How to Tell Whether a Bearing Is Loaded the Way It Was Meant to Be

Why this matters

There are two ways to load a bearing wrong and only one of them is obvious. Everybody knows an overloaded bearing dies early. Far fewer techs know that a bearing carrying too LITTLE load also fails, and fails in a way that reads like a lubrication problem, so it gets a grease schedule instead of a diagnosis.

This procedure builds two independent readings of the same bearing and makes them agree. Either one alone will lie to you.

Step 0: make the machine safe before anything else

Isolate every energy source and lock out under 29 CFR 1910.147. Then wait, and look: a fan wheel coasts for minutes after the contactor drops, so verify zero rotation rather than assuming the switch did it. A spring-loaded or gravity-tensioned motor base holds stored energy under that same standard, so restrain or de-tension it before the last belt comes off, or the base will move on its own. Belt, sheave, shaft and coupling guarding is required by 29 CFR 1910.219 in general industry and 29 CFR 1926.300(b) on a construction site, and goes back on before the machine runs. If any part of this trip involves the motor terminal box, that is electrical work under 29 CFR 1910.333(b)(2), proved dead by live-dead-live per NFPA 70E-2021, 120.5.

When you turn the shaft by hand to feel it, keep your hands on the shaft or coupling hub and out of the sheave nip, where the belt runs onto the groove. A nip point will take a finger even at hand speed. Do not wear gloves for this: near anything that turns, a glove is an entanglement hazard, not protection.

The two columns

Column one is computed. What the geometry says each bearing should be carrying, in pounds of force, from named sources.

Column two is observed. What the installation and the removed bearing physically show.

Steps 1 through 4 build column one, steps 5 and 6 build column two, and the last section reconciles them. You need both: the computed number rests on assumptions you cannot see (the tension the drive was set to, where the sheave actually sits, whether the floating bearing can float), and the observed evidence has no scale at all.

Step 1: establish which bearing locates and which floats

Find which bearing holds the shaft axially and which is meant to slide in its housing as the shaft grows with heat. The documentation says which; on a pillow-block fan shaft the locating unit usually has its inner ring positively clamped.

Skip this and everything downstream is misread. A floating bearing corroded or fretted into its housing cannot float, so thermal growth becomes an axial load nobody calculated, rising as the machine warms. That reads exactly like an overload, and re-checking belt tension finds nothing wrong, because nothing is wrong with the belt.

Step 2: list every force with a named source

Write them down. Each one gets a source you can point at:

• Belt or chain pull, from the drive
• Weight of the rotating assembly, and where along the shaft it acts
• Axial thrust from pressure across a wheel or impeller, from the equipment manufacturer's data
• Coupling reaction from misalignment, on a direct-coupled machine
• Any force from connected piping, ducting or a rigid connection that grew with temperature

If you cannot name a force's source, do not put a number on it. If it clearly exists and you have no number, carry it as an open item without one: it will explain a disagreement later.

Step 3: run the statics

For a load applied between the bearings at distance b from one of them, with span L, that bearing takes F x (L - b) / L and the other takes F x b / L.

For a load overhung beyond the bearings at distance a, with span L, the near bearing takes F x (L + a) / L and the far bearing takes F x a / L in the opposite direction. The near bearing carries more than the entire applied load. If your two overhung reactions add up to the applied load rather than differing by it, you have modelled it as if it sat between the bearings.

Forces acting in different planes combine as a vector, not a sum. Belt pull from a motor mounted beside the fan is horizontal; weight is vertical; the resultant is the square root of the sum of their squares.

Step 4: bound it at both ends

Two numbers from the bearing catalogue, not one.

The upper bound is the basic dynamic load rating C, which is what the rating-life calculation compares against.

The lower bound is the manufacturer's minimum load. The conventional order of magnitude is around 1 percent of C for ball bearings and roughly double for cylindrical rollers, but it is a real catalogue value that rises with speed and depends on the lubricant, so read yours rather than treating the rule of thumb as an answer. Below it, the elements entering the load zone are not driven hard enough by traction to roll, so they skid.

Step 5: measure what makes the computed number wrong

Column one assumed things. Go check them, machine locked out and base restrained:

• Belt tension, with a tension tester against the drive manufacturer's tension-versus-deflection table for the actual section and span. Not by thumb; a drive set by feel is commonly well off the table in the tight direction.
• The overhang distance, measured. A sheave changed at some point, or a bushing fitted the other way round, moves it.
• Soft foot and base condition, dial indicator, machine locked out. A base pulled flat by its bolts is applying a force nobody wrote down.
• Sheave groove wear, because a worn groove lets the belt ride low, changing the effective pitch diameter and so the tangential force for the same torque.

Step 6: read the evidence on the bearing itself

If you have the failed bearing, the loaded arc on the ring that does not rotate is your observable. Under a pure radial load with normal internal clearance, that arc runs roughly 120 degrees. At zero clearance it opens toward 180 degrees. Under preload it goes wider still. A markedly wider arc says clearance was lost, from a tight fit, thermal growth, or both. A short, smeared, uneven path rather than a clean matte band is the signature of skidding, which is the underloaded case.

A clean band gives direction and arc, never pounds; that is what column one is for. Mapping a damage pattern to a cause is a sibling article's subject.

Pulling and cleaning that bearing brings its own hazards. A puller under load is stored energy: stay out of line with the screw, use a shield or a restraining strap, and never heat a puller that is under tension. Solvent cleaning is an inhalation and skin route, so use the glove class named on the safety data sheet and ventilate rather than relying on a shop fan. If you cut a ring off, that is a guarded wheel, a face shield and respiratory protection for the dust, not a bare grinder on a bench.

