Keyways Stripping Again? The Practical Side of Shaft-Hub Connection Design

Why Keys Keep Stripping On My Bench

I have lost count of the small servomotor shafts I have seen chewed up by a keyway that was simply too shallow for the job. A colleague once called me at nine in the evening because a 12 mm key on an indexing table kept shearing every three weeks. We pulled the coupling, measured the keyway depth with a vernier, and found it 0.15 mm under spec. That tiny depth error cut the loaded flank area by almost a fifth on a small key, and the crush stress quietly climbed past the yield point of the hub material. This article is the set of rules I actually use on the shop floor, not the clean table from a textbook.

Keyways are cheap, simple, and universally disliked by maintenance crews. They fail in two boring ways: the key shears in the middle, or the hub bore opens up because the flank pressure is too high. Shaft twisting is rare if the shaft is sized for the torque, so in practice you are fighting crush stress on the key flanks, not shear. Keep that thought in your head through all the numbers below.

Which Connection Actually Fits Your Build

Before you pick anything, decide whether you even need a key. For light duty, a precision interference fit (H7/p6 or similar) alone can carry the torque without a key, and it removes the stress concentration entirely. I use pressed fits on small encoder pulleys and timing sprockets up to a few newton-meters. For medium torque where you still want to disassemble the unit on the bench, a parallel key with a sliding fit is the honest choice. For heavy reversing loads, splines leave keys behind, but splines cost money and need special cutters, so most non-standard machines end up with keys anyway.

That table is my own rough band, not a standard. Your ratio of hub length to shaft diameter changes all of it. A long hub spreads the load and lets you use a smaller key; a short hub on a fat shaft is where the math gets ugly.

Keyway Dimensions: The Numbers You Actually Need

For a parallel key in metric machines, I basically live on ISO 6887 (better known outside Europe as the DIN 6885 sizes everyone copies). The rule of thumb nobody prints big enough: the key cross-section grows with shaft diameter, but the key length is your free choice, and length is where most designs quietly save themselves. A typical 20 mm shaft takes a 6×6 mm key; step up to a 40 mm shaft and you are looking at a 12×8 mm key. The width follows the shaft, the depth follows the width, and the length is sized by the hub you have available.

Here is the trap with keyway depth. The depth in the shaft is measured from the edge of the keyway to the opposite side of the bore center line, and you absolutely must check the hub keyway is deep enough on the assembled bore, not just on the drawing. I have seen drawings with the hub depth dimension missing, machinists cut it to a round number, and then the key sits proud of the bore and the hub never pulls home over it. The shaft keyway depth tolerance for a metric key is usually +0.1 mm / −0.05 mm territory, and if your shop works in inches, convert with care, do not round.

Fits and Tolerances Nobody Talks About On Drawings

The classic recommendation is a snug fit between key and hub keyway, and over the years I have settled on a pattern that works across most of my builds. The key-to-shaft-keyway fit should be tight enough to hold the key when you drop the shaft vertically with a light tap, which in practice is a transition or light press fit. The key-to-hub-keyway fit should be a sliding fit so the hub can be assembled, but not a loose rattle. If the hub keyway is cut by broaching, measure it on the machine; a used broach will cut 0.02–0.04 mm narrow and someone will later file the key to force it in, which ruins the flank contact.

Material matters more than the fit. On a mild steel hub with a 45 steel key, the crush pressure limit is roughly 120–150 MPa for steady load and you should halve that for reversing load. If your hub is cast iron, treat 60–80 MPa as the ceiling and make the hub keyway longer to compensate. I would rather use a longer key in a cheap material than a short key in a fancy one, because a bent key is cheaper to replace than a cracked hub boss.

The Shear And Crush Check, Worked With Real Numbers

Let me run a quick example so the formulas sit somewhere concrete. Say we drive an indexing turret at 60 rpm through a gearmotor delivering 40 N·m, on a 28 mm shaft. That puts us in the 8×7 mm key family per the standard table. If the hub boss is 40 mm long, the effective key length in the 28 mm shaft is roughly that whole 40 mm, minus the run-out radius at each end, call it 36 mm usable.

The force on the key flank is torque divided by shaft radius: 40 N·m / 0.014 m equals about 2857 N. Crush stress is that force divided by the project flank area, which is roughly key length times (key depth in the hub, usually around half the key height for a standard keyway). For an 8×7 key the hub depth is roughly 3.3 mm, so the area is about 36 mm × 3.3 mm, roughly 119 mm². Crush stress lands around 24 MPa, well under the 120 MPa ceiling for steel, so this design is comfortable, almost boring.

