CNC machining house: how a job actually runs

• 1
  What a CNC machining house actually controls
• 2
  Step one: the DFM review that decides the process
• 3
  How workholding and datum choice drive accuracy
• 4
  When a 5-axis CNC machining house is the right call
• 5
  Holding ±0.005 mm in a production run
• 6
  The measurement plan, not just the machine
• 7
  Which process fits your part

A CNC machining house turns a 3D model into a metal or plastic part by removing material with computer-controlled cutters. That much is simple. What separates one shop from another is the set of decisions that sit between the model and the finished part: stock size, workholding, tool selection, axis count, and the measurement plan.

Those decisions are made before the spindle ever turns. A shop that reviews the drawing carefully will catch a thin wall, a deep pocket, or a datum that cannot be reached with a probe. A shop that does not will cut the part anyway and argue about it later.

The machine is rarely the limiting factor. A 3-axis mill with good fixturing holds tighter results than a 5-axis machine running a part that was never set up properly. Axis count expands what shapes are reachable. It does not fix a bad setup.

• 1
  GeometryUndercuts, deep cavities and angled faces decide the axis count.
• 2
  Tolerance±0.005 mm demands temperature control and a defined datum scheme.
• 3
  VolumeOne prototype and 10,000 parts need different processes.

Design for manufacturing is not a courtesy. It is where the shop decides whether your part is milled, turned, or milled and then turned on a mill-turn center. A shaft with cross-holes is a mill-turn job. A flat plate with pockets is a 3-axis job. A housing with features on five faces is a 5-axis job.

During the review, the shop checks wall thickness against the material. Aluminum 6061 tolerates thin walls better than 316 stainless. A 0.8 mm aluminum wall is routine. The same wall in 17-4PH will deflect and chatter.

Tool reach sets another boundary. A pocket 60 mm deep with a 6 mm corner radius needs a long, slender cutter. That cutter deflects. The shop may need to open the corner radius, reduce the depth, or accept a slower pass with a smaller stepover.

At GreatLight, DFM feedback and a quotation come back within 12 hours. The point is to raise these issues before the material is cut, not after.

• 1
  Corner radiusKeep it at least one third of the pocket depth when possible.
• 2
  ThreadsSpecify class and depth; blind holes need room for the tap.
• 3
  Surface finishRa 0.8–1.6 μm is a normal machined callout, not a default.

A part is only as good as the surface it is held on. The first operation establishes the datum. Every later operation is measured from it. If the datum face is not flat, or not accessible to the probe, the tolerance stack drifts.

Soft jaws machined to the part profile are common for production. For a one-off prototype, the shop may use a vise with parallels or a custom fixture plate. Both work. The choice depends on how many parts follow and how repeatable the load has to be.

Thin plates warp when they are unclamped. The shop relieves stress by roughing, letting the part rest, then finishing. Aluminum 7075 and 2024 move more than 6061. If the drawing calls for a flatness of 0.02 mm over a 300 mm plate, that relaxation step is not optional.

The same logic applies to hole position. Drill, then ream, then measure. Drilling alone leaves a hole 0.05 mm oversize or worse.

A 5-axis machine moves the cutter and the part at the same time. That lets the shop reach five faces in one setup, cut compound angles, and use a shorter, stiffer tool on contoured surfaces. Shorter tools deflect less, which shows up directly in the finish.

The trade-off is programming time and cost. A 5-axis setup for a simple bracket is wasted effort. Use 3-axis when the part has features on one or two faces and the tolerances are normal.

Use 5-axis when the part has curved surfaces that must blend, when the number of setups has to be reduced to hit a hole-to-hole tolerance, or when the geometry cannot be reached from three directions at all. Turbine blades, impellers, and medical housings fall into this group.

At GreatLight there are 16 simultaneous 5-axis machining centers, 12 four-axis mills, 27 three-axis machines, and 16 mill-turn centers. The largest travel is 4,000 × 400 × 150 mm. A Ø400 mm rotary table handles round parts that need angular features.

• 1
  3-axisPlates, brackets, simple pockets, one or two faces.
• 2
  4-axisCylindrical parts with slots or flats around the axis.
• 3
  5-axisContoured surfaces, compound angles, tight multi-face position.

±0.005 mm is 5 micrometers. A human hair is roughly 70. At that level, the machine, the tool, the coolant, and the room temperature all matter. A shop that quotes that tolerance must also control where the part is measured and at what temperature.

Carbide tools wear. On a long run, the shop compensates by measuring the first part, then checking at intervals, then adjusting the offset. That is in-process monitoring, and it is the only way the last part matches the first.

Surface finish and tolerance are linked. A Ra 0.2–0.8 μm finish usually means a finishing pass with a small stepover and a sharp tool. Push the feed rate and the finish degrades to Ra 1.6–3.2 μm, which may still be acceptable for a bracket but not for a seal face.

Every part is inspected before shipment at GreatLight, with reports available on request. Raw material certificates, in-process checks, and final inspection records are the evidence trail a buyer should ask for.

A CMM report means little if the datum scheme does not match the drawing. The shop should define which features are datums, how the part is aligned, and which dimensions are reported. Ambiguity here causes good parts to be rejected and bad parts to pass.

For simple parts, calipers and micrometers are enough. For a first article, a CMM with a stated program is better. For a production run, a fixture gauge is faster and more repeatable than measuring every feature on every part.

Ask what happens when a dimension drifts. A shop with a plan stops the run, checks the tool, and re-cuts. A shop without one ships the parts and hopes. The historical late-delivery probability at GreatLight is below 2%, and that comes from catching problems early, not from rushing at the end.