What a Keyway Does and How It Fails

Why this matters

A wallowed keyway is one of the few faults where the obvious fix is reliably wrong. The joint is loose, the key is battered, so the shop fits a new key, maybe a longer one, and the machine comes back on the same clock. The reason is that the key almost never failed from being too small. In most service failures the key is carrying a small fraction of what it could, and the damage comes from a duty the joint class cannot handle at any key size. This article is an inspection record, filled in for one real joint, and the record is what makes that visible.

Before you pull the hub

Isolate and lock out under 29 CFR 1910.147, confirm zero rotation rather than assuming the switch did it, and restrain a spring-loaded motor base before belts come off, since the tensioned base is stored energy under the same standard. A puller under load is stored energy as well: never stand or put your face in line with the puller screw, use a shield or a restraint, and relieve the load before you release anything.

If you heat a hub to get it off or to shrink it on, use a controlled heater within the manufacturer's temperature cap rather than an open torch, with gloves rated for that contact temperature. Heating a plated or painted component releases fume you must not breathe; cadmium plating on older hardware specifically generates cadmium oxide fume, which requires local exhaust ventilation and respiratory protection under a written program per 29 CFR 1910.134 alongside 29 CFR 1910.1027. If you go the other way and cold-shrink a shaft or a key with dry ice or liquid nitrogen, that is a cryogenic contact burn hazard needing insulated gloves and a face shield, and it is also an oxygen-displacement hazard: the boil-off displaces breathable air, so this is done with ventilation and never in a pit, a vault or any other confined space.

What a key actually does

A key sits half in a slot cut in the shaft and half in a slot cut in the hub, and it transmits torque by bearing pressure on its side faces. Not by shearing across a single plane, although that is a second failure route it is also checked against. The distinction matters because the two are checked with different areas and the bearing check is usually the one that governs.

Two forces to get straight.

The tangential force on the key is the torque divided by the shaft radius:

F = T / (d / 2)

The shear area is the key width times its length, because a shear failure would part the key along the shaft-hub interface across its full width.

The bearing area is half the key height times its length, because only the half of the key standing proud of the shaft is pressing on the hub slot.

For a square key those two areas differ by a factor of two, so bearing stress is always exactly twice the shear stress on the same joint. The reason a square key still sizes out sensibly is that the allowable compressive stress of a steel is roughly twice its allowable shear stress, so the two checks land in about the same place. That relationship is derived for the standard square key sitting half in each member; a flat key, a gib-head key or a Woodruff key changes the proportion and the check has to be redone rather than carried over.

And one thing a key does not always do: on an interference-fitted hub, the fit carries the torque by friction and the key is anti-rotation insurance. The same key in a clearance-fitted hub is carrying every bit of the torque by itself. The same part is doing two different jobs, and only one of them is what the arithmetic below describes.

The inspection record

The record filled in

A belt-driven blower, sheave on a motor shaft, 5 hp at 1750 rpm, hub loose and clattering on start.

• Shaft diameter at the seat: 1.500 in, measured
• Standard key for that range: 3/8 in square, and that is what is fitted
• Key length / hub length: 2.25 in / 2.25 in
• Key material: plain low-carbon key stock; file test shows it is harder than the shaft
• Duty: the blower serves a modulating damper and cycles roughly forty times a shift, so stop-start with reversing torque on deceleration
• Torque at rated: 5252 x 5 / 1750 = 15.0 lb-ft = 180 lb-in
• Tangential force: 180 / 0.75 = 240 lbf
• Shear area: 0.375 x 2.25 = 0.844 in squared. Shear stress = 240 / 0.844 = 284 psi
• Bearing area: 0.1875 x 2.25 = 0.422 in squared. Bearing stress = 240 / 0.422 = 569 psi
• Hub-to-shaft fit: hub bore 1.503 in against a 1.500 in shaft, so 0.003 in of diametral clearance. This is a clearance fit; the key is carrying all of it
• Key side clearance: 0.004 in on the loaded side with a feeler
• Wallow in the shaft keyway: 0.012 in of material displaced along the loaded face
• Wallow in the hub keyway: negligible
• Key itself: straight, unbent, no measurable deformation

Reading it: the arithmetic that proves the wrong thing

Plain low-carbon key stock has a yield strength in the tens of thousands of psi. At 569 psi of bearing stress this key is running roughly sixty times below its limit. Run the standard check and it passes so comfortably that the check is meaningless.

