employed (as in side view schematic below):
--bottom of sledge--
O O rollers O O
----level ground----

That description is the basis for every kind of roller-bearing or ball-bearing on the market. But all have the same drawback, in that the rollers or balls generally must not be allowed to touch each other, inside the bearing. Consider this side view sketch: C' C' C' C' which is supposed to represent directional arrows of rotation for some adjacent rollers or balls (clockwise). Note carefully while the right side of a given roller may be described as rotating "downward" in that sketch, the left side of the adjacent roller is rotating "upward". Imagine the friction if they touched each other! For this reason, almost all bearings come with an extra part called a "cage", which prevents the rollers or balls from touching each other (only bearings that are used at low speeds don't especially need cages). However, while cages prevent the greater friction, we still have the lesser friction of the rollers or balls rubbing against their cages! Thus do all bearings eventually wear out and must be replaced, even when properly and regularly lubricated.

I'm sure the bearing industry loves the inherent modest lifespan and guaranteed future replacement-sales associated with the traditional designs, and so this Idea is not going to be well-received by them. Nevertheless...consider the following sketch: |C'C,| in which two adjacent rollers contact each other and rotate in opposite directions. They would (in this example) descend in-sync together in-between the two vertical walls. Or they would rotate-in-place while the walls moved up or down in sync. That fact is no good for bearings, since the two "walls", otherwise known as the "inner race" and the "outer race", must be able to move independently, which the synchronized motion of these two rollers cannot accommodate. So consider this sketch: |C'C,C'| with three adjacent contacting rollers. This situation is equivalent to a single roller between the walls, in which the walls are free to move independently, while the rollers rotate in place. And now, this last sketch shows the relevance of that:
--outer race--
C' C' C' C'
_C, C, C,
C' C' C' C'
--inner race--

I cannot show the true spacing of the rollers in that sketch as accurately as I would like. Within each "layer" of rollers (horizontal rows here), NONE touch each other. However, every roller in the central layer must contact two rollers above and two rollers below, just as the upper layer contacts the outer race, and the lower layer contacts the inner race. In this design, every roller is locked into its position, relative to its neighboring rollers (remember that the inner and outer races are circular, so that the three layers of rollers completely surround the inner race, and there are no open ends). The rollers can never rub each other the wrong way, and no cage is needed.

For ball bearings, the situation is somewhat more complicated, but the goal can still be accomplished. Look again at that last sketch, and pretend it is a cutaway view of a portion of a ball bearing (the races curve toward and away from you in the +Z and -Z directions), and not a side view of a portion of a roller bearing. As a cutaway view, there are portrayed four whole parallel rings of balls that comprise the lowest layer that contact the inner race (all curving toward and away from you). No balls in any of these rings contact each other Then there are three whole parallel rings of balls in the middle layer. While again no balls in these rings contact each other, nevertheless each of those balls nestles between and contacts four balls of the lowest layer. Finally, the upper layer also consists of four whole parallel rings of mutually non-contacting balls, and again each ball in the middle layer nestles between and contacts four balls in the upper layer. Now take a moment to recall that the inner and outer races must accommodate their adjacent rings of balls, typically by having curved channel/guideways in which those balls can nestle and roll. One consequence of having these three interlocked layers of nestling balls is that this particular bearing design can do duty as either the normal motor-axle type, or also as an "axial thrust" type, able to support the load of a vertical axle. (Ordinary axial thrust bearings have cone-shaped rollers, and it will be very difficult to make this tri-layer-bearing Idea work with conical rollers -- so it is a Good Thing that tri-layer ball bearings can do double duty.) Anyway, you should be able to see how this design obviates any need for a cage.

A variant tri-layer ball-bearing design can have each ball in the middle layer contacting only three balls in each of the other two layers, instead of the previously-described four. This will require that the middle-layer balls be physically smaller than the balls in the other layers (in order that every group-of-three-balls, in the upper and lowest layers, be associated with a nestler), but that won't affect the functionality of the bearing at all. Indeed, if desired, many variants are possible in which the balls or rollers in each layer are different sizes from the other layers. All will work, provided that no two balls in any one layer can contact each other. And, of course, one could design five-layer bearings, seven-layer bearings, and so on, but that would probably be wastefully redundant, because they offer no advantage over tri-layer designs, and require many more component parts.

