How Load and Life Relate on a Rotating Component
Why this matters
"Reduce the load and it will last longer" is true and almost useless as stated, because the exchange rate between load and life is different for every component on the same shaft, and because the load a component actually feels is often not the load you just reduced. A shop that cuts shaft power by a fifth and expects a bearing to last twice as long has quietly assumed two things: that the bearing's load runs through the shaft torque, and that the relationship it half-remembers applies to the part in its hand. On a belt drive, the first assumption is usually wrong. This article is about the exchange rate, and about the case where a real load reduction bought almost nothing.
The safety state this work happens in
Everything below happens on a stopped machine. A belt-driven fan wheel coasts for a long time after power is removed and carries enough stored rotational energy to take a hand into the drive, which is why isolation for this work sits under 29 CFR 1910.147: lock and tag the energy-isolating device, then confirm the wheel has come to a complete stop by observation before any guard comes off. Belt tension is measured and set with the drive stopped and locked, never by pressing on a moving belt. The guard goes back on before anything is energized, and the direction check afterwards is made from outside the plane of rotation.
The machine and the signal
A belt-driven exhaust fan in a light industrial space had eaten three sets of fan-shaft bearings in about twenty months. The work was not sloppy: each replacement used the specified bearing, the shaft measured within tolerance, the housings were clean, and the lubricant was the specified grade applied at the specified interval. Nothing in the record pointed anywhere except "this machine is hard on bearings."
Somebody had heard that bearing life goes as the cube of load, and proposed the obvious fix: slow the fan down and cut the load. A sheave change dropped the fan speed to about 0.93 of its previous speed, which by the fan relationships cuts shaft power to about 0.80 of what it was, roughly a fifth less. That article's derivation and its conditions are covered elsewhere in this library and are not repeated here.
The fourth bearing set failed on roughly the same schedule as the first three.
Why a fifth less power bought almost nothing
Three separate errors were stacked in that expectation, and they are worth separating because each one is common on its own.
The power reduction was not a torque reduction of the same size. Shaft power goes with the cube of speed on a fan, but torque is power divided by speed, so torque goes with the square. At 0.93 speed the torque ratio is 0.93 squared, about 0.865 - roughly 13 percent less torque against 20 percent less power. Anyone carrying the 20 percent figure forward into a torque-driven calculation is already overstating the gain by half.
The bearing's load does not run through torque anyway. On a belt drive, the fan-shaft bearing load is dominated by belt tension acting on the sheave, plus the weight of the wheel and sheave, plus whatever aerodynamic thrust the wheel develops. Transmitted torque only enters through the portion of tension the drive needs to carry it. The general method for computing the load where the bearing actually is, from the actual load path, is owned by a sibling article in this library and is worth reading before any life estimate; see the references.
The tension was never re-set. Belt tension is a value someone installs, not a value the drive negotiates. Reducing the torque requirement by 13 percent lowers the tension the drive needs; it does nothing whatsoever to the tension already in the belts. Nobody re-tensioned. The bearing load path was therefore unchanged, and the only real gain from the whole exercise was the revolution count: basic bearing rating life is expressed in millions of revolutions, so converting it to hours divides by speed, and running 7 percent slower yields about 1.075 times the hours at the same load. Seven and a half percent more life, against an expectation of double.
What was actually setting the life
The tension was being set by feel at every replacement, and by feel it was set high, because a slipping belt is a complaint and a tight belt is not. When the drive was finally checked against the manufacturer's force-deflection specification on a stopped, locked-out drive, the installed tension was well above what the drive required even at the original speed.
Setting it to the drive's specified value dropped the calculated dynamic equivalent load at the fan-shaft bearing to about 0.70 of its prior value. For a ball bearing, ISO 281 basic rating life scales as the load ratio raised to the third power, so the life multiplier is 1 divided by 0.70, cubed, which is about 2.9 times. Combined with the 1.075 from the slower speed, the total is about 3.1 times the original rating life.
Two hedges belong in the same breath as that number. First, the exponent is 3 for ball bearings and 10/3 for roller bearings, so on a roller-bearing machine the same 0.70 load ratio gives about 3.3 times rather than 2.9 - do not carry a ball-bearing exponent onto a roller arrangement. Second, ISO 281 basic rating life is a statistical 90-percent-reliability figure derived at constant load and constant speed under catalog conditions, and it does not include lubricant film quality or contamination; the modified rating life applies a life-modification factor that can move the answer further than the load change did. "About three times the rating life" is a defensible sense of scale for a decision. It is not a service date.
