Spline CNC: how spline profiles are actually cut on CNC wells

• 1
  How a spline profile is generated on a spline CNC well
• 2
  Milling, hobbing or grinding: what each route can hold
• 3
  Runout, lead error and the tolerances that actually matter
• 4
  Material and heat treatment effects on spline accuracy
• 5
  Spline CNC route vs what it can hold

A spline is not a shape the tool can copy. It is a shape the machine computes. On a spline CNC well the control takes the tooth count, module or diametral pitch, pressure angle and root radius, then generates the flank point by point along the axis. The cutter or wheel never sees the finished profile in one pass.

The well itself is the work envelope. On our 4,000 × 400 × 150 mm travel machines a long shaft can sit in one setup, which matters because every re-chuck adds runout. Short splines on a Ø400 mm rotary table are a different problem: the table index has to hold angular position while the tool moves in Z.

Two families of motion exist. Form milling uses a cutter shaped to the gap, so the flank comes from the tool geometry. Generative motion uses a standard cutter and lets the control interpolate the flank. Form tools cut faster but lock you to one tooth form. Generative motion is slower and far more flexible when the drawing changes.

The practical consequence: a spline that fits in a standard involute family can be generated from a few numbers. A spline with a modified root, asymmetric flank or non-standard pressure angle needs a dedicated tool path and a first-article check before anyone runs a batch. That check is where most spline programs are won or lost.

• 1
  Standard involuteGenerated from tooth count, module and pressure angle
• 2
  Modified root or flankNeeds a custom path and a first-article sign-off

Milling a spline on a 3-axis or 4-axis machine is the fastest way to a prototype. The end mill follows the flank with small stepovers, often 0.05–0.15 mm, and the result is a usable part in hours. The limit is the root. A small root radius is hard to reach with a spinning cutter, and the flank finish lands around Ra 1.6–3.2 μm.

Grinding after heat treatment is the route for anything that has to hold size after hardening. A formed wheel removes 0.05–0.30 mm of stock and leaves Ra 0.8–1.6 μm on the flank. The wheel has to be dressed to the profile, so grinding cost scales with tooth form complexity, not with part count.

Between those two sits hard milling with carbide or CBN tooling. It suits medium batches where grinding is too slow but the hardness is above what HSS can handle. We use it on 4140 and 4340 shafts where the flank tolerance is ±0.005 mm and the part is too long to grind in one pass.

Wire EDM and broaching cover the edges. Wire EDM gives an accurate profile on a hardened blank but is slow and leaves a recast layer that has to be removed. Broaching needs a dedicated tool and a minimum batch to justify it. For one-off work, neither usually wins.

• 1
  Prototype3-axis or 4-axis milling, 0.05–0.15 mm stepover
• 2
  Post-hardeningForm grinding, 0.05–0.30 mm stock, Ra 0.8–1.6 μm
• 3
  Long hardened shaftHard milling on a 4,000 mm travel machine

Total runout is the number most drawings forget. A spline can be perfectly sized on the flank and still fail in the assembly because the pitch circle runs out of true by 0.03 mm. On a well, runout comes from three places: the chuck or arbor, the part itself, and thermal drift during the cut. Fixing the first two is setup work. The third one is machine time.

Lead error is the other silent variable. A spline is a helix, and the control has to keep the tool on that helix as the axis moves. If the feed rate is too high on a long part, the flank drifts and the tooth gets a taper that never shows on a short sample. On a 500 mm spline, we keep the finishing pass slow and let the machine hold the lead rather than push the cycle time.

We hold ±0.005 mm on spline flank and pitch dimensions for parts that are inspected on the machine, and Ra 0.8–1.6 μm on ground flanks. Those numbers assume the part is not re-chucked after the spline is cut. Every additional setup adds roughly 0.01 mm of runout risk, which is the whole budget on a tight spline.

Inspection follows the same logic. We check raw material before the first cut, monitor the flank during the run, and do a final check before shipment. On splines we report tooth-to-tooth spacing, total runout and flank finish. If your assembly is sensitive to backlash, ask for the lead measurement too.

• 1
  Total runoutWatch the pitch circle, not just the flank size
• 2
  Lead errorGrows on long splines if finishing feed is pushed

Heat treatment moves a spline. Case hardening and through hardening both change the part, and the amount depends on section size and quench direction. A spline cut to final size before hardening will not be to final size after. Either grind after hardening or leave stock and accept the distortion.

For 4140 and 4340 shafts, we normally cut oversize, harden, then grind the flank. For 17-4PH (SUS630), which can be aged to around 40 HRC with much less movement, milling after aging is often good enough and avoids a second grinding setup. The choice follows the drawing tolerance, not the material name.

Stainless 303 and 304 cut cleanly but gall on the flank if the tool dwells. We keep the feed steady and avoid spring passes on those grades. Titanium TC4 (Ti-6Al-4V) needs lower surface speed and more coolant; the flank finish suffers before the size does, so check Ra on the first article.

Aluminium splines are usually a fit-and-clearance problem, not a wear problem. 6061-T6 and 7075 machine fast and hold size well, but they wear quickly in a running spline joint. If the part will cycle, anodizing helps the surface and changes the dimension by a few micrometres, which has to be in the drawing.

• 1
  4140 / 4340Cut oversize, harden, then grind the flank
• 2
  17-4PH (SUS630)Age first, then mill if tolerance allows
• 3
  6061-T6 / 7075Fast to cut; anodizing shifts size slightly