I often build things (3D printers, skateboard wheels, bicycle parts) that use off-the-shelf ball-bearings. These ball bearings are available in nominal mm sizes, and to the limit of my measuring tools, tend to be ground exactly to size. For example, a bearing with 15mm ID and 24mm OD will have ID and OD of exactly 15 and 24mm, to my ability to measure.

It's also easy to buy linear or rotary shafts ground to nominal diameters. For example, I can buy 15mm ground rotary shafts. These are usually ground to a +/- tolerance which means half of them won't fit the bearings, or the more precise ones are ground to an upper spec of 15mm and very small tolerance smaller. Usually, even with the 15mm upper spec, the size is too close to 15mm for a bearing to slide on. I have to polish the shaft down with sandpaper (destroying the nice finish) in order to get an bearing that can be assembled by sliding, such as for a bicycle hub axle.

In cases where you want a bearing to have a close sliding fit on a shaft, it seems two things are possible: Either you could buy special ball bearings with slightly over-nominal size (i.e. 15.15mm ID), or you could buy shafts with slightly under nominal size (14.85mm). However, there doesn't seem to be a common "standard undersize" shaft or "standard oversize" bearings for each nominal mm size; neither thing seems to be common. This leaves a world of 15mm shafts and 15mm bearings which don't fit each other except by interference fit.

To get a bearing and a shaft which fit each other with a close sliding fit, is it more normal to use an undersize shaft or an oversize bearing? If the former, why does it seem so hard to find shafts commonly available to fit nominal bearing sizes?

  • $\begingroup$ Just checking: you are deburring the ends of the shaft right? And perhaps even chamfering it with a file? $\endgroup$ – DKNguyen May 15 at 7:56

Shafts and bearings usually use interference or transition fits - they shouldn't slide and for installation you should use force - some kind of press or clamping tool would be best, eventually if such tools are not available - few gentle hits with a hammer (you shouldn't hit the bearing directly to avoid damage, use something, for example piece of wood, to spread the force onto the whole bearing). Alternatively, you can heat/cool one of the parts to make use of thermal expansion.

As for the undersize/oversize I suggest you to read some basic information about engineering tolerances. The nominal size is obviously only theory, so we use tolerances to describe the allowed variation in the actual dimenisions. Depending on the type of tolerance we allow them to be for example only higher than nominal, only lower, always higher by certain number, either lower or higher etc. And when we assemble 2 parts that have certain tolerances we have engineering fit.

As for what's more common - it's more common to have hole-basis fit in which:

  • the hole (in this case inner race of the bearing) is as close to nominal size as possible but never smaller (tolerance H)
  • the shaft has tolerance selected for desired fit - so if you want sliding you would make it undersized, or if you want interference fit you would make them oversized
  • 2
    $\begingroup$ omg no! please don't hit it with a hammer! Use an arbor press, or a vise, or a clamp... anything $\endgroup$ – Pete W May 14 at 19:15
  • $\begingroup$ @PeteW I agree that's the best method when available - I edited the answer to highlight that. But I don't think hitting the bearing through for example a piece of wood with moderate force, appropriate to the size and type of the bearing would severly damage the bearing. Although I would never suggest such method for proffesional, I think it's best alternative for hobbyists for which such tools aren't available. $\endgroup$ – elementiro May 14 at 19:41
  • $\begingroup$ A light touch with a rubber dead blow mallet never hurt a garage project $\endgroup$ – jko May 14 at 19:49
  • $\begingroup$ Seen a bearing heater used on a 12" diameter bearing (expensive) and the heater ´stat had failed - the bearing came out glowing. Foreman was NOT happy we thought it was :) $\endgroup$ – Solar Mike May 15 at 6:33
  • $\begingroup$ Get some long-ish balsa blocks or something and drill some holes into it just slightly larger than the shaft diameter. Then tap it in. Better to not use something like a hammer with a handle that acts as a lever. Instead use a mass you directly hold in your hand for more control. Like a hunk of brass, though if you are banging on a wood block with a hold drilled in the middle any material would do. $\endgroup$ – DKNguyen May 15 at 7:58

Typically you want an interference fit on a bearing ID since the shaft is usually the element that's rotating and whatever the bearing OD goes into will have a looser fit. If your shaft is rotating and has a loose fit with the bearing ID it can start to score and cause a whole bunch of issues. You can typically find undersized dowels which could work in your instance.

Standard practice is to heat the bearing with a bearing heater, or cool the shaft, prior to assembly. The thermal expansion (or contraction) allows just enough clearance for assembly. Shaft fits are also dependent on the load and application of the bearing, usually based on ISO clearances. This is a useful resource from a reputable manufacturer. For a 15 mm bearing with line to line fit or a little tighter you may be able to just use an arbor press and some elbow grease.

  • $\begingroup$ All the hubs I have fitted bearings into the bearing OD is a press fit. $\endgroup$ – Solar Mike May 14 at 19:26
  • $\begingroup$ The inner race of the bearing will be captured in operation with an axial bolt/spacer/lockring, so there's no chance of slippage. But the bearing needs to be a slip fit on the shaft so that it can be assembled. $\endgroup$ – BetterSense May 15 at 0:04
  • $\begingroup$ @BetterSense fyi any play will wear out the bearing faster from impact loads on fewer points of contact. If you want the full design life out of it, you need those accurately controlled interferences that seem unrealistic when you look at the spec numbers. $\endgroup$ – Pete W May 15 at 14:06

Bearing manufacturers will give values for interference based on the application. This does not involve buying shafts off the shelf. It involves setting the spec and having a machinist create the right sized shaft. Pump and motor manufacturers make their shafts to the right specification, which is most certainly not by buying them off the shelf. We are talking micrometers as the measurement accuracy required. All the major bearing manufacturers have guides for this (SKF, FAG, etc.). A millwright's handbook provides guidance on installation procedures. Not all bearings can be put on in a press, so field techniques also exist (yes, hammers are involved).

Obviously there is in-depth engineering that goes into setting the standards and the resultant specs, but no engineering knowledge is necessary to use the guides. The skillset for installation is a mechanic's, not an engineer's.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.