Next, because the innards of tri-layer bearings consist entirely of smoothly rolling parts, very little or even NO lubrication will be needed. Their inherent low internal friction will let them function for a long long time. And that lack of lubricant brings us to another important feature that ordinary bearings cannot offer: Electrical conductivity. See, most lubricants are non-conductors of electricity. Should there be any voltage build-up in a rotating part (static electricity does happen!), eventually it will be able to force its way through the bearing to the outside mounting equipment. In doing so a small electric arc will occur, and any welder knows that uncontrolled electric arcs can damage materials, such as the innards of a bearing. This phenomenon is well documented for ordinary bearings, and is just another aspect of the wear-and-tear that they must suffer.

To be fair, I should mention that the lubricant is not always at fault, with respect to arc-damaged bearings. Graphite, for example, is a pretty good lubricant AND a good conductor of electricity. A more subtle culprit is the material from which bearings are made. This is usually a hard steel, and it happens that most metals are almost always and quite Naturally coated with a micro-thin layer of oxide -- a layer that frequently is harder than the metal itself. And that layer is ALSO a nonconductor of electricity! So,when electricity flows between two bare metal objects, a damaging arc effect can still occur, simply because the current has to blast through the oxide layers on those two pieces of metal. What this means is, "Why do I say that tri-layer bearings can offer an electrical conductivity advantage over ordinary bearings, assuming both are made from the same materials?"

Well, there is indeed little advantage if both kinds of bearings are made of exactly the same materials. However, let me digress for a moment, to discuss some alternate materials. Copper, for example. For minor electrostatic arcs, two pieces of copper generally exhibit no damage after the event. This is because copper is NOT normally coated with an oxide layer; the metal reacts poorly with oxygen in the air at ordinary temperatures. So, with no coatings to blast through, no arc-damage is only to be expected. But am I now suggesting that we make tri-layer bearings from copper? Not really, the metal is too soft to handle the mechanical stresses for which bearings are normally used. But it is not so bad an idea, to make and use some copper tri-layer bearings together WITH hard steel tri-layer bearings, so that the steel bearings carry the stress, and the copper bearings carry electricity. Remember, the extremely low internal friction of tri-layer bearings will let such unstressed soft copper roll and roll and roll for a long long time, with minimal wear-and-tear (even more minimized, perhaps, if graphite lubricant is used!). Thus they could carry quite heavy electrical loads between stationary and rotating parts....

Before this digression ends, I note that a few other metals, like gold, silver, and platinum, also seem to normally not be coated with a micro-thin oxide layer. If it wasn't for the expense (and the fact that some of them are even softer than copper), they could used just as easily as copper, for conductive tri-layer bearings. But one metal in this oxide-free category, iridium, is quite interesting in that it is a pretty hard material. Yes, I know it is way too expensive to make whole bearings from it, but what about electroplating? We could take all the parts for a hard steel tri-layer bearing, strip off the oxide coating, and then electroplate a micro-thin iridium coating. When assembled, the bearing should have no problem at all handling static electricity, because there will be no oxide layer preventing current flow until a damaging arc occurs. --And why can't we do this iridium trick with ordinary bearings? Because of the cage! Friction between the cage and the other parts will eventually wear off the iridium layer, allow an oxide layer to form, and then arc-damage will be as possible as on any much cheaper ordinary non-iridium-plated bearing.

In closing I focus on that word "assembled", because assembling a tri-layer bearing poses an interesting challenge. We don't want any chance that rollers or balls can escape from a finished bearing, after all! So how do we get them inside the bearing in the first place? Well, there are ways, some more subtle than others. For example, the least subtle involves a giant press that FORMS the last side of the outer bearing race, after all the other parts are in place. Simpler and less expensive ways exist, too, but I shall refrain from discussing them openly. I happen to think tri-layer bearings should earn me some money, even if the existing industry likes the current wasteful situation.

