A gearbox input shaft, 25 mm diameter, carrying a deep-groove ball bearing. The bearing was pressed on with a k5 fit — interference of 0 to +15 μm. That’s the textbook fit for a rotating shaft bearing. The bearing seized on the shaft after 500 hours. Not the bearing’s fault — the shaft had worn where the bearing sat. The inner ring had fretted against the shaft, leaving a reddish-brown oxide and a 0.02 mm groove. The bearing had been spinning on the shaft instead of with it.
Fretting on a bearing seat usually means one of two things: the fit was too loose, or the load was higher than the fit’s capacity. The shaft was k5 — nominally correct. The customer measured the shaft: 25.000 mm. The bearing bore: 25.008 mm. That’s a clearance of 8 μm, not interference. The bearing had been sliding on the shaft since day one. The k5 fit was on the drawing. The actual parts were assembled with a loose fit.
What interference is actually needed
For a bearing inner ring on a rotating shaft, the fit must prevent the ring from creeping relative to the shaft under load. The required interference depends on the radial load and the bearing size. A rough rule for light to medium loads: 1 μm of interference per mm of shaft diameter. For a 25 mm shaft, that’s 25 μm of interference. The k5 fit provides 0 to +15 μm. That’s marginal. The actual parts had clearance.
The precise method is in ISO 286 and the bearing manufacturer’s handbooks. The load F, the bearing bore diameter d, and the bearing width B determine the tangential stress at the bore. The required interference is:
Δd = (0.8 × F × k) / (π × B × E)
Where F is the radial load, k is a factor based on the hub/shaft ratio (about 0.5 for a solid shaft), and E is the modulus of elasticity. For a 5 kN radial load on a 25 mm bore with 15 mm width: Δd = (0.8 × 5000 × 0.5) / (π × 15 × 207000) = 0.0002 mm. That’s 0.2 μm. Almost nothing. So why did the bearing creep?
Because the bearing inner ring is thin. It flexes elastically under the rolling element load. Even a tiny clearance opens a path for the ring to creep relative to the shaft. Creep happens at 1-2 μm of clearance, and the fretting follows. The bearing doesn’t need much interference — it needs any interference at all. A clearance fit on a rotating shaft is a fretting machine.
What fixed it
The shaft was re-machined to m6 (+8 to +28 μm). The new bearing pressed on with 15-20 μm of actual interference. The fretting stopped. The bearing stayed put. The shaft ran for three years without the seat wearing.
But the deeper fix was measuring, not assuming. The original drawing said k5. The machinist turned the shaft to 25.000 mm — within the k5 tolerance band (0 to +15 μm) — but the bearing was 25.008, so the parts didn’t fit as intended. The tolerance on the bearing bore is a minus tolerance (0 to -8 μm), meaning the bearing bore is 24.992 to 25.000 mm. The machinist ignored the bearing’s actual dimension. A k5 shaft at 25.000 mm against a bearing at 25.008 doesn’t exist — the bearing bore tolerance doesn’t go that way. The measurement showed the bearing was a cheap import with a plus tolerance, out of spec. The lesson: measure the actual parts, don’t trust the class symbol.
Fit selection table
| Application | Shaft fit | Interference (25 mm shaft) |
|---|---|---|
| Bearing, rotating shaft, light load | k5 / m5 | +0 to +20 μm |
| Bearing, rotating shaft, heavy load | m6 / n6 | +8 to +30 μm |
| Bearing, stationary shaft | j6 / k6 | -5 to +15 μm |
| Gear hub, keyed | H7/p6 | +2 to +25 μm |
| Coupling hub on motor shaft | H7/m6 | +0 to +25 μm |
Assembling a press fit without damage
Heating the bearing inner ring (to 80-100°C, never more — above 120°C the metallurgy changes) lets it drop onto the shaft without hammering. Or use an induction heater. Or press with a proper arbor press — never a hammer, the load path through the rolling elements damages the races. And check the fit before assembly: measure the shaft and the bearing bore, calculate the actual interference, and confirm it’s within the intended band. Ten minutes of measuring beats a fretted shaft.
A rotating bearing needs interference, not clearance. k5 on the drawing is a promise, the parts are the reality — measure both. The fretted shaft wasn’t a bearing problem, it was a 8 μm clearance that let the ring creep. Press fits fail quietly, then loudly.