Worked example: the fan whose far bearing kept failing

An air-handler fan shaft on two pillow blocks. Bearing span 24 in. Sheave overhung 5 in beyond the drive-end bearing. Fan wheel weighs 120 lb and sits 10 in from the drive-end bearing, between the bearings. The drive delivers 7.5 hp at 900 rpm through a 12 in pitch diameter sheave. The motor sits beside the fan, so belt pull is horizontal and weight is vertical. The far bearing has been replaced three times in two years; the drive-end bearing has never been touched.

Column one.

Torque is 5252 x hp / rpm in lb-ft, so 5252 x 7.5 / 900 gives 43.8 lb-ft, which is 525 lb-in. Divide by the 6 in pitch radius and the tangential force is 87.5 lbf. The drive manufacturer's shaft-load table for this arc of contact gives about 1.7 times tangential, so belt pull on the shaft is about 150 lbf.

Belt pull is overhung 5 in beyond a 24 in span:

• Drive-end reaction: 150 x (24 + 5) / 24 = 181 lbf
• Far-end reaction: 150 x 5 / 24 = 31 lbf, in the opposite direction
• Check: 181 minus 31 is 150, the applied load. Correct for an overhung case.

Wheel weight, 120 lb between the bearings at 10 in from the drive end:

• Drive-end share: 120 x 14 / 24 = 70 lbf
• Far-end share: 120 x 10 / 24 = 50 lbf
• Check: 70 plus 50 is 120. Correct for a load between the bearings.

Vector sums, belt horizontal and weight vertical:

• Drive-end bearing: square root of (181 squared plus 70 squared) = 194 lbf
• Far bearing: square root of (31 squared plus 50 squared) = 59 lbf

The bounds. The catalogue rating for this insert size reads about 6,000 lbf, and one percent of that is 60 lbf. Treat both as illustrative and pull the real minimum-load line, because this is the trap named above: the requirement rises with speed, and at 900 rpm on a pillow-block insert it usually sits well below one percent of C. A computed 59 against an assumed 60 is two numbers inside each other's uncertainty, not a finding. Skidding is also mainly a high-speed behaviour, so treat a low computed load here as a reason to pull the real curve and read the raceway, not as a diagnosis.

What column one points at. The drive-end bearing is at 194 lbf against a 6,000 lbf rating, 3 percent of C and entirely comfortable. The far bearing is at 59 lbf, in the neighbourhood of any minimum-load line you are likely to read for this insert at this speed. Column one cannot settle it from there. What it can do is name which bearing to suspect and which direction to suspect it in, and the direction is under-load, not over-load.

Column two settled it. The last failed far bearing showed a short smeared path on the outer ring with no clean matte band, no spalling, and discoloured grease. The drive-end bearing, pulled for inspection, showed a clean 130-degree band on the outer race centred on the belt-pull direction, which is a bearing being loaded the way it was meant to be.

The trap. The obvious field fix is to tighten the belts, because more tension raises the far bearing's load. Run it: double the belt pull to 300 lbf and the far bearing goes to the square root of (62 squared plus 50 squared), which is 80 lbf, comfortably over the minimum. But the drive-end bearing goes to the square root of (362 squared plus 70 squared), which is 369 lbf. Using the ball bearing load-life exponent of 3, the life ratio is (194 / 369) cubed, or 0.15. You have bought the far bearing its minimum load by throwing away 85 percent of the drive-end bearing's calculated life. That is not a fix, it is a transfer.

What actually addresses it is geometric or specification-level: move the wheel toward the far bearing so more of its weight lands there, or take the case to the equipment manufacturer for a bearing with a lower minimum load requirement at this speed. Both are design decisions, not field adjustments, and that is the honest answer to give the customer.

Reconciling, and what a disagreement means

Put the columns side by side. Three outcomes:

They agree. The computed load is between the bounds and the load zone is a clean band of about the expected arc, centred where your resultant points. The bearing is loaded as intended, and whatever failed it is not load, so go looking at lubrication, contamination or fits.

Column two shows more load than column one predicts - a wider arc, a heavier band, or a band pointing somewhere your resultant does not. There is a force in the path you have not named. Usual candidates, in the order they turn up: a drive tensioned above the table, thermal growth reacting against a locating shoulder or a jammed floating bearing, a rigid duct or pipe that grew.

Column two shows less load than column one predicts - the smeared, non-uniform path. Either your computed number is too high, usually because the drive is looser than you assumed, or the bearing spends real time at a duty you did not model: on a variable-speed machine, normally turndown rather than nameplate.

Run only column one and you get confident arithmetic about a machine that is not built the way you assumed. Run only column two and you get the example: three replacements of a part that was never defective, at a shop that never checked whether the load reaching it was above the line the catalogue draws.

References

• ISO 281 for the rating life relationship and the load-life exponents used in the trade-off above; bearing manufacturer catalogue for the basic dynamic rating C and the minimum load requirement, which is speed and lubricant dependent
• 29 CFR 1910.147 (hazardous energy and stored energy in a tensioned drive), 29 CFR 1910.219 and 29 CFR 1926.300(b) (power-transmission guarding, general industry and construction), 29 CFR 1910.333(b)(2) (electrical work), NFPA 70E-2021, 120.5 (live-dead-live)
• 29 CFR 1910.134 (respiratory protection program), for the solvent and dust routes named in the pulling step
• Drive manufacturer documentation for tension-versus-deflection values and shaft-load factors, both of which vary with belt section and arc of contact
• See related: What a Bearing Load Actually Is; Why a Bearing Fails in a Way That Names Its Own Cause