Quality control

Why CNC parts fail inspection

Rejections are rarely random. The same handful of defects accounts for most failed batches, and each one has a known process cause you can raise with the factory.

Rejections are rarely random. A handful of defects accounts for most failed batches, and each one has a known process cause a buyer can name in an email. "The parts are bad" gets a discount; "the bore is 0.03 mm oversize" gets a corrective action.

1. Burrs

Symptom: raised metal on edges and hole exits, enough to cut a glove or stop a part seating flat.

Root cause: burr formation is normal in cutting. The defect is a missing or inadequate deburring step, and hand deburring at the end of a shift is never uniform.

Prevention: specify the edge condition instead of writing "deburr all edges". A note naming a maximum edge break turns an opinion into a measurement. Ask whether deburring is a programmed operation or hand work.

2. Chatter

Symptom: a regular wavy pattern on a wall or floor, and Ra readings that fail even where geometry passes.

Root cause: vibration in the machine-fixture-tool system. Long tool overhang, thin walls, unsupported sections, or a radial depth of cut that excites a natural frequency of the setup. Rigidity matters more than removal rate.

Prevention: shorter overhang, support under thin walls, a lighter radial cut at higher spindle speed, and a separate finishing pass. A finish requirement on the drawing gives the programmer a reason to slow down; without one, cycle time wins.

3. Tool marks

Symptom: visible parallel lines, a step where two passes overlap, or a witness mark on a face that should be clean.

Root cause: usually a stepover or feed decision: too coarse a stepover for the finish callout, or a worn tool whose effective radius no longer matches its programmed path.

Prevention: state the finish as an Ra value and name which faces are visible. A factory that does not know which surface is exposed finishes all of them to the same standard, or none.

4. Thread damage

Symptom: galled or torn threads, a thread that starts and binds, or a go gauge that will not enter. Most common on aluminium, stainless and blind holes.

Root cause: chip packing in blind holes, the wrong tap for the material, a forming tap where a cutting tap was needed, or damage after tapping in handling.

Prevention: specify thread class and go/no-go gauge inspection, and require protective caps on critical threads in packing. Ask whether the thread is cut or formed — formed threads tear differently and need a different gauge.

5. Bore out of tolerance

Symptom: a reamed or bored hole that is oversize, tapered along its length, or oval rather than round.

Root cause: a worn or chipped reamer, coolant not reaching the cutting edge, a drill that wandered so the reamer followed a bent path, or a two-point gauge that cannot detect ovality. A part measured hot also reads small and grows after cooling.

Prevention: measure with a three-point or air gauge after the part has cooled, track reamer life by tool count, and gauge every part if the bore locates.

6. Flatness and parallelism out of tolerance

Symptom: a plate that rocks on a surface plate, or a face that measures parallel at the edges and off in the middle.

Root cause: internal stress. Removing material from one side of a plate releases residual stress from rolling or heat treatment and the part bows. Vise clamping can also spring a thin part flat during machining and let it relax afterwards.

Prevention: for thin plates, rough machine, stress relieve, then finish. Specify flatness with the measurement method, because flatness differs on a granite plate and on a CMM.

7. Surface roughness above the callout

Symptom: Ra reads 1.6 µm where the drawing says 0.8 µm, while dimensions pass.

Root cause: feed per tooth too high for the required Ra, a worn insert, or a material temper that does not machine cleanly.

Prevention: state the material temper and the Ra value together, with the instrument and cut-off length. A 0.8 µm result taken with a different cut-off is not comparable, and the argument that follows is about the instrument.

8. Mixed material

Symptom: a batch that machines, anodises or weighs differently from the last, or hardness outside the specified range.

Root cause: stock of the same shape but a different grade in the same rack, 6061 delivered where 6082 was ordered, or cheaper stock bought to protect a margin. This is the defect that matters most to an end user, because it often passes inspection and fails in service.

Prevention: require a material certificate per lot, checked against the purchase order, and test independently where certificates look photocopied. Where strength is critical, add a hardness check to incoming inspection.

9. Dimensional drift: good first article, bad production run

Symptom: the first-article report is perfect. Three hundred parts later the batch is rejected, with every sampled part out on the same dimension.

Root cause: the first article proves the setup, not the process. Two mechanisms dominate:

• Thermal growth. A cold machine removes heat differently from the same machine after four hours of cutting. Spindle and part expand, the dimension moves, and the first hour of parts is not the afternoon.
• Tool wear accumulation. A small change in effective cutter radius shifts the dimension slightly on every part. Over a few hundred parts the drift exceeds the tolerance band, and because it is gradual, no single part looks like an outlier.

Prevention: pair first-article approval with an in-process check plan — a stated frequency, dimension and gauge — plus a rule for trending readings. Ask for the in-process sheet from the last batch of a similar part. A factory that can produce one controls the process; one that cannot is running to the final sample and hoping.

Summary

Defect	Common root cause	Prevention
Burrs	Hand deburring, honed or dull tools	Name the edge condition; deburr in the programme
Chatter	Tool overhang, thin walls, weak fixturing	Finish limits that force a lighter cut
Tool marks	Stepover and feed set for cycle time	Specify Ra; identify visible faces
Thread damage	Chip packing, wrong tap, handling damage	Thread class, go/no-go gauges, capped threads
Bore out of tolerance	Worn reamer, drill wander, hot measurement	Track reamer life; measure cooled parts
Flatness / parallelism	Residual stress, clamping spring	Rough, stress relieve, finish; state the method
Surface roughness	Feed per tooth, worn insert, wrong temper	State Ra with temper and cut-off
Mixed material	Substituted or confused stock	Certificate per lot; hardness check
Dimensional drift	Thermal growth, tool wear	In-process plan with trend limits

Used as a checklist before a purchase order, this list turns most quality arguments into a conversation about one named process step — the difference between finding out after the shipment that a factory could not hold a tolerance, and finding out while the parts are still on the machine.