 — Vernon, May 13 2004

Electrically damaged bearings http://www.vibanaly...ctric/electric.html

Tri-Layer Roller Bearing http://www.nemitz.net/vernon/rollerb.gif
This is a side view schematic.. No means of keeping the rollers from gradually coming out the sides of the bearing, while rolling, is shown. [Vernon, Oct 21 2004]

Tri-Layer Ball Bearing http://www.nemitz.net/vernon/cutaway.gif
Cutaway view. For each group of five balls in the X pattern, I didn't take the time to erase those four little arcs, inside the center ball. You can pretend that the 4 balls are either behind or in front of the center ball (you can also pretend the balls are transparent :). [Vernon, Oct 21 2004]

Three Rows of Balls in Two Layers (twice) http://www.nemitz.net/vernon/ballstak.gif
This sketch attempts to show how the middle layer of balls (one row) can contact either 4 or 3 balls in an adjacent layer (two rows). You can easily see why the middle layer must have smaller balls, when each contacts 3 balls in an adjacent layer. Again, you can pretend that the middle layer is either above or below the other layer. [Vernon, Oct 21 2004]

If you're not logged in, you can see what this page looks like, but you will not be able to add anything.

Annotation:

a picture paints a thousand words.
 — neilp, May 13 2004

So you replace the friction of the cage with the friction of rubbing against the side wall of the part holding them all together? And this part now has to take any load placed on the axle that would otherwise force the middle set of balls to squirt out sideways? I don't see the advantage, and furthermore, I don't even think this will work at all. In any case, you can't hide behind the "I might patent this so I can't discuss the particulars" excuse because that's not nice, for one, and for another, I'd happily sign a non-disclosure agreement, and discuss this off line.
 — oxen crossing, May 13 2004

 This doesn't work, because of the misalignment that will be caused by operating pressures. The bearings, roller or otherwise, will be forced between their neighbours and jam.

 Existing bearings need to be captive, to prevent the problems [o c] has pointed out.

For a practical example: You can walk across a swimming pool full of marbles much more easily than you can on a single layer of marbles.
 — Zanzibar, May 13 2004

Bearings like this have been built (and patented).
 — ldischler, May 13 2004

 [oxen crossing], only roller bearings would have the side-wall rubbing problem, but they do have that already in existing roller bearings. I have wondered about ways to deal with that, and have thought that perhaps something like this roller shape (____) (imagine that the upper horizontal line is also present) could allow the side walls to be curved just like a ball-bearing race. That would allow rolling, low friction, AND retention of the rollers. Only the center layer of a tri-layer roller bearing poses a problem with respect to side walls, and I have a notion or two regarding that, too.

 Next, I do not understand how you think, in a tri-layer ball bearing, that the middle layer of balls is going to squirt out sideways. Each is too closely surrounded by six or eight balls in the other two layers, for this to happen (I did indicate that I was unable to portray that).

 [Zanzibar], such jamming as you mention can only happen if there is any freedom of motion OTHER than rolling, for the balls or rollers. The described design offers no such freedom. Also, recall that I wrote something about different-sized balls or rollers in each layer. It would be sensible if, with respect the the middle layer, the inner layer had smaller-diameter parts and the outer layer had larger-diameter parts. Since each ring has the same number of balls or rollers, and since larger-diameter rings also have greater circumferences, using such graded sizes would consume much of the possible free space into which they might jam-up. Regarding the pool of marbles, the side walls of the pool affect their ability to rotate. This would also be true if a single layer of marbles on the floor of the pool included no free space into which they could roll. (Not to mention that the annotation that follows yours essentially claims that the idea DOES work.)

[ldischler], I did do some looking for Prior Art, when I first came up with this idea. I didn't find anything. Can you provide more information? Thanks!
 — Vernon, May 13 2004

I recall seeing the idea in a search years ago. The patent (or maybe there was more than one) was an old one, not searchable by keywords. It might have been used as a toothless planetary gearbox instead of a bearing, but the geometry would be the same. Another method of getting rid of the cage is to use gear teeth on the ends of rollers. These would mesh with gear teeth in the race.
 — ldischler, May 13 2004

 [ldischler], if the idea was patented all those years ago (and I really would not be surprised, because this Idea seems simple enough for that), then by now the patent should have expired, and tri-layer bearings should be readily available (and I should have found them when I looked for Prior Art). Unless there really is some kind of industry conspiracy to suppress extreme-lifespan bearings. I'd really like some concrete evidence that this is Baked, before I delete it. And, if there is a conspiracy, shouldn't this Idea stay up, just to help foil it? If enough engineers knew that siezed, blown, or otherwise self-destructed bearings could become a rare thing, that would be worth it!

