What Over-Tensioning Does to the Bearings Either Side

Why this matters

Belt pull does not stop at the sheave. It runs down the shaft into the bearings either side of it, and because the sheave hangs outboard of them rather than sitting between them, the bearing nearest the sheave carries more than the belt is pulling. That is a lever, and levers do not care how carefully you set the tension after the fact. The reason this matters more than most over-tightening errors is that nothing announces it: the drive runs, the machine makes its numbers, the belt looks fine, and eighteen months later a bearing that should have run for years is replaced and read as a bad bearing. The cost was booked on the day someone gave the adjuster one more turn.

Isolate before you measure any of this

Lock out and tag the motor at its disconnect. A wheel or flywheel holds rotational energy after the power is gone, which is stored mechanical energy under 29 CFR 1910.147, so verify zero rotation by eye and block the wheel or close and secure the damper before reaching inside the guard, because a windmilling wheel back-drives the sheave. Release any spring-loaded tensioner with the tool the design provides and stay out of the arm's swing path. If a bearing temperature reading is part of your check, take it with a non-contact infrared thermometer aimed through the guard opening with the guard closed; a hand on a running bearing housing is both a burn and a caught-in exposure. Opening an electrical enclosure to reach the disconnect is electrical work under 29 CFR 1910.333(b)(2), proved live-dead-live per NFPA 70E-2021, 120.5, and the guard goes back on before restart under 29 CFR 1910.219 in general industry or 29 CFR 1926.300(b) on a construction site.

The load does not follow the machine's duty

The first thing to get straight is that belt pull is essentially constant while the drive is assembled. A belt's total loop length is fixed, so when torque rises the tension redistributes between the two strands rather than being added to the loop: the tight side climbs by roughly what the slack side gives up. The sum of the two changes very little, and the sum is what the shaft feels.

So a fan running at 30 percent of its rated output carries the same shaft load as one at 100 percent. A machine idling overnight carries it. A drive shut down but still belted carries it. There is no light-duty condition in which an over-tensioned drive gets a rest, which is exactly why over-tension shows up as bearing life and never as a performance complaint.

From belt pull to bearing reaction: the overhang lever

Take the two bearings supporting a motor or fan shaft, and call the one nearest the sheave the near bearing. Let a be the distance from the near bearing's centerline out to the sheave's center plane, and b the distance between the two bearing centerlines. For a belt pull W applied at the sheave, simple statics gives:

near bearing reaction  = W x (1 + a / b)
far bearing reaction   = W x (a / b), acting in the opposite direction

Two things fall out of that, and both are counter-intuitive enough that they are worth saying plainly.

The near bearing always carries more than the belt pulls. Not the same, more. With the sheave hung a quarter of the bearing span outboard, it carries 125 percent of the pull. Push the sheave further out and the factor climbs without any tension changing at all.

The far bearing is loaded backwards. Its reaction opposes the pull. It is small in absolute terms, but it is pure lever moment, so it is proportionally the most sensitive thing on the shaft to where the sheave sits. It is also the reason a shaft can behave oddly at very low pull: rolling bearings have a minimum load below which the rolling elements skid rather than roll, and a lightly loaded far bearing sitting near that threshold is its own separate problem.

Both reactions act in a fixed direction, set by where the belt spans run. With a rotating inner ring and a stationary outer ring, that means the same arc of the outer raceway takes every load cycle for the life of the machine. Damage is not spread around the race; it accumulates in one sector, which is why an over-loaded belt-drive bearing so often shows a localized band of raceway distress.

Why a modest load rise is not a modest life reduction

Rolling bearing life is conventionally expressed as basic rating life, the number of revolutions 90 percent of a population of identical bearings will reach under a given constant load and speed with adequate lubrication and clean operating conditions. Under that convention, life varies with the load ratio raised to the third power for ball bearings and to the power of 10/3 for roller bearings. Every one of those conditions matters: it is a statistical life for a population, not a prediction for your bearing, and it says nothing about contamination or a lubricant that was wrong for the temperature.

What the exponent means in practice is that load errors are amplified. A 25 percent load rise leaves a ball bearing at 1 divided by 1.25 cubed, or 51 percent of its rating life. Half the life for a quarter more load. Nobody adjusts a belt with that trade in mind, and the arithmetic is doing it whether they do or not.

Worked example: two errors, one bearing

A motor with a sheave overhung a = 2.0 in beyond the drive-end bearing, bearing span b = 8.0 in, ball bearings both ends. Correct commissioning tension puts the resultant belt pull at 100 lb.

Baseline. Near bearing 100 x (1 + 2.0 / 8.0) = 125 lb. Far bearing 100 x 0.25 = 25 lb, opposed. Call these the reference loads; every ratio below is against them.

