Diagnostic CNC Plastic Inspection: How We Verify Polymer Parts

• 1
  Why plastic is harder to inspect than metal
• 2
  Tactile or optical: choosing the right diagnostic method
• 3
  When to measure: the timing rules that decide the result
• 4
  What tolerance is realistic on a machined plastic part
• 5
  Feeding inspection data back into the machining process
• 6
  Inspection method selection by feature type

A machined aluminium bracket holds its shape once the tool leaves. A POM or PEEK part often does not. Polymer chains relax after cutting, and the released stress shows up as bow, twist or a change in a bore diameter. Diagnostic CNC plastic inspection exists because a measurement taken at the wrong moment can be wrong by more than the tolerance you are trying to hold.

Three properties drive most of the difficulty. Elastic modulus is low, so a probe tip pushes the surface instead of reading it. Thermal expansion is high, so a part measured at 25 °C and a part measured at 30 °C are not the same part. Moisture absorption moves dimensions over days, not seconds. Each one needs a different countermeasure.

The practical consequence is that inspection on plastics is a timing problem as much as a metrology problem. We measure twice on stress-sensitive features: once after roughing and once after the part has rested. The gap between the two readings tells us how much the geometry is still moving.

None of this makes plastic unmeasurable. It means the measurement plan has to be written for the material, not copied from a metal job sheet.

• 1
  Low modulusTactile probing deflects the surface unless contact force is kept low.
• 2
  High thermal expansionCTE on unfilled polymers is roughly 5–10× that of aluminium.
• 3
  Moisture uptakePA and POM grow as they absorb water; condition before final measurement.

Contact probing is still the reference method for most machined plastic parts. A CMM with a low-force strain-gauge head can touch a PEEK or ABS surface without leaving a witness mark and without reading a deflected surface. We keep contact force in the low millinewton range on soft polymers, and we approach holes from both sides so the probe does not drag across a burr.

Optical methods suit a different job. A laser scanner or structured-light system captures thousands of points per second, which makes it good for freeform surfaces, thin walls and features that would bend under any stylus. The trade-off is that optical data struggles at sharp internal corners and on translucent or glossy materials unless the surface is prepared.

For most production plastic parts we combine the two. Optical scanning covers the overall form and thin sections; tactile probing confirms the critical dimensions, datum features and any GD&T callout. The combined report shows where each number came from.

There is a cost side too. A full optical scan on a complex housing can take longer to process than a short tactile program on six critical features. The inspection plan should name the features that actually matter, not measure everything because the scanner can.

• 1
  Tactile winsBores, datums, tight GD&T, anything with a hard tolerance.
• 2
  Optical winsFreeform surfaces, thin walls, high point density, no-touch needs.
• 3
  HybridScan for form, probe for the critical few. Standard on our plastic jobs.

Measuring a freshly cut plastic part is the most common source of false rejections. The part is still warm from cutting and still releasing internal stress. A bore that reads 9.98 mm at minute two can read 10.01 mm an hour later. Neither number is wrong; they describe different states of the same part.

Our baseline rule is to let the part rest at controlled temperature before final measurement. For thin-wall or high-stress geometries the rest period is longer, because relaxation follows a curve rather than a step. We log the time between last cut and measurement so the report is traceable.

Conditioning matters for hygroscopic polymers. PA and POM absorb moisture from shop air, and the growth is measurable over days. If the drawing specifies a conditioned state, the part has to reach that state before the final inspection, not after.

In-process checks sit between these two extremes. They are not for final acceptance. They exist to catch a trend early, while the machine can still be adjusted, and they deliberately use looser limits than the final report.

• 1
  Never accept hotWarm parts read undersize on outside dimensions and oversize on bores.
• 2
  Log the clockRecord time from last cut to measurement on every report.
• 3
  Condition hygroscopic gradesPA and POM should reach the specified moisture state first.

On a stable, filled polymer such as carbon-fibre reinforced PEEK, a well-supported feature can hold ±0.005 mm on a small critical dimension. That is the tight end, and it assumes the part is measured in the same thermal state it will be used in. On unfilled soft polymers, the achievable window is wider, often ±0.05 mm or looser.

Geometry matters more than material name. A bore through a thick boss is easier to hold than the same bore through a 1.5 mm wall. Thin walls deflect under cutting force and then spring back, so the finished dimension depends on how the tool path was sequenced. Short, rigid features are simply easier to measure and to make.

Feature size matters too. Very small holes below roughly 0.5 mm are hard to probe reliably and hard to cut cleanly in soft plastic. Long unsupported spans sag. If a design puts a tight tolerance on a flexible feature, that is a signal to revisit the design, not to tighten the inspection.

This is why we treat tolerance as a conversation, not a checkbox. If a callout is tighter than the material can reasonably support, the honest answer is to say so at the DFM stage rather than at the inspection bench.

• 1
  Filled beats unfilledGlass and carbon fibre reduce creep and hold tolerance better.
• 2
  Rigid beats thinThick bosses and short spans hold tighter than thin walls.
• 3
  State mattersTolerance only means something at a defined temperature and moisture.

Diagnostic inspection is only useful if the numbers go somewhere. A report that sits in a folder changes nothing. We feed measurement results back to the machinist and the programmer, so a drift on a bore becomes a cutter compensation change on the next run.

The most valuable signal is the trend, not the single reading. If five parts in a row read 0.02 mm oversize on the same feature, that is a process offset. If one part reads oversize and the next reads nominal, that is likely material or clamping variation. The two cases call for different fixes.

Tool wear on plastics behaves differently from metal. A sharp cutter leaves a cleaner edge and less stress, but it also wears faster in filled grades. Monitoring surface finish alongside dimensions catches the point where a tool starts rubbing instead of cutting, which shows up as heat and then as dimensional drift.

Closing the loop is what turns inspection from a gate into a control. The report is the record; the adjustment is the product.

• 1
  Compensate, don't just recordA consistent offset goes straight back into the tool offset table.
• 2
  Watch finishRising Ra in filled plastic often precedes a dimensional drift.