A pump drive shaft coupling kept failing on a 30 kW diesel-driven pump. The original design used a straight-sided spline — 8 teeth, 40 mm major diameter, with a keyway cut through the hub for extra “safety.” The coupling hub cracked at the keyway corner after 6 months. The replacement used an involute spline with the same major diameter. The crack disappeared. The difference wasn’t the material — it was the tooth geometry and how the load got into the hub.
What the two spline types are
A straight-sided spline has teeth with parallel flanks. The sides of the teeth are flat, perpendicular to the root circle, and the tooth tip is a flat land. It’s cheap to make (broach or mill), tolerant of misalignment, and used in agricultural equipment, old machine tools, and low-speed drives.
An involute spline has teeth shaped like gear teeth — the flanks are involute curves. The teeth center on the pitch circle, and the load transmits through the involute flank contact. It’s made with gear-cutting tools (hob, shaper, or broach with involute form). It costs a bit more per piece but carries load better and centers more precisely.
Why the straight-sided spline cracked
Two compounding problems.
The keyway was the weak point. Cutting a keyway into a splined hub removes material from the hub wall at the worst location — the root of a tooth where the stress is already highest. The keyway corner is a sharp stress raiser. The hub was 25 mm wall thickness at the tooth root, and the keyway cut it down to 18 mm at one spot, with a sharp corner. Under the 30 kW diesel’s torque pulses, the corner cracked and propagated.
Straight-sided splines don’t center well on the flank. They center on the major diameter (if the fit is close) or they wobble. The wobble under load makes the load concentrate on one or two teeth instead of spreading across all 8. Two teeth carried the 30 kW torque spike; the stress at those roots was 3-4x the average. The crack started at the loaded tooth root.
The involute fix
Involute splines center on the tooth flanks. The 30° pressure angle flank contact self-centers the hub on the shaft. The load distributes across all the engaged teeth — with the involute form, even with slight angular misalignment, the flank contact spreads the load. The 8-tooth involute spline (30° PA, flat root) carried the same torque with a 1.5x stress reduction at the roots. No keyway, no stress raiser, no crack.
The math: straight-sided spline stress concentration at the root corner was about 2.5-3.0 (Kt). The involute flat-root spline has a Kt around 1.3-1.5. Combined with the load distribution improvement (2 teeth → 6 teeth engaged), the actual stress at the critical point dropped by about 60%. The hub wall stayed 25 mm and never cracked again.
The sizing formula
The standard involute spline capacity check (based on ANSI B92.1) uses the transmitted torque and the permissible tooth load:
T = (F × D_p × L × N × K) / 2
where T is torque, F is the allowable tooth load per unit length (about 20-30 MPa for hardened steel on hardened steel, derated for dynamic loads), D_p is the pitch diameter, L is the engagement length, N is the number of teeth… the full AGMA calculation is longer, but the practical sizing rule is simpler:
For a 30 kW pump at 1500 RPM, torque = 30,000 / (2π × 1500/60) = 191 Nm. A 40 mm pitch diameter involute spline with 30 mm engagement and 8 teeth, at a conservative allowable flank pressure of 25 MPa, handles about 400 Nm. Safety factor 2.1. The straight-sided version at the same geometry, with 2 teeth effectively engaged, handled about 120 Nm. That’s why it failed.
Fit classes and when they matter
Involute splines come in fit classes that define how tight the hub is on the shaft. The main ones:
| Fit | Clearance (sliding) | Use for |
|---|---|---|
| Sliding fit (H/h) | 0.02-0.10 mm | Axially moving splines, gear shifts |
| Close fit (H/f) | 0.01-0.03 mm | Fixed hubs, easy assembly |
| Press fit (H/k) | Interference to 0.02 mm | Fixed hubs, high torque, no axial movement |
The pump drive used a close fit (H/f). The hub slid on by hand with a light tap. Under load, the flank contact holds the torque and the fit prevents radial play. A sliding fit would have added 0.05 mm of radial play, which under torque pulses creates impact loading at the flanks — that wears the teeth over time. For a permanently coupled pump, close fit is right.
Broaching vs. hobbing and tolerances
Involute spline hubs are broached — the broach cuts the full tooth form in one pass. The hub bore must be drilled and reamed to the correct root diameter first. Shafts are hobbed or ground. The tolerance class (like 5H/4h) determines the fit. The practical tolerances: pitch diameter within 0.02 mm for a close fit, tooth spacing within 0.01 mm. Broached hubs hit these numbers consistently.
The cost difference: an 8-tooth involute broach costs more than a straight-sided spline cutter, and the hub broaching operation needs the right machine. For one-off parts, the tooling cost is real. For production runs over 100 pieces, the per-part cost delta is small. The pump drive was a production pump — the tooling amortized fine.
When straight-sided is still fine
Straight-sided splines aren’t wrong everywhere. They’re fine for: low-speed drives (under 500 RPM), light loads (below 5 kW), applications where the hub needs to slide (the flat flanks tolerate axial movement), and where the shaft/hub are both soft steel. A 4 kW conveyor drive with a straight-sided spline, correctly sized with the load spread across the teeth, runs for a decade. The failure case was 30 kW with a keyway added on top — the combination of high load and a stress raiser.
The coupling cracked because a keyway cut into a splined hub removed the wall at the tooth root, and straight-sided splines concentrated the load on two teeth. The involute spline spread the load, removed the keyway, and dropped the root stress 60%. Size splines by the engaged-teeth load, not the major diameter. And never cut a keyway into a spline that’s already carrying torque.