Error one: tensioned by feel, 50 percent high. Belt pull 150 lb. Near bearing 187.5 lb, far bearing 37.5 lb. Both bearings see the same 1.5 load ratio, so both land at 1 divided by 1.5 cubed, or 30 percent of their reference life. Seventy percent of the life is gone and the drive shows no symptom at all.

Error two: the sheave set for clearance instead of against the shaft shoulder. Say it ends up 1.0 in further out, so a = 3.0 in. At correct 100 lb tension the near bearing factor becomes 1 + 3.0 / 8.0 = 1.375, giving 137.5 lb against the 125 lb reference. That is only a 10 percent load rise, and cubed it is (125 / 137.5) cubed = 75 percent of reference life. A quarter of the bearing's life for one inch of shaft position, at perfectly correct tension.

The far bearing takes that same inch far harder. Its factor goes from 0.25 to 0.375, a 50 percent load rise, leaving it at 1 divided by 1.5 cubed, or 30 percent of reference life. Absolutely it is still only carrying 37.5 lb, so whether it becomes the first failure depends on how the two bearings are rated relative to each other, and on a typical machine the drive end is the larger of the two. The point is not which one dies first; it is that the far bearing's exposure is entirely a function of a dimension nobody records.

Both errors together. Belt pull 150 lb with a = 3.0 in: near bearing 1.375 x 150 = 206 lb. Against the 125 lb reference that is a load ratio of 1.65, so life is 1 divided by 1.65 cubed, or 22 percent of reference. Roughly a fifth. Neither mistake felt like a mistake at the time, and neither is visible on any reading the machine produces.

What it looks like in the field. Not a squeal and not a spike in current. It looks like a bearing replacement at a fraction of the interval the same machine used to run, with the replaced bearing showing raceway distress concentrated in one arc rather than spread around the race, and a running temperature that has been quietly higher than its neighbours for months. The second replacement fails on the same schedule as the first, because nobody measured the tension or the overhang either time.

The number that overrides all of this

Fan, blower and gearbox makers publish a maximum allowable overhung load, and it is always stated at a specific distance from the shaft shoulder or bearing face, because the moment is what the shaft and bearing arrangement were designed around. A published overhung load quoted without that distance is meaningless: the same figure applied one inch further out is a different load on the bearing and a different bending moment on the shaft. When such a number exists, it is the gate, and rating-life arithmetic is only there to tell you how much margin you are eating inside it.

Shaft bending is the other half of this and it has its own article; a belt pull large enough to matter to a bearing is also deflecting the shaft, which shows up at the seal lip long before it shows up as a broken shaft.

What changes the answer

A heavy overhung wheel changes which term dominates. On a large fan, the wheel's own weight and residual imbalance may be the sizing load and belt pull a secondary term. Recompute rather than assuming the belt is the story.

Pull direction matters on some machines. Where the drive is arranged so belt pull opposes the weight of an overhung wheel, the resultant is the vector difference and adding tension can reduce net bearing load over part of the range. This is a real effect and a bad thing to assume without checking the geometry, because getting the sign wrong doubles the error.

Roller bearings use the 10/3 exponent, which is steeper still, so the same load error costs more life than the ball-bearing arithmetic above suggests.

How to verify you got this right

• Record the overhang. With the drive locked out, measure from the bearing housing face or the shaft shoulder to the sheave center plane and write it down. It is the one dimension in this whole chain that nobody has, and it can only be measured with the machine open.
• Check the sheave is against its shoulder or at the maker's stated position, not wherever it slid to when the bushing was pulled up. Locating a sheave for belt-line convenience and then correcting the belt line by moving the other sheave is the cheaper trade.
• Compare the two bearing housing temperatures with a non-contact infrared thermometer, guard closed, at the same running condition. Equal-duty bearings that differ persistently are telling you about a load split, and the near bearing running hotter than the far one is expected, not a finding.
• Verify the tension is a measured number, not a feel. The reserve you need is set by peak torque and drive geometry, which is a separate article, and every pound above it is being paid for at the third power.

References

• 29 CFR 1910.147 for mechanical isolation and stored rotational energy; 29 CFR 1910.333(b)(2) and NFPA 70E-2021, 120.5 for enclosure work and live-dead-live proving
• 29 CFR 1910.219 (general industry) and 29 CFR 1926.300(b) (construction) for guarding of belts, pulleys and sheaves
• ISO 281 for basic rating life and the load-life exponents; bearing manufacturer engineering data for minimum load and permissible operating temperature
• Fan, blower and gearbox manufacturer data for maximum allowable overhung load at a stated distance from the shaft shoulder
• See related: What Belt Tension Actually Controls; Why Shaft Deflection Matters More Than Shaft Strength; What a Bearing Load Actually Is