How they confirmed it, and what they wrote down
The fifth bearing set ran past the point at which every earlier set had failed, which is the only confirmation that means anything here, and it took longer than the diagnosis did. The change that made it durable was not the sheave and not the bearing: it was writing the specified deflection force and deflection distance on a tag at the drive, so the next person to tension it has a number instead of a thumb. The sheave change stayed in place because the slower speed still met the required airflow, but on this machine it was worth about seven percent, not the doubling it was sold as.
The exchange rates, side by side
Each of these is stated with the condition it was derived under, because none of them survives being carried onto a different component.
| Component | Life responds to | Relationship | Derived under |
|---|---|---|---|
| Rolling-element bearing | Dynamic equivalent load at that bearing | Life in revolutions goes as the load ratio to the third power for ball, 10/3 for roller | ISO 281 basic rating life, 90 percent reliability, constant load and speed, catalog conditions, no lubrication or contamination factor |
| Rolling-element bearing, in hours | Speed, separately from load | Hours scale inversely with speed at constant load | Same, converting millions of revolutions to hours |
| V-belt | Tension and bending stress per pass around each sheave | Steep, but the exponent is specific to belt section and construction | Manufacturer drive-rating tables; route to them rather than assuming a number |
| Belt, cycle count | Belt speed and belt length | Bends per hour go with belt speed divided by belt length, times the number of sheaves | Geometry only, holds generally |
| Motor winding insulation | Sustained winding temperature | Aging rate roughly doubles per 10 C sustained increase | Thermal-aging rule of thumb, valid within the insulation class's normal operating range, not an extrapolation |
| Sheave and shaft fretting | Fit condition and reversing load | Not a life exponent at all; it is a threshold behaviour | Observation, not a rating standard |
The last row matters more than it looks. Not every failure mode has an exponent. A fretted bore, a loose bushing or a wiped keyway does not get 2.9 times better because you dropped the load 30 percent; it either has enough clamp and enough fit to stop moving, or it does not.
What would have changed the conclusion
If this had been a direct-coupled machine rather than a belt drive, the original reasoning would have been much closer to right, because on a direct-coupled pump or blower the bearing load is dominated by hydraulic or aerodynamic reaction plus rotor weight, and reducing the duty genuinely reduces that reaction. The failure here was applying a load-life relationship across a load path it does not travel, which is the same class of error as applying any coefficient outside the geometry it was measured on.
If the bearings had been failing with contamination signatures rather than fatigue signatures, the load-life exponent would have been the wrong tool entirely: a contaminated bearing is on the modified-rating-life side of the calculation, where the life factor swamps the load term. Reading which failure mode you are actually looking at is covered by the bearing-failure articles in this library, and it belongs before any life arithmetic.
How to verify you got this right
- Name the load path before you name the exponent. Write down what actually pushes on the bearing: belt tension, rotor weight, hydraulic reaction, thrust, coupling reaction. If your proposed change does not appear in that list, your predicted life gain is zero.
- Check whether your change requires a second action to take effect. A torque reduction only reaches the bearing if somebody re-tensions the belts to the new requirement, on a stopped and locked-out drive, against a specified deflection force.
- State which exponent you used and why. Ball or roller changes the answer by more than ten percent on the same load ratio.
- Convert to hours deliberately. Rating life is in revolutions. If the machine also slowed down, that is a separate and additive gain, and if it sped up, it is a separate loss that people routinely forget to subtract.
- Sanity-check against the failure signature. If the parts coming out do not look like load fatigue, the load-life calculation is answering a question nobody asked.
References
- ISO 281, rolling bearing dynamic load ratings and rating life, for the basic rating life relationship, its exponents and the conditions it is defined under
- 29 CFR 1910.147, control of hazardous energy, for isolation and stored rotational energy on fan and coupled-drive work
- Manufacturer drive-rating and force-deflection tension data for V-belt life and installed tension specifications
- See related: What a Bearing Load Actually Is; How a Belt Drive Transmits Torque; Why a Bearing Fails in a Way That Names Its Own Cause; How a Fan Law Changes What the Drive Has to Deliver