CNC Machining of Shells: How Thin-Wall Parts Stay in Tolerance

• 1
  CNC Machining of Shells: What Makes the Geometry Different
• 2
  Wall Thickness, Tolerance and Where the Limits Sit
• 3
  How 5-Axis Setups Change Shell Machining
• 4
  Fixturing, Chatter and Sequence Control
• 5
  Shell Machining Reference by Material and Wall

A shell is a part where most of the material has been removed and what remains is a wall. Housings, covers, motor cans, battery trays, gearbox casings and instrument enclosures all share this shape. The machined features that matter are usually a flange, a sealing face, a bearing bore and a set of mounting holes. The skin between them carries the load and closes the volume.

The machining problem is not cutting the outside. It is that the stiffness of the part drops as the wall gets thinner, and the part is no longer strong enough to hold its own shape against cutting force. A solid block at 50 mm thick barely deflects under a roughing pass. A 1.5 mm wall at the same cut depth rings, springs back and leaves a taper behind the cutter. The metal is still there, but the support behind it is gone.

That is why shell work is judged by how the part is held and how the cut is sequenced, not by spindle speed alone. Once a wall is free on both sides, every subsequent pass pushes it around. The order of operations decides whether the finished wall measures 2.00 mm or drifts to 2.35 mm on one side of the bore.

A useful rule before quoting: measure the longest unsupported wall span and divide it by the wall thickness. Under 20:1 the part behaves like ordinary prismatic work. Between 20:1 and 50:1 you need light passes and extra support. Above 50:1 the design itself is fighting the process.

For aluminium shells, 1.0 mm is a common minimum on a short wall and 1.5 mm is comfortable for anything longer than 100 mm. Stainless and steel push that up, because cutting forces are higher and spring-back is worse. On 304 or 316, plan on 2.0 mm minimum and expect to slow the finishing pass. Titanium behaves differently again: it holds a thin wall once formed but cuts hot and work-hardens if the tool rubs.

Tolerance follows the same logic. We hold ±0.005 mm on bores, spigots and sealing faces where the fit is defined. On the free wall between those features, that number is not meaningful, because the wall moves when you release the vise. A realistic callout on an unsupported 1.5 mm wall is ±0.05 mm, and even that depends on part length and how the fixture is built.

The practical approach is to tolerance only what functions. Put tight limits on bore diameters, flange flatness and hole positions. Leave the outer skin at general tolerance and let the finishing pass follow the shape. A drawing with ±0.01 mm on every surface of a thin housing will be quoted high and still argued over at inspection.

Surface finish tracks the same trade-off. Sealing faces and sliding bores often need Ra 0.8–1.6 μm. Outer cosmetic surfaces are fine at Ra 1.6–3.2 μm as machined, or Ra 0.2–0.8 μm if the part is later anodized and must show a uniform sheen. Asking for a mirror finish on a 1 mm wall means many light spring passes, which adds cost and can still chatter.

A 3-axis machine can cut the outside of a shell and one open face. The moment a part needs features on four or five sides, each additional face becomes another setup, another vise position and another chance to lose datums. On a thin wall, reclamping is often what pushes the part out of tolerance, not the cut itself.

A simultaneous 5-axis center keeps the part in one fixture and tilts the tool to reach the sides. The gain is not just fewer setups. It is that a short, rigid tool can reach a deep internal pocket at an angle instead of hanging out over the wall. Short tools chatter less, so the wall can be finished in fewer passes. Our shop runs 16 simultaneous 5-axis machining centers alongside 12 four-axis mills and 27 three-axis machines, so the setup is matched to the geometry rather than forced onto one platform.

Access still has limits. A deep internal cavity with a narrow opening and a sharp internal corner cannot be reached by any practical cutter. Tool reach, shank diameter and corner radius set the floor on internal detail. If a pocket is deeper than four times the cutter diameter, expect either a larger corner radius or a design change.

Maximum part size is 4,000 mm on our largest travels. Big shells usually go on the 4,000 × 400 × 150 mm machine, while mid-size housings fit the 750 × 1,150 × 550 mm and 600 × 600 × 600 mm envelopes. Small instrument shells run on the 500 × 500 × 450 mm or 500 × 310 × 200 mm platforms, which hold tighter geometry on short walls.

Chatter in a shell is usually a support problem, not a speed problem. The wall vibrates because nothing is behind it. Soft jaws machined to the actual part profile support far more of the surface than a standard vise, and a low-melt fixturing compound or a matched support block fills the cavity during the roughing stage.

Sequence matters as much as support. Rough the inside first while the outside is still solid, then flip and rough the outside, then finish in a light pass on both sides. Leaving a sacrificial web until the last operation keeps the part rigid for most of the cycle. Cut the web away with a light finishing pass and the wall never sees a heavy load.

Cutting data should follow the wall. A 1.0 mm aluminium wall finishes well at 0.2–0.5 mm radial depth with a high spindle speed and a sharp, polished flute. Pushing a 6 mm cutter at full width will deflect the wall and leave a visible witness mark. For stainless, drop the radial depth further and keep the tool moving; dwelling on a thin wall work-hardens the surface and dulls the cutter.

Coolant and chip evacuation matter in closed shells. Chips trapped inside a cavity get re-cut and scratch the finished wall. Through-spindle coolant or a directed air blast clears the pocket. On deep housings we plan the toolpath so chips exit the same way they entered, not into a corner where they sit.