CNC lathe treatment: how a turning operation actually cuts metal

• 1
  What happens at the insert tip during CNC lathe treatment
• 2
  Which features a lathe can and cannot produce
• 3
  Chucking, datums and the errors they create
• 4
  Where the achievable tolerance actually comes from
• 5
  Turning versus milling: which process fits the feature

In turning the workpiece spins and the tool moves. The spindle holds the stock in a chuck, collet or between centers and rotates it at a set surface speed. A single-point insert feeds along the Z axis (longitudinal) or the X axis (facing and diameter). The insert shears off a chip whose thickness is set by feed per revolution, not by spindle RPM. That distinction matters: doubling RPM raises cutting speed, while doubling feed raises chip load and cutting force.

Heat leaves the cut mostly through the chip. When the chip runs too thin, heat builds in the insert and the workpiece instead. That is why a light finishing pass at high RPM can burn a surface that a heavier pass would leave clean. The practical fix is to keep feed above the nose radius limit, usually 0.05 mm/rev or more for a 0.4 mm radius insert.

Three forces act at the same time: cutting force pushing down on the tool, radial force pushing the tool away from the work, and feed force along the axis. Radial force is the one that ruins diameter control. A boring bar with a 4:1 length-to-diameter ratio deflects roughly 16 times more than the same bar at 1:1, so deep bores need a larger bar or a lower depth of cut.

A lathe is a rotational machine. Any surface that is a solid of revolution comes off it naturally: outside diameters, bores, faces, grooves, chamfers, spherical ends, tapers and threaded sections. The tool never needs to reach around a corner, because the part turns past it. That single fact drives most of the cost difference between turning and milling.

Cross holes, slots, flats and polygon shapes sit off the rotation axis. A plain two-axis lathe cannot cut them. On a mill-turn center they can be cut in the same setup, and we run 16 of those machines. That keeps one datum for the whole part, which is why a turned-then-milled shaft usually holds better concentricity than the same part run on two separate machines.

Wall thickness sets another boundary. Below about 1 mm on a 50 mm diameter aluminum tube, chuck clamping pressure distorts the bore. Spring passes and soft jaws help, but the part may still come out oval. For very thin rings, the honest answer is often to turn oversize, stress-relieve and finish after clamping pressure is released.

Length-to-diameter ratio decides whether the part is turnable at all. Between centers with a tailstock, 10:1 is routine. Past 20:1 the bar starts to whip and you need a steady rest or a different process. Long flexible shafts are usually ground, not turned, for that reason.

How you hold the part decides the tolerance you can hold. A three-jaw scroll chuck is fast but repeats to roughly 0.05 mm. A four-jaw independent chuck needs indicating but can be dialed to 0.01 mm or better. Collets are the middle ground: fast, and typically repeat within 0.02 mm on clean bar stock.

The datum is the surface the dimensions are measured from. In turning it is usually the face you machined first. If the part is flipped for a second operation and the new face is not square to the first, every axial dimension shifts. We face the second side in the same setup whenever the geometry allows, which removes that error source entirely.

Thermal growth is the quiet one. A spindle running for two hours grows a few micrometres, and a 300 mm steel shaft grows about 0.003 mm per 1 °C. For work at ±0.005 mm, that is a real share of the budget. Warm-up cycles and in-process gauging keep it under control on tight jobs.

Chip control is a setup item, not a housekeeping item. A birdnest of aluminum chips wrapping the bar will mark the finished diameter. Through-tool coolant or a peck cycle with a chip break at 0.15–0.25 mm/rev usually clears it.

A quoted tolerance is the sum of several errors, not a single machine spec. Spindle runout, tool wear, thermal drift, clamping distortion and measurement uncertainty all contribute. On a well-kept lathe, spindle runout of 2–3 μm is normal, and that alone eats half of a ±0.005 mm band.

Tool wear moves the cut in one direction. A carbide insert turning 304 stainless might wear 0.02 mm over 100 parts if the speed is pushed. That is why the first and last part of a long run can differ even when the setup never changes. Offset compensation every 20–50 parts keeps the diameter centered instead of drifting to one edge of the band.

Surface finish is a separate budget from size. Theoretical Ra from feed and nose radius is roughly feed squared divided by 32 times the radius, so a 0.8 mm radius at 0.1 mm/rev gives about Ra 0.4 μm in theory. Real values land higher because of vibration and built-up edge. We hold Ra 0.8–1.6 μm as a normal turned finish and Ra 0.2–0.8 μm when the callout demands it.

Measurement matters as much as cutting. A micrometer on a hot part reads small; a CMM on a flexible shaft may bend it. Let the part stabilize and agree on the gauging method before the first article, not after.