Oh, and I agree about the gear thing. But gears involve some sliding contact during their operation, which means more wear-and-tear than pure rolling.
 — Vernon, May 13 2004

vernon no offence but if a picture paints a thousand words as nelip has quoted please draw more diagrams. like 1 or 2 for this.
 — engineer1, May 13 2004

Most bearings are destroyed by improper lubrication, not by cage friction. And BTW, this idea can't be used with ball bearings--at least with any kind real-world load--because it would result in point contact. Also, this wouldn't work with heavy-duty bearings--such as spherical roller and taper roller bearings.
 — ldischler, May 13 2004

 A picture paints a thousand words, but a thousand words does not a picture make.

 Vernon, please try to make shorter posts, or at least change your format a bit. If there's a concept that can be introduced in one sentence but requires a paragraph to explain fully, use the one sentence and refer to a descriptive paragraph later. Try using a synopsis at the beginning (you've done it before and it worked, why have you strayed again?).

 These tri-layer bearings will not work. Picture a vertical loading pushing down on a horizontal axle. The rollers on the bottom of the bearing will see a force pushing them apart, and the axle will tend to move downward. This downward motion will allow the rollers at the top to move closer together. To prevent this, you'd need a cage to support the loads, again introducing friction.

Sorry, I have to give this a (-).
 — Freefall, May 13 2004

OK, I admit I skimmed a bit here and there (not as much as usual), and on closer inspection, I see how the middle balls might not squirt out. They do transmit, however, that desire to the inner and outer layers of balls which then feel the need to rub really too hard on the outside of their keeper channels, a force not existent in single layer balls.
 — oxen crossing, May 13 2004

 [engineer1], I do have some more ordinary drawings, but haven't had a chance to scan them into a small-size Web-image format, for linking. [ldischler], I am beginning to think that you also are not really visualizing what I am visualizing here. It CAN work for balls!

 In the meantime, let me try another ASCII sketch: O O O O O O O O O O O O O O O O Above are two rows of eight balls each, none touching. Imagine a Really Big Bearing, disassembled so you can poke your head into its middle and look down upon the balls in the outer race. So, these rows are occupying mere portions of two side-by-side (parallel) circular channel-tracks of that outer race, in which the balls would roll. (Also let's ignore gravity, so the balls will stay put and not roll downhill toward the middle of those portrayed rows, as this description continues.) Now pick any adjacent pair of balls in the first row, and the adjacent pair in the second row. These form the corners of a square, and in the center of that square, resting on those four balls, imagine a fifth equal-sized ball. That is just one ball of the middle layer; the sketch offers locations for seven such balls, all in a row, none in that row touching any of its row-mates.

 Now mentally turn that image inside-out, so that the two rows of portrayed balls are ABOVE the single row of imagined balls. If in the original mental image the two portrayed rows were the outer-race layer, and if the single row of imagined balls were the middle layer, then now I am asking you to consider the two portrayed rows as the inner-race layer. Each ball in the middle layer IS locked into place by eight surrounding balls! None can pop out unless some of those eight balls are free to move in some fashion other than synchronized rotation. Such free motion is prevented by the fact that in a bearing, balls are regularly spaced in a complete circle. And here we have three layers of regularly spaced balls in five rings (in an X-patterned cross-section), and NO free space left over. Yes, I know that this means the balls have to be just the right size, relative to the space between the inner and outer races. Years ago I worked out a formula for that; if I can find it (and convince myself that it isn't wildly erroneous) I will post it.