That is the trap, and it is worth naming because it is where good technicians talk themselves out of the real finding. The calculation is correct and it is answering a question this joint does not pose. It is a steady-torque calculation, and the duty field on the record says the torque is not steady.

Here is what the record actually says happened. Every time the drive decelerates, torque reverses. With 0.004 in of side clearance and 0.003 in of fit clearance, the key does not simply change which face it presses on. It travels across the gap and arrives. The peak contact stress in that arrival is set by the closing speed and the local stiffness of the two members, not by the rated torque, and it can be many times the steady figure. Repeat it forty times a shift and the loaded faces of the slot get worked, which opens the clearance, which lengthens the travel, which raises the next impact. On top of that, the small relative motion inside the clearance is a fretting mechanism that removes material continuously without any gross yielding at all.

Neither of those appears anywhere in a bearing-stress calculation, and no key size removes them. Doubling the key width halves the stress in a check that was already passing by sixty times and does nothing about the clearance the key crosses.

Why the damage is in the shaft

Read the last three lines of the record together: the shaft keyway is wallowed 0.012 in, the hub keyway is not, and the key is undeformed.

A key is supposed to be the sacrificial member. Key stock is specified softer than the shaft and the hub so that when a joint is abused, the cheap, three-minute part deforms first and tells you. On this joint the file test says the key is harder than the shaft, so the shaft has been acting as the sacrificial member instead. The shaft is the part that requires the machine to come apart, the rotor to be balanced again and the bearings to be disturbed.

This is a specification failure that costs nothing to avoid. When a key is made up from bar stock at the bench, check that what came out of the rack is key stock and not a piece of hardened flat, and keep hardened material out of a joint whose whole design assumes the key gives up first.

What the key length is actually doing

The joint above has a key length of 2.25 in on a 1.500 in shaft, which is 1.5 times the shaft diameter. That is not a coincidence. For a square key whose width is about a quarter of the shaft diameter, which is what the standard sizes give, shear and bearing come out about equally critical at a length near 1.5 times the diameter, and that is also about the hub length most components carry.

Going longer buys much less than it looks like it should, because the load does not distribute evenly along the key. The shaft twists slightly along the length of the joint, so the end where the torque enters takes disproportionately more of it, and material at the far end is barely engaged. A key twice as long is not twice as strong.

The same caution applies to fitting two keys at 180 degrees. Load sharing between them is imperfect, and design practice credits a second key with well under a doubling. If a joint genuinely needs two keys, it has outgrown keys, which is the next section.

Changing the joint class

The durable answers all change what carries the torque rather than resizing what already was not the limit.

Two things not to do. Do not weld up a keyway and re-cut it without settling the metallurgy first: welding puts a heat-affected zone and residual tensile stress at the surface, which is exactly where a shaft's stress lives and where fatigue starts. And do not simply drive a slightly oversize key into a wallowed slot to take up the gap, because the slot is no longer a rectangle: the loaded face is worked and the corners are rolled, so the oversize key bears on the rolled edges rather than on a flat, which concentrates the very load you were trying to spread.

References

• ANSI/ASME B17.1 for standard key and keyseat proportions by shaft diameter range
• Trade-standard mechanical design references for the shear and bearing checks on a parallel key, and the length convention at which the two come out equally critical
• Component manufacturer documentation for taper bushing and keyless locking assembly clamping capacity, which is set by contact pressure rather than by any key calculation
• 29 CFR 1910.147 (hazardous energy, stored energy in a tensioned drive and a loaded puller); 29 CFR 1910.134 and 29 CFR 1910.1027 (respiratory protection program; cadmium fume from heated plating)
• See related: How a Shaft Actually Carries Torque; Why Shaft Deflection Matters More Than Shaft Strength