What Preload Does in a Bearing Arrangement

Why this matters

Preload is the only load in a machine that nobody applies from outside. It appears when the internal clearance in a bearing goes to zero and keeps going, and from that point every rolling element is carrying force before the machine does any work at all. Sometimes that is exactly what a designer wanted. More often it arrives uninvited, from a fit that was slightly tight and a temperature difference nobody thought about, and it turns a correct-looking rebuild into a motor that will not stay running.

Before you touch it

Reading current on a running motor is energized electrical work under 29 CFR 1910.333(b)(2), which means an arc-flash risk assessment and the personal protective equipment that follows from it under NFPA 70E-2021; clamp on a conductor outside the enclosure where the installation allows it rather than opening a live panel to reach one. When you take it apart, isolate and lock out under 29 CFR 1910.147 for the mechanical side and 29 CFR 1910.333(b)(2) for the electrical side, and prove dead with the live-dead-live sequence of NFPA 70E-2021, 120.5. Wait for zero rotation rather than trusting the switch, and restrain a spring-loaded motor base before pulling belts, since that base is stored energy. Take temperature readings on a running machine with a non-contact infrared instrument from outside the guard, never by reaching past one. Arc-rated gloves are correct for the electrical task and are exactly what you must not wear near a shaft that can turn, so keep the two tasks separate and never carry the gloves from one into the other. When you mount a bearing with heat, use a controlled induction heater and hold at or below the manufacturer's stated cap, which for standard dimensionally stabilized bearing steel is commonly around 120 C, with gloves rated for that contact temperature. An open torch is uncontrolled and will alter the steel.

The call

A belt-driven blower motor comes back three weeks after a bearing rebuild. It starts fine, runs fine, and about twenty minutes in it trips its thermal overload. Reset it and it does the same thing. Nothing is obviously wrong: the belt tension checks against the drive table, the sheaves line up, the load has not changed, and the motor was quiet on the bench.

Do not replace the device that told you the truth

The tempting move is a new overload relay, because the motor "runs fine" and the relay is what stops it. Establish why the device opened before you treat it as the fault. A thermal overload that opens on a motor which is genuinely drawing more current than it should is doing precisely its job, and replacing it reaches the same end state as jumpering it, one step slower and with a part number on the invoice. The end state is a cooked winding.

So the first measurement is running current against the nameplate, at the same point in the run where it trips. On this motor it was climbing steadily through the run rather than sitting flat, and it crossed the nameplate figure a few minutes before the trip. That is not a relay problem. Something in the machine is getting harder to turn as it warms up.

Second measurement, non-contact from outside the guard: bearing housing temperature climbing well past what the same motor ran at before the rebuild, and still climbing when the trip came.

What preload actually is

A rolling bearing leaves the factory with radial internal clearance, a small amount of free play between the rings measured before it is mounted. Clearance classes step from C2 (tighter than normal) through CN (normal) to C3 and C4 (progressively looser). The class is stamped on the box and it is a duty decision, not a part-number detail.

Two things eat that clearance in service, and they add.

The interference fit on the shaft stretches the inner ring. Press an inner ring onto a shaft that is larger than its bore and the ring grows, which closes clearance directly. For a solid steel shaft and a standard-section bearing, roughly 75 to 80 percent of the diametral interference shows up as lost radial clearance. That fraction is derived for that geometry: a hollow shaft or a thin-section bearing changes it, and on a thin-section bearing the ring simply follows the shaft more closely.

The inner ring runs hotter than the outer ring. Heat generated in the contact leaves through the housing, so there is a temperature gradient outward, and the hotter inner ring grows more than the outer. The clearance lost is the coefficient of thermal expansion times the raceway mean diameter times the temperature difference.

When the sum of those exceeds the mounted clearance, the residual goes negative. Negative clearance is preload, and it does not know whether anyone intended it.

Where the clearance went

The rebuild used a CN bearing off the counter. The original had been C3.

Bore 50 mm, outside diameter 90 mm, so raceway mean diameter is (50 + 90) / 2, which is 70 mm.

Unmounted clearance. For this bore bracket the standard clearance table gives a CN band of roughly 6 to 23 micrometres. Take the midpoint, 14 micrometres. Read the bracket your bore actually falls in; the bands step at bore boundaries.

Fit loss. Shaft measured over the seat, bearing bore measured, both with a micrometer at matched temperature: 15 micrometres of interference. At 75 percent, that is 11 micrometres of clearance gone.

Mounted clearance: 14 minus 11 = 3 micrometres.

Thermal loss. Inner ring 15 C hotter than outer, coefficient for steel 11.7 x 10 to the minus 6 per C, on the 70 mm mean diameter:

11.7e-6 x 70 x 15 = 0.0123 mm = 12.3 micrometres

Note what that calculation is and is not. It is a diametral clearance change computed on the raceway mean diameter, and it assumes the outer ring sits at the housing temperature. Using the bore diameter instead of the mean diameter understates it by nearly a third, which is a common way to get a comfortable-looking answer.