------
I see I was writing that while [Freefall] and [oxen crossing] were doing their own posts. A synopsis would in this case constitute a picture, which is not yet available. When I have such a picture, I can edit the main text to reference it as a synopsis. Regarding the tendency for the inner axle to push all the rollers or balls about, I do understand the notion, but am not convinced that the opportunity for it to happen is really there. I'm thinking that the parts are too-well enmeshed to have such freedom to dislocate. (But if they can do any sort of non-rotational sliding, then your scenario does seem likely.) Obviously the way to find out is to build one (yet since [ldischler] indicated that that has been done...?) Regarding inner-race and outer-race balls rolling along not-exactly-in-the-center of their tracks, yes, I am aware that this would happen, and all this means is that NARROW tri-layer bearings are not likely to be built. Unless really small and very tough balls are used. Which may be inherently superior, because many small balls offer more points of contact for force to transmit through the bearing. (Yet such aren't made today, in ordinary bearings, because all those small balls have to be caged, and "more points of contact" also means that such a cage would wear out faster.)
 — Vernon, May 13 2004

I agree, they have nowhere to go parallel to the rotation, but the outside balls will still put extreme pressure on the inside lip of the race perpendicular to the rotational motion, once loaded. This will cause serious drag, as the natural axis of rolling for a ball in a corner is 45 degrees off one rolling on the floor right next to it. You have sliding, just like a cage, only worse, because it's taking up, what, half? the load with the edge of the race, as opposed to the entire load going into the perimeter of the race (analogous to the tread of a tire).
 — oxen crossing, May 13 2004

 [oxen crossing], that is indeed a most excellent problem with this idea. Thanks! (Now to decide if this should persist; there are plenty other wildly wrong ideas here that have yet to die...and this one was rather subtly flawed, indeed.)

[half], I was invoking the notion that something patented is supposed to be something that works, however poorly. Yes, I know that there are a lot of patented things that cannot work at all, but they are relatively few compared to the total.
 — Vernon, May 13 2004

 Keeping this Idea alive for a while does leave open the opportunity for additional insights. Please consider the following sketch: / \ -O \ / This attempts to portray a middle-layer ball and the axes of rotation of the balls that surround it (but shallower slashes would be better). It seems to me that since that middle ball only makes POINT contact with its surrounding balls, there will be more rolling and less sliding than has been previously indicated. Certainly there is an interesting synchronization in those axes, that implies less-than-awful possibilities, with respect to pure rotation of that middle ball. Also, if the minor radius of the toroidal track is slightly larger than the radius of the ball, then a ball in that track will only be contacting the inner (or outer) race at just one point, and again rolling is more likely than sliding.

Next, consider this:
/ \ / \ / \
-O- O -O
\ / \ / \ /
which tries to show a similar axis-of-rotation thing for even more rings of balls than was originally described in the main text. The point I wish to make here is that only part of the overall load is being directed against the sidewalls of the bearing; most goes through the middle in the normal way. (And the reason I mentioned the unlikeliness of NARROW bearings in a prior annotation was because the obvious cure to side-wall loads is...thicker side-walls, thus defying the definition of "narrow".)
 — Vernon, May 13 2004

I vote we donate a few bucks each to buy Vernon a scanner.
 — RayfordSteele, May 13 2004

[RayfordSteele], I do have access to a scanner. I just have been too busy to do any scanning. Perhaps on Saturday...DONE!
 — Vernon, May 14 2004

[Freefall], if you see this, would you care to comment about the notion of combining ordinary steel bearings with tri-layer copper bearings? That is, the steel bearings would carry the load, so that there is nothing to dislocate the parts in the copper bearing. Thus, even if tri-layer bearings do have your stated flaw (which at this writing remains to be verified), the copper sort could still usefully replace brushes, for carrying electrical current between stationary and rotating parts.
 — Vernon, May 14 2004

The coppeer--ball-bearing is an interesting idea to replace stationary brushes in a slip-ring commutator system, but unfortunately copper tends to gall in that application, even if the loads are light. In addition, balls tend to have small areas of contact, which could lead to point damage during a high-current start. Roller bearings would be an improvement, but without a cage, even in a low-load environment, they would likely become misaligned and jam.
 — Freefall, May 14 2004

 [Freefall], thanks. Can you be more specific about "tends to gall in that application, even if the loads are light"? I specified practically no physical load, since ordinary steel bearings would be there to handle that. Are you saying that copper can't roll on copper without being damaged? Also please recall that I originally mentioned using graphite as an electrically conductive lubricant in such copper bearings. If that galling only occurs AS the copper rolls while conducting, then wouldn't the graphite help prevent it?

