# Ball Bearings

## Contents

The basis of modern civilization, including the basis for the GVCS, is ball bearings. See how they are made:

# YK Discussion on IRC

The attached file is a text document of the roller bearing discussion between YK and logari81. Below is a summary of the key points discussed.

# Discussion

[13:49] <logari81> YK: I know quite a lot about ball bearing design
[13:50] <logari81> YK: I have significant experience with elastohydrodynamic lubrication
[13:53] <logari81> YK: I also have some experience with cageless designs (I have done a traineeship at INA Schäffler KG which is a major manufacturer of cageless roller bearings, so I have heard stuff about this design also)
[13:55] <logari81> FactorChris: what are the dimensions that we are mostly interested in?
[13:56] <YK> Ok, sounds good. I'd need the specifics on what questions are to be answered in the proposal though.
[13:57] <YK> Design dimensions are within 12 inches for the entire ball bearing, at least for the project I'm associated with.
[13:59] <YK> Logari, can you provide some information concerning what to look for in cageless designs?
[14:00] <logari81> YK: 12 inches outer diameter is quite big
[14:01] <logari81> YK: the problem with cageless is that the sum velocity at the contact point between the rollers is zero, so theoreticaly no lubricant enters the contact
[14:02] <logari81> what is the rotational speed?
[14:02] <YK> I meant only as a general statement- to correct myself, the outer diameter is more like 6 inches for the largest one required, at this point in time at least.
[14:03] <YK> The expected rpm is low, starting from a few hundred at the motor shaft then geared down through single stage spur.
[14:04] <logari81> one nice idea that I have seen is having a small hole in the rollers and a cage made by wire through this holes
[14:05] <logari81> this could work for relatively big bearings
[14:06] <YK> What size range would you estimate "big" at?
[14:07] <logari81> hmm, I think 6 inch outer diameter is big enough
[14:09] <YK> What design element allows sealing of the roller bearing for lubrication holding purposes?
[14:10] <YK> Also, how well do you think the wire cage would hold up under axial load?
[14:11] <logari81> there are commercial bearings with integrated seal, but it would be easier to use a commercial radial seal independent of the bearing
[14:12] <logari81> YK: the cage itself shouldn't take any of the load, its function is only to seperate the rollers or balls
[14:13] <logari81> YK: the critical loads on the cage are usually centrifugal, for high rotational speeds
[14:13] * logari81 goes afk, will be back later
[14:34] <logari81> back
[14:45] <logari81> AFAIK smaller balls/rollers are finished with tumbling:
[14:45] <logari81> http://en.wikipedia.org/wiki/Tumble_finishing
[14:45] <logari81> and bigger ones are turned+ground
[14:46] <logari81> I think tumbling requires quite a lot of know how gained through lots of try and error
[14:46] <logari81> turning n grinding are more straightforward processes
[14:53] <YK> What design points do you have in mind for the inner and outer races?
[14:55] <YK> Also, for the aforementioned outer diameter of roller bearings, which method of finishing would you lean towards?
[14:56] <logari81> YK: first of all are we talking about ball bearings, cylindrical roller bearings or tapered roller bearings?
[14:57] <logari81> YK: are you aware that all big bearing manufacturers provide CAD drawings for their product?
[14:57] <FactorChris> back
[14:58] <logari81> of course they don't include the exact profile corrections for the races and the rollers but it is a good start
[14:58] <FactorChris> Logari. As I understood it, manufacturing of ball bearings was the main interest.
[15:02] <logari81> FactorChris: I do not have plenty of time, but if there is interest I could help with design issues at least
[15:02] <YK> cylindrical roller bearings, and yes, I've taken a look at skf's site for design considerations.
[15:04] <logari81> YK: I think for cylindrical roller bearings most of the design effort is put onto the rollers, with the races having a quite theoretical cylindrical form
[15:05] <YK> What would you say to having the outer race be comprised of 3 moderate thickness plates?
[15:07] <logari81> for axial loads the geometry of the design of the flanges is also important
[15:07] <logari81> you mean one ring and two separate flanges at the sides?
[15:07] <YK> yes
[15:08] <YK> Leakage comes to mind, but it appears so much simpler to fabricate
[15:08] <logari81> YK: like the inner ring of these here http://www.skf.com/skf/productcatalogue/jsp/viewers/imageViewerJs.jsp?image=0601f12.gif&file=1_4_2&maincatalogue=1&lang=en ?
[15:08] <logari81> but with two separable flanges?
[15:09] <YK> yes, similar to that iamge
[15:09] <YK> yes
[15:09] <logari81> I don't find it a bad idea, but I wonder why nobody does it like this
[15:09] <logari81> I think it should work
[15:10] <YK> then now, what would you say to having one of the outer plates be the frame of the robot?
[15:10] <logari81> but concerning the machining I am not sure that it is easier
[15:11] <logari81> having one middle ring and two flanges means that you have to grind 6 faces
[15:12] <YK> One side of one flange, one side of the other flange, and the inner part of the ring, making 3 faces, isn't it?
