Aerospace CNC Processing: How Precise Parts Get Made

• 1
  What aerospace CNC processing actually controls
• 2
  Why 5-axis setups change the tolerance story
• 3
  Material behavior sets the real boundary
• 4
  Reading a tolerance callout before you quote
• 5
  Inspection and documentation on a machined part
• 6
  From prototype to production without re-qualifying
• 7
  Choosing a route by part geometry and tolerance

Aerospace CNC processing is subtractive machining of flight and ground-support hardware: brackets, housings, fittings, actuator bodies, sensor mounts, structural ribs. A rotating or stationary cutter removes material from a solid block, and the machine follows a toolpath generated from your CAD model. Nothing about that is exotic. The difficulty is that the parts are thin, the materials are stubborn, and the tolerances are tight enough that the machine, the fixture and the cutter all have to agree.

Three things decide the result. First, geometric accuracy: can the machine reach the position the toolpath asks for. Second, surface integrity: does the cut leave a subsurface layer that behaves like the parent metal. Third, repeatability: does part 40 match part 1 without an operator adjusting offsets. A shop can hit ±0.005 mm on a one-off and still fail the run.

Accuracy on paper is a stack, not a single number. Machine positioning, thermal drift over a long cycle, fixture stiffness, tool deflection and spindle runout all contribute. A 4,000 mm travel gantry and a 500 mm cube machine do not hold the same numbers, and neither does a 0.8 mm end mill pushed through titanium at a feed rate meant for aluminium.

The practical takeaway for a designer: tight tolerance should be a decision, not a default. Put ±0.005 mm where the part functions, and leave the rest at general tolerance. That single habit removes more cost from an aerospace program than any negotiation over hourly rates.

Every time a part moves to a new fixture, you add a datum shift. On a three-axis machine a complex bracket may need four or five setups, and each one re-establishes position from a feature that already carries its own error. Five-axis machining keeps the part in one clamp and rotates the tool or the table instead. Fewer setups means fewer accumulated datum errors.

That matters most on parts with features on non-orthogonal faces. A housing with ports at 30° and 60°, or a rib pattern that wraps around a curved wall, is awkward to reach in three axes. Reach the feature once and the relationship between the bore and the mating face stays inside the machine's own positioning budget instead of inside the sum of five fixture setups.

Simultaneous five-axis also lets the cutter approach at an angle. Instead of a full-width cut with a long overhang, the tool tilts and uses the side of the flute. Cutting forces drop, chatter becomes manageable, and thin walls survive. On titanium and Inconel this is often the difference between a finished part and a scrapped one.

Five-axis is not automatically better. A simple plate with holes on one face is faster and cheaper on a three-axis mill. We keep 27 three-axis machines for exactly that work. The question is whether the part has features that force multiple orientations. If it does, five-axis usually wins on both tolerance and total cycle time.

Aluminium 6061-T6 and 7075 machine cleanly and hold tight tolerance with modest effort. They are the default for brackets, panels, enclosures and prototype housings. The catch is thermal. Aluminium expands roughly twice as fast as steel, so a part that measures perfectly on a warm machine can move out of tolerance at 20 °C inspection. Let the part stabilize before the final cut.

Titanium Ti-6Al-4V and Inconel behave differently. They conduct heat poorly, so the cutting edge absorbs temperature and wears fast. They also work-harden: a cutter that rubs instead of cutting hardens the surface and the next pass gets harder still. Sharp tools, constant feed, generous coolant and no dwelling in the cut. Speeds drop to a fraction of aluminium rates.

Thin sections are the other boundary. A wall under 1 mm in aluminium or 0.8 mm in titanium will deflect away from the cutter unless it is supported. Options are leaving sacrificial stock and finishing in a second pass, adding temporary tabs, or packing the cavity with a machinable support. None of them are free, and all of them beat a scrapped part.

Magnesium AZ31B and AZ91D cut very fast and hold good accuracy, but chip handling is a safety matter, not a housekeeping one. Fine magnesium swarf ignites easily. That is a shop-floor control, not something the designer manages, but it explains why not every supplier quotes magnesium.

Plastics and composites round out the list. PEEK and POM hold tolerance well. Carbon fibre machines with diamond-coated tooling and dust extraction. Delamination at the exit face is the usual failure and the fix is a support plate or a backing material.

A drawing that shows ±0.005 mm on every dimension is not a tighter part. It is a more expensive part with the same function. Tolerance is a cost driver, and it should map to what the assembly actually needs. A mounting hole pattern that locates a bracket needs tight position, not tight diameter. A clearance hole needs neither.

Position tolerance is where most of the value sits. A true position callout of Ø0.05 mm MMC on a bolt circle tells the shop that hole-to-hole relationship matters. Hole diameter at ±0.1 mm tells them nothing important. Splitting the callout this way often lets the shop use a faster drill cycle and still pass.

Surface finish behaves the same way. Ra 0.8–1.6 μm is a normal machined finish and comes off the cutter. Ra 0.2–0.8 μm needs a finishing pass, sometimes a smaller stepover, and it takes time. Ra 1.6–3.2 μm is as-machined and is fine for most non-sealing surfaces. Specify fine finish only on sealing faces, bearing bores and sliding surfaces.

Datum structure deserves a second look. If the drawing names A, B and C on three mutually perpendicular faces, the inspector can set up the part and check it. If the datums are implicit or contradictory, the shop and the inspector will interpret them differently and the argument happens after the parts are cut. Name the datums the way the part sits in the assembly.

Tolerance only means something if it is measured. A machined part should leave the shop with evidence: raw material certificate, in-process checks at the critical operations, and a final dimensional report. On aerospace work the report is often the deliverable, not an extra. If your program needs first article inspection, say so at quoting, because it changes the routing.

CMM inspection is the standard for position and profile. For a Ø400 mm rotary table part or a long 4,000 mm frame, the CMM has to be large enough or the part has to be fixtured in sections. Both approaches are valid, but they generate different reports. Agree on which one before the run starts.

In-process monitoring catches drift before it becomes scrap. A bore that trends 0.003 mm over 30 parts is a warning about thermal growth or tool wear, not a random event. Stopping the run and re-setting the offset costs minutes. Finding the trend at final inspection costs the whole batch.

We inspect 100% of parts before shipment and provide reports on request. For prototype quantities that report is often a simple dimensional sheet. For production runs it becomes a control plan with sampling frequency and a record of every offset change.

The transition from one prototype to a 10,000-part run is where aerospace programs lose time. A prototype machined on a five-axis center with a bespoke fixture is not automatically producible on the same machine at volume. The tooling, the cycle time and the inspection plan all change. Plan the bridge early.

One approach is to design the prototype with production in mind: keep the same datums, avoid features that only exist to make the prototype easy, and use the material you intend to ship. If the prototype is 7075 and production is 6061, the tolerances will not transfer. Material substitution is a re-qualification, not a shortcut.

For low and mid volumes, five-axis machining is often the production process itself. We run 16 simultaneous five-axis centers, 16 mill-turn centers and 12 four-axis mills, with no minimum order quantity, so one part and 10,000 parts route through the same quality system. That removes the re-qualification step entirely for machined parts.

Where volume climbs past the point where chip-to-chip time dominates, casting or die casting becomes the better route and machining moves to finishing operations. The decision point is usually the cost of a roughing cycle versus the cost of tooling amortized over the run. Run both numbers before committing.