 Next, I do understand that rollers would be more conductive than balls, but I also have previously mentioned something about using lots and lots of small balls, which should offer practically the same or even more contact area than an ordinary number of ordinary-sized rollers.

Finally, while I know that alloyed copper is practically always rather less conductive than pure copper, nevertheless alloyed copper is also usually somewhat harder than pure copper. Perhaps, depending on anticipated electrical load, some tri-layer bearings could be made from alloyed copper, as additional galling-prevention (not to mention lasting longer even while running somewhat warmer due to small resistance losses).
 — Vernon, May 14 2004

[Vernon], I think the number of contacting points has a big effect on friction - not just the contra-rotating areas. More bearing element means more friction, and more heat, which also kills bearings.
The load bearing ability will be reduced by the additional force created by the non-perpendicular contact direction - if all bearings are the same diameter, then 30% reduction in load.
The size of the bearing will be increased by another layer. In many cases, bearings must fit on a shaft, and then inside a housing. The housing would need to be bigger.
I can see your idea is possible with cylindrical roller bearings, but not with any other types. The ball bearing idea, as others have mentioned already, will cause large slip angles. Tapered, and barrelled bearing also will not work.
I may be wrong, but I always thought that the reason for the cage was to stop grouping of the bearings. If they all collect together (quite likely), then there will be a gap in another area. This would cause uneven loading and possibly premature failure.
This brings me to a final point: bearing tolerance. More layers means less precision.

But the idea is novel: I wonder if there are any practical uses?
 — Ling, May 14 2004

 Consider this sketch...

 fishbones fishbones fishbones

 idea idea idea.

fishbones fishbones fishbones.
 — zigness, May 15 2004

My fingers are tired of assembling connecting rod needle bearings without cage, and they work better than the ones with, because more needles, the more load they stand. The reason for a cage in ball bearings is because they DISSASEMBLE if the balls group togheter. Or, how do you think they are mounted between the runs?. Anyway +, because I like the mental image of all the rollers simultaneously seizing, dissasembling and dissapearing in the engine's insides. (I shouldn't be so mean, I know).
 — finflazo, May 15 2004

 [Ling], friction is a major component of sliding, not rolling. Friction is very minor in most situations involving pure rolling. So, it doesn't matter much how many balls or rollers are rolling, so long as all are rolling and none are sliding. (Total mass of all those parts is likely a factor, too, since getting them to start rolling does require overcoming inertia. In a way, this is an argument FOR lots of smaller parts....)

 OK, I don't mind the notion that the cage in an ordinary bearing has a more important purpose of preventing clumping of parts, than of reducing rubbing between those parts. It was obvious that the tri-layer idea would work as an alternative means to prevent rubbing, but I must admit some uncertainty about the possibility of clumping.

Regarding lesser precision, I am reasonably sure that this is not a problem for tri-layer roller bearings. I can see a possible problem in ensuring that dozens of balls are all exactly the same size, but for rollers, well, if we can't just cut reasonably equal lengths from a constant-diameter rod, there are always lathes.
 — Vernon, May 17 2004

I checked about the cage: I was wrong, it is to reduce friction.
But there is another type of friction - rolling friction. A roller under load will deform slightly (a bit like a tyre on a car). A larger wheel has less rolling friction. But If the bearing is oscillating quickly, the rollers should be small as you mention.
For tolerances, the manufacturers would have two more measurements to worry about (the diameters of the second and third layers), and these would be at an angle which is determined by those measurements. Obviously, precision would not be as good as a single layer.
For clumping, the tri-layer bearing should never, ever do this, since then the outer layer of rollers would drop towards the centre layer. Any gap between same-layer bearings would result in the other layers moving in the radial direction.
Finally, wouldn't the number of bearings in each layer need to be the same? The gap between rollers in the outer layer would be bigger. So the outer layer might need to be a larger diameter to keep the contact angle reasonable.
 — Ling, May 17 2004

 [Ling], I have added three linked drawings. The number of balls or rollers in each row of each layer are indeed the same, and I have already annotated about using larger parts in the outer layers. However, this actually depends on the size of the gap between the bearing races. A small gap means large numbers of small parts, and in this case it is OK for all balls or rollers to be the same size (see "Tri-Layer Roller Bearing" link).