[15:12] <logari81> if one of them is part of the housing, you spare one grinding
[15:13] <logari81> making 4 you mean
[15:13] <logari81> aaah
[15:13] <logari81> I am talking only about the planar surfaces, I meant 6 planar + 2 cylindrical
[15:14] <YK> why 6 planar? I can only think of 2, one face of one plate and one face of the other outer plate.
[15:15] <logari81> what about the mate surface between the side flanges and the middle ring?
[15:16] <logari81> but all this depends on the precision you want to achieve
[15:16] <YK> as for the machining, I can only see the need for torching 3 holes, 1 for each plate, excluding mount holes.
[15:17] <YK> why does one need to grind the mate surface?
[15:19] <FactorChris> guys, this discussion is excellent, is it possible you could capture it in an email, wiki, or google docs of somekind?
[15:19] <logari81> if you want to achieve a decent accuracy you have to, but maybe you don't need such high quality
[15:19] <logari81> in any case you have to grind the races
[15:20] <logari81> and also think that the rollers will also slide against the side flanges, so, it would be nice if these faces were ground
[15:22] <logari81> bearings are usually the most critical element of a machine, so if I would make them home made, I would make them as good as I can
[15:22] <YK> yes, making a minimum of 2 planar + 2 cylindrical surfaces. can you explain how grinding the mating surface helps accuracy? [I'll copy/paste this into a text document then put it up on the wiki, FactorChris. Under what heading though?.]
[15:25] <logari81> the total width of the bearing is t_L + t_M + t_R, by grinding all six planar faces of L,M,R you achieve the maximum accuracy for each of t_L, t_M, t_R
[15:28] <logari81> especially if you want a roller element bearing to carry some (small) axial load also, the meaning of accuracy in width dimension is not to underestimate
[15:28] <logari81> width direction*
[15:29] <logari81> but this is also something that can be improved in later versions, you can start with the simplest case
[15:30] <YK> Ok, I understand now about the grinding. Do you think a single row cageless roller bearing design would work well in conjunction with this triple plate idea?
[15:32] <YK> (or, in light of the theoretical lubrication problem, is the cage too important to avoid)?
[15:32] <logari81> the design should work. A minimum condition however is that the races and the rollers are ground
[15:33] <logari81> and if cageless (full complement) the rollers should have as low roughness as possible
[15:34] <YK> if caged, what are the design elements of the cage?
[15:35] <logari81> one other idea is to make the side flanges of bronze, it is easier to manufacture in general
[15:36] <YK> what are the tradeoffs associated with bronze?
[15:37] <logari81> the cage could be everything that keeps the rollers apart, think of wedges put between the rollers and kept together by two further side flanges
[15:38] <logari81> bronze maybe will wear a bit faster if not well lubricated, but in a non-ground case, think of wear like running in
[15:40] == chully [~chully@c-69-136-0-188.hsd1.in.comcast.net] has quit [Read error: Operation timed out]
[15:41] <YK> can you expand on the wire cage design you mentioned earlier, and other possible cage designs that involve simple fabrication techniques?
[15:41] == chully [~chully@c-69-136-0-188.hsd1.in.comcast.net] has joined #opensourceecology
[15:46] <logari81> YK: the simplest cage is made from a plastic tube with diameter fitting between the outer and inner ring
[15:47] <logari81> you cut a part of the tube as wide as the bearing and you open rectangular windows in radial direction
[15:47] <logari81> I mean one window per roller
[15:47] <YK> how long would such plastic fittings last?
[15:49] <logari81> I think the problem with this design is not the lifetime of the cage
[15:50] <logari81> the problem is that the spacing between the rollers is quite high, so you have less rollers in total to carry the load
[15:51] <logari81> in this more sophisticated cage http://en.wikipedia.org/wiki/File:Cylindrical-roller-bearing_din5412-t1_type-n_ex.png the spacing between the rollers is reduced to a minimum
[15:52] <logari81> but if you don't have mould injecting or something like this is more difficult to achieve such complexity
[15:52] <logari81> or you come up with some other clever idea
[15:53] <logari81> you can take some ideas from here http://skindustry.tradeindia.com/Exporters_Suppliers/Exporter12110.169520/Ball-Bearing-Cage.html
[15:54] <logari81> this is more like the "tube" variant:
[15:55] <logari81> I meant this
[15:58] <YK> Both look difficult to machine. Could you expand on the lubrication requirement difference between caged and cageless designs?
[16:00] <YK> This seems to be the critical junction in design.
[16:02] <logari81> actually between the race and the roller you expect to have very little metal contact, normally the lubricant film should separate the two surfaces. However between two rollers, you will most likely have metal contact, so you need to have a very low surface roughness to minimize this effect and possible fretting wear
[16:04] <logari81> in other words the contact roller-roller is more roughness-sensitive than the contact roller-race
[16:05] <YK> What would you approximate as the general lifetime difference between caged and cageless roller bearings? How much of this difference can be minimized given low roughness rollers?