Residual: 3 minus 12.3 = negative 9.3 micrometres. About 9 micrometres of preload that nobody applied, specified or measured.

The same motor with the class it left the factory with

Redo it with C3, whose band for this bore bracket runs roughly 18 to 36 micrometres. Midpoint 27.

• Mounted: 27 minus 11 = 16 micrometres
• After thermal: 16 minus 12.3 = positive 3.7 micrometres

Still clearance, and not much of it. That is the point of C3 rather than an argument against it. C3 is not generous, it is the class that leaves a small positive clearance at operating temperature on a machine with an interference fit and a real thermal gradient, which is why it is specified so widely on electric motors and belt-driven equipment. Substituting CN because it was on the shelf does not remove a safety margin, it removes the entire operating clearance.

Why twenty minutes and not immediately

The trip timing is the diagnosis confirming itself. Preload raises friction torque. Friction torque generates heat. Heat raises the temperature difference between the rings. A larger difference adds more preload. Each round is small, and the machine spends twenty minutes going round it before the current crosses the relay's threshold.

A machine that is over-preloaded from the moment it starts trips on start, and that is a different fault, usually mechanical interference or a mounting error. A machine that trips after a warm-up period is telling you the mechanism is thermal, and there are not many candidates: lost bearing clearance, a locating and floating arrangement where the floating end cannot float, or something rubbing that grows into contact.

What deliberate preload buys, and what it costs

Preload is not a defect. It is applied on purpose in plenty of arrangements, and it buys three things.

Stiffness. A preloaded bearing deflects less under a given load, because there is no clearance to take up first and because more rolling elements are in contact. That matters wherever shaft position under load has to be held.

Position and reversal control. With clearance, the shaft moves through it every time load direction reverses. Preload removes the free play and the impact that goes with it.

Skid prevention. At high speed and light load, elements can slide into the load zone instead of rolling. Preload keeps them loaded and rolling.

The costs are equally concrete. Friction torque rises and so does heat, which is the effect above. And because preload adds to the load every element carries, calculated fatigue life falls: the load-life exponent that punishes an external overload punishes an internal one identically.

Preload also changes contact geometry in an angular contact bearing, raising the operating contact angle. The axial load factors used to convert a combined load into an equivalent load are specific to contact angle, so a preloaded pair is not described by the factors printed for the unloaded nominal angle.

Setting preload with a number instead of a feel

Where preload is intended, it has a method with a measurable output. Four are in common use:

• Measured axial displacement, moving one ring a stated amount against the other
• Measured starting torque, turning the assembly and reading the torque required
• A ground spacer or a calibrated shim stack, which fixes the preload geometrically
• A spring giving a known force, most often a wave washer behind one outer ring

The last one is different in kind from the other three and worth understanding. A rigid preload set by a spacer converts thermal growth directly into a large preload increase, because the geometry cannot yield. A spring preload holds roughly constant force as things grow, absorbing the change in its own deflection. That is why a wave washer appears behind the floating-end bearing on so many small motors, and why leaving it out during a rebuild because it "looked like a shim" changes the machine's thermal behaviour rather than just its parts count.

"Snug it up until it feels right" produces a preload with no number attached, which cannot be checked, recorded or reproduced by the next person.

What would have changed the diagnosis

A machine where preload is intended. On a spindle designed around a preloaded angular contact pair, a rising current with temperature can mean preload has been lost through wear, not gained, and the evidence reverses: the fault shows as position error and chatter rather than as heat.

A floating end that cannot float. Same trip timing, same rising current, entirely different cause: the axial growth of the shaft has nowhere to go and is being reacted as thrust. Check that the floating outer ring is free in its housing before you conclude anything about clearance class.

A hollow shaft. The fit-loss fraction above assumes a solid steel shaft. On a hollow shaft the shaft yields to the ring rather than the other way round, less of the interference converts to lost clearance, and the same fit produces less preload.

A larger machine with a bigger raceway diameter. The thermal term scales with mean diameter, so the same 15 C gradient costs a bigger bearing proportionally more clearance. The clearance class question gets more important as machines get larger, not less.

References

• ISO 5753 for radial internal clearance classes and the bands by bore bracket, which step at bore boundaries
• ISO 281 for the rating life relationship, which is where the life cost of an internal load appears
• Bearing and motor manufacturer documentation for the shaft and housing fit tolerances, the clearance class specified for the machine, and the mounting temperature cap
• 29 CFR 1910.333(b)(2) (energized electrical work), NFPA 70E-2021, 120.5 (live-dead-live) and the arc-flash risk assessment; 29 CFR 1910.147 (hazardous energy and stored energy in a tensioned drive)
• See related: Radial, Thrust and Combined Loads; What a Bearing Load Actually Is