Regarding rolling friction, yes I know this is due to deformation of the rolling object under load. Obviously a small deformation is a greater percenatage (or ratio) of a small roller than of a large roller -- but equally obviously, a lot of small rollers will distribute the total load better, and so less deformation per roller will happen. I wonder if anyone has analyzed this to determine the optimum number of rollers or balls, to achieve the minimum deformation ratio, in a given ordinary bearing.
 — Vernon, May 17 2004

I vote for. The idea is a novel one. This doesn't necessarily mean I think it would work in a real world application, but the idea has potential. As for the contact points at the side, of the race, assuming cylindrical bearings, these could be minimized, by geometry. There wouldn't be no friction, but there'd be less friction.
 — mtoonsdale, May 17 2004

 I don't think I would make them out of any sort of copper alloy. Much too soft to last.

Too busy to do any scanning, and yet he can write a mile a minute...
 — RayfordSteele, May 17 2004

[RayfordSteele], copper or copper-alloy tri-layer bearings would ONLY be used for carrying electrical loads between stationary and rotating parts. Physical loads would require other bearings made of steel or other hard stuff.
 — Vernon, May 17 2004

 Two layers could be enough sometimes.

 //consider the following sketch: |C'C,| in which two adjacent rollers contact each other and rotate in opposite directions. // [They could] // rotate-in-place while the walls moved up or down in sync. //

 Say the walls (the inner race and the outer race) both move 1 cm. This will not be exactly the same angle, since they have different diameters. If they are nearly the same diameter they won't rotate much relative to each other unless the bearings rotate a lot. That could be a friction problem.

However, if the rollers are the same diameter as the inner race, the outer race must be more than three times that, and will rotate less than a third of the angle that the inner race does. So most of the inner race rotation will not be matched by the outer race rotation.
 — caspian, Dec 19 2004

I admit I haven't finished reading this, but based on the pictures and the first 5 paragraphs, I'd say it's pretty clever, regardless of practical problems. It might not be suitable to replace all current bearings, but that's hardly saying it has no use. For instance, it sounds like these could be easier to put together for prototypes and hobby stuff, when you need a custom bearing.
 — tiromancer, Dec 20 2004

 I tried unsuccessfully to get this to work. I didn't have a proper outer race, and the axle and 2 layers of 5 rollers were cut from some plastic tube (so basicly it was just 11 lengths of plastic tube and trying to hold them together with my hands) but they had a great tendency to pack into a hexagonal structure, and I'm pessimistic that a better setup would work.

I think the bearings could be unstable after all. I'm imagining this happening: Axle moves down, the rollers within each layer at the bottom are forced apart, which pushes the rollers together (within each layer) at the top until they touch. This takes more space at the top, but it's available as the axle moves down. Alternatively, it stops when rollers at the bottom fit into the widening gaps in their neighbouring layers.
 — caspian, Dec 24 2004

[caspian], yes, I do recognize the possibility. It is because the design depends on friction to keep the rollers or balls in place. Which means the solution is to use a different shape of roller (micro gear teeth, perhaps) or ball, such that neighbors interlock and prevent slippage.
 — Vernon, Dec 24 2004

 "Also, recall that I wrote something about different-sized balls or rollers in each layer. It would be sensible if, with respect the the middle layer, the inner layer had smaller-diameter parts and the outer layer had larger-diameter parts."

Don't think that would work the balls would move at different speeds and interfere with the movement of adjoining balls. but what the hell do i know anyway.