Shear is nowhere near as exciting: the same force over the full key cross-section (8 mm × 7 mm, so 56 mm²) gives about 51 MPa, against a yield shear in the 140–160 MPa range for a 45 steel key. So the decision here is governed by contact pressure, as with most small machines. The story changes the moment the hub is aluminium. A 6061-T6 hub with a crush limit around 45–55 MPa will be at about half its limit in this example, acceptable, but only just, and the moment someone adds a radial load or a second machine to the same shaft, the flank will start to brinell.

T = 40 N·m, r = 14 mm
F = T / r = 40 / 0.014 = 2857 N
A_crush = L_eff × h_hub = 36 × 3.3 = 119 mm²
sigma_c = 2857 / 119 ≈ 24 MPa    (steel limit ~120, AL6061 ~50)
A_shear = 8 × 7 = 56 mm²
tau = 2857 / 56 ≈ 51 MPa    (45 steel yield ~400, tau_y ~ 160)

Keep a margin, because the handbook numbers assume perfect uniform contact. Real keyway flanks touch over maybe two thirds of the nominal length after years of fretting and dust. I usually design for a crush stress under 60 percent of the material limit, and I tell the team never to trust a drawing that promises more than that without a measured flank contact check.

Assembly Damage Is The Real Enemy

Most stripped keyways I inspect were damaged in assembly, not in service. The classic scene: a technician drops a pulley over the shaft and it will not slide past the proud key, so out comes the hammer, and the key nose gets battered until the hub finally pops over it. That smash leaves a burr at the keyway entry, the burr then works like a cutting tool against the hub flank, and within a month you have fretting and a loose fit. My rule on the bench is simple: if the hub does not slide over the key with hand force on a light tap, stop and check the depth and the burr, do not force the assembly.

Gibb head keys help here because you can pull them out and seat them after the hub is in place, but they are heavy, and the protruding head tends to catch on guards and belt covers on small machines. For small servomotor pulleys I increasingly prefer a radial screw locking an interference-fit hub, no key at all, because the shaft is short and the axial space is valuable. Honestly, the keyless route is underused on non-standard machines; it kills the stress concentration and removes a failure point for a few euros of machining time.

My Field Checklist Before Anything Ships

1. Measure the actual keyway depth in the shaft with a depth micrometer, plus the hub depth on an assembled sample; compare both to the drawing, allow +0.1/−0.05 mm.
2. Check the key fits the shaft keyway with a light tap, and slides in the hub keyway without a file or a hammer.
3. Deburr the keyway entry on both ends; a chamfer at each end prevents the assembly load from shearing a corner off the key.
4. Run the crush stress number against half the yield limit of the weakest material (aluminium hubs fail first, nearly always).
5. Verify the key length is in the hub, not floating in air; a key that overhangs the hub on the driven side does no work and vibrates.
6. If the machine reverses or indexes, add a setscrew or a locking collar to stop axial key walk; a loose key walks out of the keyway within weeks.

One last point that costs people real downtime: put the keyway position on the drawing relative to a clock angle on the shaft end. I have seen couplings keyed at arbitrary angles, so the technician fights against a mismatched hub keyway and files material off the key to make it fit, which is exactly how you lose flank contact. A simple angular dimension, say keyway at 12 o’clock when viewed from the drive end, saves twenty minutes on every assembly and prevents a whole class of misaligned-key mistakes.

Keys look like the least interesting part of a machine, but they are the part that fails silently and stops production at the worst possible moment. Size them by crush stress against the real hub material, check the assembled depth before you torque anything, and treat the key flank like the precision surface it actually is. If you build the habit of measuring the keyway on one sample from every batch, you will catch the machining drift that would otherwise cost you a shaft skid on the floor months later.

A practical note on key standards if you work across suppliers: an ISO metric key and an ANSI inch key of the same nominal width are not interchangeable, and I have watched a maintenance crew mix them on a retrofit because both measured close with a caliper. The flank angles differ slightly and the inch key sits at a different depth, so the hub ends up rocking on the key instead of seating. When you buy off-the-shelf pulleys or couplings, ask the vendor which keyway standard the bore is cut to, and mark it on the drawing title directly. It sounds like trivia until you have two suppliers machining the same hub bore to different standards and the machine loses torque on the second shift.

Do not forget axial position either. A keyway that runs the full shaft length gives you flexibility, but it also turns the whole key into a stress raiser under bending. On longer shafts I prefer two short keyways, one at each drive point, rather than one long slot, because the bending stress concentrates less and replacement keys are cheaper to cut short. Small choices like this are what separate a machine that runs for years from one that lives on the spare-parts shelf.