CNC machining realistic inspection: matching the method to the tolerance

• 1
  Why CNC machining realistic inspection starts at the drawing
• 2
  What each inspection method can and cannot resolve
• 3
  Temperature, clamping and material effects on the reading
• 4
  How many parts to measure, and when 100% is the wrong choice
• 5
  Choosing an inspection method by tolerance and feature

Inspection is usually treated as something that happens after the spindle stops. In practice the measurement plan is set the moment a tolerance is written on the drawing. A ±0.1 mm callout on a 20 mm bore is routine work. The same callout on a 3 mm wide slot in a soft plastic pocket is a different job, because the probe, the clamping force and the temperature all start to matter as much as the machine.

Realistic inspection means three things line up: the tolerance band, the geometry that has to be checked, and the gauge that can physically reach it. Miss one and the report becomes paperwork. The number on the page looks fine, but nobody has proved the part.

This matters most for engineers working to ±0.005 mm or better. At that level the difference between a good part and a rejected part is often the measurement setup, not the cut. A part that measures out of tolerance on a surface plate may sit dead center on a CMM with the right datum alignment.

• 1
  Tolerance bandHow much room the drawing actually leaves after stack-up
• 2
  Feature accessWhether a probe or stylus can reach the surface without collision
• 3
  Gauge resolutionRule of thumb: gauge resolution should be 10× finer than the tolerance

Calipers and micrometers are the fastest tools on the floor, and for a lot of turned and milled features they are enough. A 0-25 mm micrometer with 0.001 mm resolution reads a shaft diameter reliably to about ±0.01 mm when the operator has good technique. Push it toward ±0.005 mm and the reading starts to include the operator, the part temperature and how the anvil sits on the surface.

Height gauges and bore gauges sit in the middle. They cover more geometry than hand tools and cost far less time than a CMM run. The weak point is the datum. If the drawing calls out position relative to a machined face, a height gauge setup that references the wrong face will produce a repeatable number that is repeatably wrong.

CMM and optical measurement handle the hard cases: true position, profile of a surface, concentricity, and anything with a compound angle. They are the only realistic route to verifying a ±0.005 mm position callout on a 5-axis part. The trade-off is cycle time and fixturing. A CMM program that takes 25 minutes per part does not belong on a 10,000 piece run unless the sampling plan says so.

• 1
  Hand toolsBest for single features at ±0.01 mm or looser
• 2
  Height and bore gaugesGood coverage, sensitive to datum choice
• 3
  CMM and opticalNeeded for true position, profile and compound angles

Aluminum expands about 23 × 10⁻⁶ per °C. On a 200 mm aluminum part, a 5 °C difference between the machine and the inspection room moves the length by roughly 0.023 mm. That is four times a ±0.005 mm band. If a shop measures a warm part straight off the machine and compares it to a drawing at 20 °C, the reading is not wrong, it is just answering a different question.

Clamping does the same thing in a different direction. Thin walls, plastic parts and long slender shafts deflect when they are held. The classic failure is a bore that measures round in the fixture and oval on the surface plate. Both readings are real. The question is which one matches how the part works in service.

Material choice changes the risk profile too. Titanium and stainless move less with temperature but cut with more tool pressure, so residual stress can release after machining and shift a feature by a few thousandths of a millimeter. Plastics can absorb moisture and grow. For ABS, POM, PA and PEEK parts, a settle period before final inspection is often more useful than a tighter gauge.

We plan inspection around these effects. Raw material check, in-process monitoring and final inspection all happen at GreatLight before shipment, and reports are available on request. What we will not do is quote a measurement the setup cannot support.

• 1
  Let parts stabilizeEspecially aluminum at tight tolerance and plastics after machining
• 2
  Fix the datum firstAgree on the datum before anyone writes a program
• 3
  Match the reading to serviceUnclamped, in-use condition is usually the one that counts

100% inspection sounds like the safe answer. On a 10,000 piece run it can be the expensive one, because the cost of measuring every part may exceed the cost of the parts. The engineering question is whether the process is capable enough that sampling will catch a drift before it produces scrap.

Process capability is the deciding factor. If a feature runs at a stable center with a comfortable margin, checking first article, then a periodic sample, then final inspection gives better control than inspecting every piece with a tool that cannot resolve the tolerance anyway. If the feature sits near the edge of the band, or the material is inconsistent, inspection on every part is justified.

The parts that need full inspection are rarely the ones people expect. A simple turned bushing at ±0.05 mm may need a check every 50 pieces. A thin-walled housing with a true position callout at ±0.005 mm may need every part on a CMM, because a single tool wear event can walk the feature out of band within a few cycles.

• 1
  Stable processFirst article plus periodic sampling plus final check
• 2
  Marginal processTighter sampling interval or 100% depending on risk
• 3
  Critical featuresFull inspection regardless of run size