Matching Tolerances of Bearings and Shafts, Bearings and Holes

• 1
  What a Bearing Fit Actually Does
• 2
  How Interference Changes Internal Geometry
• 3
  Reading the Shaft and Housing Tables Correctly
• 4
  Machining and Measuring the Seat
• 5
  When the Standard Fit Tables Do Not Apply
• 6
  Checking the Fit Before You Cut Metal
• 7
  Typical Shaft and Housing Fit Choices by Load Case

A rolling bearing is a finished precision component. The inner ring is made to a tolerance class, the outer ring to a tolerance class, and neither is meant to be reworked. What the machine shop controls is the shaft diameter and the housing bore. Those two dimensions decide how the rings sit, and the ring that carries the rotating load must not slip on its seat.

Creep is the slow failure everyone forgets. If the inner ring turns on a shaft that is too small, the ring face polishes the shaft, wear debris enters the grease, and the bearing fails long before its rating life. If the fit is too tight, the inner ring grows and the radial internal clearance drops. The balls get pinched, friction rises, and the bearing runs hot.

So matching tolerances of bearings and shafts is not about hitting one number. It is about picking an interference or clearance band that matches the load, the speed, and the temperature the assembly will see. Get the direction of that band wrong and no amount of bearing quality will save the design.

• 1
  Rotating load ringNeeds interference so it cannot creep.
• 2
  Stationary load ringCan be a light fit, sometimes a small clearance.
• 3
  Split housingUsually a looser outer fit than a solid one.

When you press an inner ring onto a shaft, the ring bore is stretched. The ring does not stay round in the free state; it follows the shaft. Every micrometre of interference becomes a small reduction in radial internal clearance. On a 30 mm bore bearing with a few micrometres of interference, that loss is modest. On a thin-section bearing, it can eat most of the clearance class you paid for.

The outer ring behaves the same way in reverse. A tight housing bore compresses the outer ring and closes the clearance further. That is why a bearing that is tight on both rings can feel smooth by hand and still seize after ten minutes of running. The clearance was already gone before the temperature rose.

Heat adds to this. A steel shaft grows about 11 × 10⁻⁶ per °C. A 40 mm shaft that runs 40 °C above ambient grows roughly 0.018 mm. If the housing is aluminium and grows faster, the outer fit loosens while the inner fit tightens. The fit you measure at 20 °C is not the fit you get at operating temperature.

None of this is a reason to avoid interference. It is a reason to start from the clearance class you need, then work back to the fit that keeps it.

ISO 286 gives you letters and numbers: h, j, k, m for shafts, H, J, K, M for holes. Lowercase is a shaft, uppercase is a hole. The number is the tolerance grade, and smaller numbers mean a tighter band. A shaft marked k5 has a tighter band than k6, but both are usually interference fits on a nominal bearing bore.

The first question is which ring rotates relative to the load. Not which part spins. A shaft can spin while the load direction stays fixed, and in that case the inner ring sees a rotating load. But if the load rotates with the shaft, the inner ring sees a stationary load and the outer ring may be the one that needs the interference. Engineers get this backwards more often than any other point on the page.

The second question is how much interference the bearing can take. Bearing catalogs list the recommended shaft and housing tolerances for each series and load case. Those tables already account for the clearance loss. If your application is unusual, such as a thin-walled housing or a plastic housing, the catalog values are a starting point and not an answer.

The third question is what the shop can actually hold. A k5 band on a 30 mm shaft spans only a few micrometres. If the lathe and grinder cannot hold that band repeatably, a nominally correct fit becomes random in production. That is a process question, not a design question.

• 1
  Lowercase letterShaft tolerance, for example k6.
• 2
  Uppercase letterHole tolerance, for example H7.
• 3
  Smaller grade numberTighter tolerance band.
• 4
  Catalog tableAlready includes clearance loss.

Turning a bearing seat is not the same as turning a general diameter. The seat usually needs grinding if the band is tight. Turning alone can hold ±0.005 mm on a good machine, but the surface finish matters too. A turned seat at Ra 1.6–3.2 μm is fine for a light fit. A ground seat at Ra 0.2–0.8 μm is what you want under an interference fit, because a rough surface loses metal during assembly and the fit loosens.

Measure the seat, not the shaft end. A shaft can be perfectly round at the outboard end and tapered near the shoulder. Use a micrometer at three positions along the seat and two directions at each position. Record the numbers. A single reading at one spot tells you almost nothing about whether the ring will seat properly.

For housings, check roundness as well as diameter. A three-jaw chuck can leave a triangular bore that a two-point bore gauge will not see. A tight housing that is out of round pinches the outer ring in three places and the bearing runs rough.

Assembly matters at this stage too. Press fits should be done with an arbor press or a controlled temperature difference, not a hammer. Heating a bearing above roughly 120 °C can change its heat treatment and its dimensions. If you need a large interference, heat the housing or cool the shaft, and stay inside the bearing maker's limits.

• 1
  Ground seatRa 0.2–0.8 μm for interference fits.
• 2
  Three positionsMicrometer along and around the seat.
• 3
  Roundness checkTwo-point gauges miss three-lobed bores.
• 4
  No hammerUse press or controlled temperature.

The catalog tables assume a solid steel shaft and a rigid, thick-walled housing. Change either and the numbers shift. A hollow shaft deflects more under the same interference, so the effective grip is lower. A thin aluminium housing expands faster than a steel outer ring, so the outer fit loosens when the assembly warms up.

Plastic and composite housings are a separate case. Their creep behaviour means an interference fit relaxes over time, and the outer ring can start to turn. A common answer is a metal insert or a retaining compound, but that is a design decision that has to be made before the parts are cut.

Very small bearings are another boundary. Below about 10 mm bore, the ring section is thin and a normal k6 fit can distort the raceway. Very large bearings go the other way: the interference needed to prevent creep becomes large, and the clearance loss has to be added back into the bearing selection.

High-speed spindles sit at the edge of all this. The fit has to hold the ring at low speed and stay safe at 20,000 rpm, where centrifugal growth and heat both change the geometry. That is a case for a specialist, not a general table.

A few minutes of checking saves a scrapped shaft. First, confirm the bearing part number and its tolerance class. A P6 bearing and a P0 bearing of the same size do not have the same bore deviation, so the same shaft fit does not produce the same result.

Second, calculate the clearance loss. Take the interference you expect, convert it to a diameter change on the ring, and subtract it from the radial internal clearance. If the remaining clearance is below the minimum for your speed and load, the design is too tight.

Third, look at the temperature. Estimate the running temperature of the shaft, the housing, and the bearing. If the difference between them is more than about 20 °C, the cold fit is not the working fit.

Fourth, decide how you will inspect. For a k5 or m5 seat, a hand micrometer is not enough on the shop floor. A bench gauge or a comparator with a setting master gives repeatable readings. Write the inspection method on the drawing so the next batch is made the same way.

Finally, test one assembly before the full run. Press one bearing onto one shaft, check the drag torque, and run it for a short period. The feel by hand is a weak signal, but a change in drag torque after a few minutes tells you the clearance is closing. That is a real failure mode caught early.