1. The Field Problem: "The Torque Was Right, So Why Did It Fail?"

A coal handling conveyor in a power plant used a sprag-type backstop on the gearbox input shaft, rated 1,500 N·m holding torque against a calculated requirement of 1,100 N·m. The margin looked comfortable — a factor of 1.36. Yet at 17 months the unit was replaced: visible sprag flattening, circumferential scoring on the inner race, and intermittent slip on the next emergency stop.

Nothing was wrong with the rating. What was missing was the duty cycle. The conveyor tripped on emergency stop or belt-slip interlock four to five times per shift, always under full load, always with the belt coasting to a stop and then recoiling into the locked sprags. The bearing was never loaded above its nameplate value — but it was loaded above its nameplate value 3,000 times a year, which is a fatigue problem, not a torque problem.

2. Mechanism: Four Things a Static Rating Does Not Describe

2.1 Transient lock torque

When a loaded inclined belt is stopped by the backstop alone, the stored elastic energy of the belt carcass is released into the locking elements. The peak torque at the moment of sprag engagement is typically 1.5 to 2.5 times the steady reverse holding torque, and it lasts milliseconds. A statically correct selection says nothing about this peak.

2.2 Contact fatigue, not yield

Sprags and races make line contact at very high Hertzian stress. Failure is not a bending break; it is subsurface-initiated pitting that grows with the number of stress cycles. Life in cycles falls steeply with torque — for sprag contact fatigue, life exponent p is roughly 4 to 5 in the relation N ∝ (T_rated/T)^p. A 90 percent torque increase does not cost 10 percent of life; it costs more than 90 percent of it.

2.3 Repeated engagement and micro-slip

Each lock event requires the sprags to tilt from freewheel to engaged position. Under repeated emergency stops with residual reverse speed, sprags engage at speed, scuff the race, and lose the oil film exactly where the load is highest. Zero-speed holding — where a hydrodynamic film cannot form — turns into boundary lubrication, and boundary lubrication under 1.5 GPa contact stress is where pitting begins.

2.4 False brinelling during idle holding

An emergency-stopped belt can sit locked for hours while the structure vibrates. Sprags indent the race at fixed positions, and the indentation becomes the origin of the next pit. Static rating tables never mention this, because it depends on how long the machine stays locked and how much vibration the frame transmits.

3. Worked Example: Life Calculation Instead of a Torque Check

Same 1,500 N·m-rated sprag unit. Measured reverse speed at engagement 0.12 m/s, elastic recoil gives a measured transient peak of 1.9 × steady torque.

Apply a misalignment or lubrication derating of 1.5 and predicted life drops to under two years — which matches the observed 17-month failure. The unit was correctly rated and wrongly applied. This is why a life calculation, not a torque comparison, belongs in the selection file.

4. Engineering Solutions

• Reduce the transient, not just increase the rating. A controlled braking ramp on the VFD, or a hydraulic brake that decelerates before the backstop takes load, cuts the peak factor from 2.0 to about 1.3 — a far cheaper intervention than the next frame size.
• Match the type to the engagement count. Roller-type backstops distribute load across several rollers simultaneously, so individual contact stress and pit growth rate are lower; sprag units win on torque density and engagement speed but are less forgiving under thousands of hard lock events.
• Protect the lubrication regime. Oil-lubricated gearbox-mounted units carry fresh oil to the contact; sealed grease units rely on the initial fill and are far more sensitive to repeated zero-speed loading.
• Prove it by test, not by catalogue. Ask for a dynamic cycling test: 10,000 lock/unlock cycles at rated torque with torque trace recorded, plus pre/post contact condition report.

5. Series Selection and Verification

Documentation to request before purchase: rated holding torque and the peak/shock basis behind it, L10 engagement cycle life at rated and at 1.5× torque, allowed engagement speed, lubrication type and replenishment interval, race and sprag hardness with case depth, and the dynamic cycling test report. If a supplier can only quote one number, the specification is not yet complete.

6. FAQ

Q: If the unit is rated 1,500 N·m and the load is 1,100 N·m, is that not a 36 percent margin?
A: A 36 percent static margin, yes. Under repeated emergency stops the governing number is transient peak (2,090 N·m) applied thousands of times, which erodes life far faster than the static ratio suggests.

Q: Does a bigger backstop solve it?
A: Partly. Upsizing raises rated torque but also raises drag torque and heat in freewheeling duty. Reducing the transient through controlled deceleration usually gives more life per dollar.

Q: How many emergency stops per year is "too many"?
A: There is no universal number — compare actual engagement count against the L10 cycle life at your transient torque. Above roughly 10 percent of available cycles per year, plan for condition monitoring.

Q: What early warning signs should maintenance watch?
A: Any reverse creep on the belt after stop, a change in the sound at lock-up, rising oil temperature in the backstop housing, and metal content in the oil are the four earliest indicators.