What Belt Tension Actually Controls

Why this matters

Tension is the only thing on a belt drive that every tech adjusts and almost nobody sizes. It gets set by feel, by a thumb press, by "another half turn on the adjuster," and then it gets blamed in both directions: loose when something slips, tight when a bearing goes. Both blames are usually wrong, because tension is not a power setting. It is a reserve, and a reserve is sized against the worst instant the drive will ever see, not against the load it spends its life at. A drive tensioned for the average is a drive that is correct on the bench and short exactly when the torque spikes, which is the one moment nobody is standing there with a gauge.

Isolate before you touch the adjuster

Lock out and tag the motor at its disconnect before any hand goes inside the guard. A fan wheel or flywheel holds rotational energy after the power is gone, which makes this stored mechanical energy under 29 CFR 1910.147: confirm zero rotation by eye, and where the wheel sits in moving air, close and secure the damper or physically block the wheel, because a windmilling wheel drives the sheave backwards through the belt. Many adjusters are spring loaded; release a spring tensioner only with the tool the design provides and keep your hands and face out of the release path, because a slipped pry bar sends the arm through its full travel. If reaching the disconnect means opening an electrical enclosure, that is electrical work under 29 CFR 1910.333(b)(2), and prove the meter live-dead-live per NFPA 70E-2021, 120.5 before trusting a zero. 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.

Five things tension does not control

Most of the confusion in the field is people expecting tension to do a job that belongs to something else.

• It does not set transmitted torque. Torque is the tight-side tension minus the slack-side tension, times the effective radius. Add the same amount to both strands and the difference is unchanged, so the drive transmits exactly what it did before while every bearing carries more. The derivation lives in the companion article on how a belt drive transmits torque.
• It does not set speed. Speed comes from the ratio of pitch diameters. Tension only affects speed indirectly, by determining whether the belt is sliding, and a belt that is sliding has already failed at its job.
• It does not correct misalignment. Pulling a misaligned belt tighter increases the side load that is already wearing one flank and drives the belt harder into the flange it is climbing.
• It does not restore a worn groove. The V wedge comes from contact on both flanks, and a belt sitting on the groove floor has no wedge left to amplify. Tension cannot manufacture flank contact that the geometry no longer provides.
• It does not quiet a belt permanently. A squeal that goes away with a quarter turn usually comes back, because the squeal was a symptom of the drive being at its friction limit and you moved it just inside the limit rather than fixing what put it there.

What it does control: the floor under the slack side

Here is the whole mechanism in one line. Friction lets the two strands hold different tensions, and the ratio of tight to slack is bounded. Installed tension does not change that bound; it sets the slack-side value that the bound is applied to.

Torque demand fixes the difference between the strands. The friction limit fixes the maximum ratio between them. Between those two constraints, a low slack side forces a high ratio for the same difference, and a high slack side allows a low one. So the question tension answers is: how much room is there between the ratio the drive is currently running at and the ratio at which the belt lets go.

That is why tension is a reserve. You are buying distance from the friction limit, and you pay for it in bearing load and belt fatigue on every rotation for the life of the drive.

The number to size against is the peak, not the running load

The demand that matters is the highest instantaneous torque the drive will see, and there are three ways it arrives:

• Starting torque. A loaded machine breaking away from rest demands more than it demands running, and the belt is cold, possibly damp, and starting from static contact.
• Cyclic torque. A single-acting reciprocating compressor or a piston pump delivers its load in a pulse tied to the compression stroke, so the instantaneous peak is a multiple of the mean every single revolution.
• Process upsets. A blocked screen, an ingested rag, a valve slammed shut.

Belt makers do not ask you to compute the pulse. They publish a service factor by driven-machine class and daily operating hours, and you multiply nameplate power by it before selecting and tensioning the drive. That factor is the peak, packaged. Tension to the nameplate load rather than the service-factored load and you have thrown away the entire reserve the selection was built around.

Tension is a setting that decays, not a property the drive keeps

Three separate mechanisms pull installed tension down, and they run on different clocks.

Seating. A new belt beds into the groove flanks over the first hours of running, which shortens the effective center-to-belt path and drops tension. Most of it happens fast.

Construction set. The belt's tensile members and rubber body take a small permanent elongation under load. This is not wear, it is the belt arriving at its working length.

Thermal and load cycling. Rubber relaxes at temperature, and each load cycle works the belt a little further into its set.

Together these are why belt makers call for a re-check after a run-in period, commonly quoted in the range of the first 24 to 48 hours of operation, and why the number your maker publishes for the section is the one to use rather than a remembered figure. Skip the re-check and the drive spends the rest of its life at whatever tension the run-in left it with, which is always lower than what you set.

Worked example: correct at spec, slipping on start

A drive on a reciprocating machine. Motor nameplate 5.0 hp at 1750 rpm. Driver pitch diameter 4.0 in, driven pitch diameter 16.0 in, center distance 12.0 in, so the driven shaft turns 1750 x 4.0 / 16.0 = 437.5 rpm.