[16:08] <logari81> a cageless bearing if well designed should theoretically last longer, because you have more elements to carry the load (this statement is under the assumption that the roller-race contact is the failure criterion)
[16:11] <logari81> I cannot do an estimation but I believe the full complement design could work well for many situations
[16:13] <FactorChris> Greetings
[16:13] <FactorChris> YK and Lorari81
[16:13] <YK> That seems to be the simpler fabrication route as well. What are the concerns associated with having too much space between rollers in a full complement roller bearing?
[16:13] <FactorChris> I just talked to Marcin about the stuff you're working on the ball bearings. He wanted to talk with you guys.
[16:14] <FactorChris> Could you send him an email with the things you've been talking about?
[16:14] <FactorChris> opensourceecology@gmail.com
[16:14] <FactorChris> and please cc me on that factorchris@gmail.com
[16:14] <FactorChris> Thanks guys
[16:15] <YK> Sure, I'll send you two the email.
[16:18] <logari81> YK: you shouldn't have to much space, because the rollers would gather all together and leave a large space at the other side. The will not remain evenly distributed.
[16:20] <logari81> YK: look another idea for a cage http://www.ina.com/content.ina.de/de/ina_fag_products/productinformation/rotativ_products/cylindrical_roller_bearings/lowfriction_cylro/reibungsarme_zylro.jsp
[16:20] <YK> Would a space equal or less than the length of 1 roller raise any problems?
[16:21] <logari81> but be aware that the cage is not connected to the outer ring
[16:21] <logari81> I have the impression that the space is like half a roller
[16:22] <logari81> but it is easy to check from commercial bearing catalogs
[16:24] <logari81> YK: take a look here for example http://medias.ina.de/medias/en!hp.ec.br/SL1818;aCZgqVzMLDrg?#SL1818
[16:24] <YK> It would appear that using that cage design would result in moving from 3 to 5 plate design. How large, relatively, is roller-flange contact friction?
[16:25] <logari81> YK: no, it is not a 5 plate design
[16:25] <logari81> YK: the cage plate should be totally independent
[16:26] <logari81> if you press it between to further plates it cannot rotate anymore
[16:26] <logari81> the solution is that the middle ring has a groove in it
[16:27] <YK> I meant the 2 additional plates to form the extra groove, while keeping the cage independent of the plates. (and assuming that no grooves are to be machined into the plates)
[16:27] <logari81> but I don't think it is a nice solution, it looks nice but it isn't
[16:28] <logari81> aha, ok, but then you could move the groove to the sides (having two grooves and two cages instead of one)
[16:29] <logari81> the problem is that you reduce the effective width of your bearing, I don't like this design, even if they sell it
[16:31] <YK> I agree- overall, I think there are better alternatives.
[16:32] <YK> What concerns arise from having too much space between flanges?
[16:32] <logari81> here you have a dxf drawing where you can probably measure the spacing between elements
[16:32] <logari81> (in case that you hadn't found it)
[16:32] <YK> I'm unable to load the page
[16:33] <logari81> YK: then try with this http://www.tracepartsonline.net/%28S%28vqwvex3qwtyfjh55p3u2kd55%29%29/global/parts.aspx?caddy=1
[16:34] <YK> what should I look for?
[16:37] <logari81> YK: nothing, unfortunately the total number of rollers is not included in the drawings
[16:40] <logari81> YK: but in the cross section drawings you have some interesting details about the rollers
[16:40] <logari81> e.g. rounding of their edges
[16:41] <YK> I can't seem to see the image. I get directed to a search page
[16:41] <logari81> YK: and having the roller diameter you can probably guess their number, try to download the cross section 2D drawings
[16:42] <YK> ah yes.
[16:44] <logari81> YK: you should get something like this http://imagebin.org/161434
[16:47] <YK> I see it.
[16:49] <YK> the flange-to-flange tolerance seems incredibly small. Is it ok to have relatively larger space between the roller and the flange?
[16:52] <logari81> YK: it depends if you want to use the bearing for axial positioning of the shaft or not
[16:52] <logari81> in general cylindrical roller bearings are not very appropriate for this use
[16:56] <YK> How well do roller bearings respond to radial force applied at a distance from the centre?
[16:59] <logari81> you mean moment loads, they are not supposed to carry any moment loads
[16:59] <logari81> in this case you need two bearings
[17:01] <YK> What will be the roller bearing's reaction to an offset shaft rotation?
[17:03] <logari81> offset shaft rotation?
[17:04] <YK> say, a shaft within the roller bearing is connected via a shaft coupling to another shaft at an angle
[17:05] <YK> *with an offset*
[17:06] <logari81> you mean a cardan coupling?
[17:07] <YK> without a cardan coupling.
[17:07] <YK> and with, that would be good information too.
[17:08] <logari81> if you have an angle other than zero you will always have a moment, so that you will need two bearings
[17:08] <logari81> with one single bearing you can tolerate very small angles I think
[17:09] <logari81> in case of cardan it becomes a bit worse because this bending moment is also variable1