[mawgadog], all the bearings in any one LAYER turn at the same speed. For an ordinary single-layer bearing, the balls or rollers don't have to rotate in place as the inner race or outer race moves; they can roll along in-between the two races. For a multiple-layer bearing, All the layers will move together in-between the races, but the relative rotation rates of each layer can indeed differ depending on the sizes of balls or rollers. It DOES work out OK.
 — Vernon, Aug 05 2005

 Less typing, more building. Build it and customers will come (if it is worthy).

 The main negative thing I can see is the extreme loading of the rollers from the 45 degree angle of a roller centered between 2 others. Brinnel, anyone?

It is a noteworthy idea, I just don't know if it will have less or more friction. And for all the complexity, cost, and weight, it had better be way less friction.
 — moPuddin, Aug 06 2005

 hey, wait a minute, why didn't I think of this before?

 //all have the same drawback, in that the rollers or balls generally must not be allowed to touch each other, inside the bearing//

says who? bike mechanics I know have been ripping out cages and packing in one fewer balls than will fit into the race for decades (one more would actually be too many, the space left over is less than one ball diameter). (in all the regular bearings on a bike: headsets, cranks, wheels) never had a problem with bunching, and makes everything move more easily and last longer. maybe this is just a high speed problem?
 — oxen crossing, Aug 06 2005

[oxen crossing], that's interesting. However, one of the goals of this Idea was to allow a completely greaseless variant (not counting powdered graphite as lubricant), which could conduct electricity between stationary and rotating parts. Also, notice that that for a single-layer bearing (as in bicycle), if the balls or rollers clump a bit, the axle may get a bit off-center (not necessarily significant; I've occasionally ridden a bike to the store with blown bearings [to get replacements, of course!], and it works, even if somewhat wobbly). But for, a multi-layer bearing, if friction fails and clumping happens, the axle gets WAY off-center.
 — Vernon, Aug 06 2005

The axle doesn't get off center at all; the stresses become very very slightly assymetrical but the rotating part is held as rigidly if the bearing race were packed to its theoretical maximum.
 — elhigh, Aug 08 2005

Ok, I stand corrected on the off-center thing, for a single-layer bearing. Thanks.
 — Vernon, Aug 08 2005

no need to get pissy Vernon i was just making an observation that I thought was correct it seems it was not.

This is novel to me I think that if you define the topology then apply it to systems like chemistry orbitals or even just maybe spin states you will have a buckyballesque new clever thing.
 — beanangel, Aug 11 2005

A little late I see. Better late then never. I've been contemplating this very question for a long time. How do you build a bearing with only... ONLY rolling friction. After working with rolamite for awhile I tried some non banded ideas I had. The first one was to use three layer of rollers but after completing two in my modeling program the bearing worked. Another thing, this same bearing will fail even if the rolling elements are confined to roll around the inner and outer raceways they can bunch up on one side and unbunch on the other. Unless you force compliance by gearing the rolling elements this happens eventually. I did in the end find a solution. If you use your second example case with two rows of rollers you can use two non independant tracks to ensure spacing. one of the rows of rollers bear the load as in a normal roller bearings. Essentially you have two rows of rolling elements both held in place by either the outer or inner receways. you have to adjust for whatever allows two row rotary bearings to work in the first place, their non linearity, but then you basically have a cage with no sliding friction. put an inner raceway in and you might have a low friction bearing.
 — RichardT, Mar 10 2010

[RichardT], thanks for the input. I've been thinking about variations on the theme of "gearing" the rollers, to solve the bunching/unbunching problem, and suspect that quite modest bumps and grooves may suffice (thereby meaning the rollers roll more than they mesh). Your alternate solution sounds intriguing; can you link an image?
 — Vernon, Mar 10 2010

what if the rollers were smaller diameter on the ends, and the cage was 2 connected ring shaped parts that only contacted the rollers in the smaller diameter areas? it would be suspended off either raceway by the wider diameter meat of the rollers. spokes to connect the cage rings could go between the rollers, and also would be of smaller diameter than the rollers to avoid dragging on the raceway. If it is a large enough bearing to justify it, the rollers could have a ball-bearing interface on the ends where the cage holds it.
 — AutoMcDonough, Mar 10 2010

back: main index