INSPECTION · September 7, 2026 · 7 MIN READ

CMM Inspection of Forgings: Datum Strategy, Probe Selection and the Draft-Angle Trap

A coordinate measuring machine resolves to 0.001 mm, but on an as-forged surface the measurement uncertainty is dominated by the part, not the machine: a 5° drafted wall deflects a 2 mm stylus by about 0.0035 mm per 0.04 mm of contact error, and coarse forged surfaces add 0.02–0.05 mm of probing scatter. Plan the datum strategy before choosing the machine, or the report will be precise and wrong.

WHERE THIS APPLIES AT FORGEPRESS CHINA

ForgePress China issues first-article reports from a CMM with the datum scheme and alignment method stated, in a 20 ± 2 °C inspection room. Capacity: HKM3 hot-die forging line 1,000–5,000 t, part weight 0.1–500 kg, maximum envelope 1,200 × 800 mm, as-forged to ISO 8062 class C with critical dimensions machined to ±0.2 mm, process control to ISO 9001, first article 4–8 weeks from die design release.

KEY NUMBERS

Probing scatter on as-forged surfaces: 0.02–0.05 mm; on machined surfaces 0.002–0.005 mm.
A 2 mm ruby stylus on a 5° draft wall produces cosine error; use a smaller stylus or a scanned best-fit.
Best-fit alignment is right for form checks, datum alignment is right for functional checks — they give different answers.
Scanning 20,000 points on a 200 mm forging takes 3–6 minutes; touch-trigger point checks take 30–90 s.

Datum strategy decides whether the report means anything

There are two legitimate ways to align a forging on a CMM, and they answer different questions. Datum alignment fixes the part to the drawing's datums and then measures — this is what a customer means when they ask whether the part conforms. Best-fit alignment minimises the deviation over the whole surface and then measures — this is what a die correction needs, because it shows where the die is wrong rather than where the part sits.

Reporting a best-fit result as a conformance result is one of the most common inspection errors in forging. The best-fit will "improve" a part by 0.1–0.3 mm simply by rotating and translating it, which hides a real die problem and simultaneously hides a real datum problem.

The draft-angle trap

Forged vertical walls have 3–7° of draft. When a spherical stylus touches a sloped surface, the contact point is not directly below the stylus centre, so the machine records a point that is offset from where the software thinks it is. With a 2 mm ruby on a 5° wall the offset is a few micrometres per contact, which is small — but the surface finish scatter is 20–50 µm, and that dominates.

The practical consequences: never assess a drafted wall's position from a single point; scan a line and fit; and never specify a tolerance tighter than 0.05 mm on an as-forged drafted feature, because the measurement cannot support it.

Feature	Recommended strategy	Expected uncertainty
Machined datum pad	5–9 points, plane fit	0.002–0.005 mm
As-forged flat surface	Scanned line or patch, least squares	0.02–0.04 mm
Drafted wall	Scanned section, then project	0.03–0.06 mm
Forged hole (cored)	Cylinder fit on 3 sections	0.02–0.05 mm
Machined bore	Cylinder fit on 5 sections	0.003–0.008 mm
Free-form contour	Point cloud vs CAD, colour map	0.03–0.08 mm

Uncertainty figures assume a 2–3 µm machine in a temperature-controlled room at 20 ± 2 °C. A forging measured on the shop floor at 28 °C will show additional thermal error of about 0.01 mm per 100 mm for steel.

Probe and stylus selection

Use the shortest, stiffest stylus that reaches the feature. A 2 mm ruby on a 20 mm carbon-fibre stem is a good general choice for forgings; a 5 mm ruby on a 50 mm stem bends measurably under 0.1 N of trigger force and adds error on steep walls.

For high-volume first-article work, scanning is faster than point-to-point and produces enough data to see form error. For production sampling of a few critical dimensions, touch-trigger is faster and adequate.

What to put in the inspection plan

Datum definition with the number of points and the fit type for each datum feature.
Which dimensions are checked as-forged and which after machining.
Alignment method for each report: datum or best-fit, stated explicitly.
Temperature at measurement and the thermal correction applied.
Acceptance rule and the handling of out-of-tolerance results, including who is notified and within what time.

Frequently asked questions

Can a CMM measure an as-forged surface accurately?

It can measure it repeatably to about 0.02–0.05 mm, which is sufficient for as-forged tolerances of ±0.3 mm and marginal for ±0.1 mm. For anything tighter, machine the datum and the measured feature first.

Should first-article inspection use best-fit or datum alignment?

Datum alignment, because that is what the drawing defines and what the customer will use. Use best-fit additionally, as a separate report, when you need to know how to correct the die — but never present best-fit as conformance.

How many points are enough for a plane?

Five to nine for a machined plane, and a scanned patch for an as-forged one. Three points define a plane mathematically but say nothing about its form, so a three-point check can pass a surface that is visibly bowed.