What the drive has to pull. Motor torque at nameplate power is 63,025 x 5.0 / 1750 = 180 lb-in in inch-pound units. The driver's effective radius is 2.0 in, so the net pull is 180 / 2.0 = 90 lb at the running load. The drive was selected with an illustrative service factor of 1.4 for this machine class and duty, read off the maker's table, so the peak the belt must actually hold is 90 x 1.4 = 126 lb.

How it was tensioned. The installer set it to a tension ratio of 5, a common working design figure, against the 90 lb running pull: slack side 90 / (5 - 1) = 22.5 lb, tight side 112.5 lb. On a running check it looked perfect and it was, for running.

Where the reserve went. Total loop tension is roughly fixed by the belt's length and stiffness, so when the torque peak arrives the slack side stays near 22.5 lb and the tight side rises to cover the demand: 22.5 + 126 = 148.5 lb, a ratio of 148.5 / 22.5 = 6.6.

What the geometry allows. Wrap on the small sheave is 180 - 2 x arcsin((16.0 - 4.0) / (2 x 12.0)) = 180 - 2 x 30 = 120 degrees, or 2.094 radians. Taking an illustrative dry rubber-on-cast-iron coefficient near 0.3 for clean flat contact at moderate speed, wedged through a 36 degree groove, the effective coefficient is 0.3 / sin(18 degrees) = 0.97. The capstan limit is e raised to (0.97 x 2.094), about 7.6. So the drive runs at 5.0 against 7.6, a comfortable 66 percent of the limit, and peaks at 6.6 against 7.6, which is 87 percent of it. Add a cold belt with overnight condensation on the flanks, where the coefficient is below the clean-dry figure, and the peak crosses. It slips on start, settles once it is turning, and reads perfect on every running check anyone makes.

Tensioning it correctly. Set the ratio of 5 against the 126 lb peak instead: slack 126 / 4 = 31.5 lb, tight 157.5 lb. Now the peak sits at 66 percent of the limit like the running load does.

What that costs. At 120 degrees of wrap the two spans sit 60 degrees apart, so the shaft load is the vector sum, not the arithmetic sum. Before: the square root of (112.5 squared + 22.5 squared + 2 x 112.5 x 22.5 x cos 60 degrees) = 125 lb, against an arithmetic sum of 135. After: 175 lb against a sum of 189. On this geometry the vector result is about 7 percent below the sum, so unlike a high-wrap drive where the two agree within a percent, here you have to resolve them. The shaft load rose 40 percent, exactly the service factor, and what a 40 percent load rise does to bearing life is the subject of its own article.

The fix that costs less. Slide the motor out to a 20.0 in center distance and wrap becomes 180 - 2 x arcsin(12.0 / 40.0) = 145 degrees, or 2.533 radians, lifting the limit to about 11.7. The original 6.6 peak ratio now sits at 57 percent of the limit with no tension increase and no extra bearing load at all. The cost is a longer belt and enough slide travel to fit it, which is a materials problem rather than a life-of-the-machine problem.

How to verify you got this right

Three checks, all with the drive locked out and the wheel restrained unless stated otherwise.

• Confirm which load you tensioned for. Find the service factor used at selection, or read it from the maker's table for the driven machine class and daily hours. If the tension came from nameplate power, it is low by that factor.
• Compute the wrap angle rather than eyeballing it. Measure center distance and both pitch diameters with a tape and calipers. Below about 120 degrees on the small sheave there is very little friction margin, and the fix is geometry, not tension.
• Re-check after run-in. Come back inside the maker's stated window and measure again. A drive that has lost tension since commissioning has not failed; it has finished settling, and this is the reading that becomes its real setting.

One reading you can take running: tach the driven shaft from outside a closed guard, standing out of the plane of rotation because a failed belt or sheave fragment leaves along that plane, with hearing protection under a program per 29 CFR 1910.95 if the space is at or above the 85 dBA action level. Compare against the speed the pitch diameters predict. A steady gap beyond about 1 percent is a drive already living past its friction limit, and adding tension is treating the symptom until you have established what took the capacity away.

References

• 29 CFR 1910.147 for mechanical isolation and stored rotational energy; 29 CFR 1910.333(b)(2) for work at an electrical enclosure and NFPA 70E-2021, 120.5 for live-dead-live proving
• 29 CFR 1910.219 (general industry) and 29 CFR 1926.300(b) (construction) for guarding of belts, pulleys and sheaves; 29 CFR 1910.95 for the occupational noise program and its 85 dBA action level
• Belt and sheave manufacturer engineering data for service factors by driven-machine class and duty hours, groove angles by section and pitch diameter, and run-in re-tension intervals
• See related: How a Belt Drive Transmits Torque; How to Tension a Belt Without Guessing; What Over-Tensioning Does to the Bearings Either Side