How to Read a CMM Inspection Report

A coordinate measuring machine (CMM) report looks authoritative: rows of numbers to four or five decimal places, each with a pass or fail against it. Those numbers are the end of a chain of decisions about alignment, probing and evaluation, and a report can only be judged by knowing what those decisions were.

What a CMM measures

A CMM does not measure features. It records the coordinates of individual points on the surface of the part, one at a time with a touch-trigger probe or as a continuous stream with a scanning probe. Everything after that is computation, in three stages:

1. Points. Discrete coordinates, each with some probing error.
2. Fitted features. Software fits ideal geometry, such as a plane, circle, cylinder or cone, to each group of points.
3. Evaluated characteristics. Dimensions and geometric tolerances are calculated from the fitted features.

Two consequences follow. First, the number and spread of points matters. A bore measured with four points at one depth yields a diameter and a center but says almost nothing about roundness or taper. Second, the fitting method changes the answer. A least-squares fit returns an average diameter. A maximum-inscribed fit returns the largest pin that would enter the bore, which is what governs assembly and is closer to how ASME Y14.5-2018 defines the mating size of a hole. On a bore with form error the two differ.

Alignment to datums

Before any characteristic is evaluated, the software builds a part coordinate system from measured features. For the results to mean what the drawing means, that alignment must use the drawing's datum features in the drawing's order of precedence: the primary datum levels the part, the secondary sets its rotation, and the tertiary fixes the last origin.

If the alignment is taken from other surfaces because they are easier to probe, every position, profile and orientation result is referenced to the wrong frame. A best-fit alignment, which shifts and rotates the measured data to minimize overall deviation from the CAD model, is legitimate for evaluating form or a profile tolerance with no datum references. It is misleading for location results, because it absorbs a location error into the fit.

Common mistake: Accepting a report that does not state how the part was aligned. Without the datum features, their order and the alignment method, the location results cannot be interpreted.

Datum features have form error of their own. A primary datum face probed with three points gives a different plane each time the points move; more points, spread over the full surface, give a stable alignment. The background on datum reference frames is in the guide to reading GD&T on a drawing.

Report columns

Layouts vary with the software, but the content of a CMM inspection report is consistent.

Table 1. Typical CMM report columns

The header matters as much as the rows: part number, drawing revision, serial number, program name and revision, machine identification, date, operator, units and temperature.

Deviation is the most useful column for engineering. A set of holes that all deviate in the same direction by a similar amount points to a datum or setup shift, not to random error, and that tells the machinist what to correct.

Position and profile results

Position

Position is reported as a diameter, twice the radial distance from true position to the measured axis. If a hole axis is off by 0.003" (0.076 mm) in X and 0.004" (0.102 mm) in Y, the radial error is 0.005" (0.127 mm) and the reported position is 0.010" (0.254 mm). The value is compared with the diameter of the tolerance zone and is never negative. The X and Y deviations listed beside it are information; the position row is the acceptance criterion.

Where the drawing specifies maximum material condition, the allowed tolerance is the stated value plus a bonus from the measured feature size. Check that the feature size was measured and that it supports the bonus claimed.

Profile

A profile tolerance of 0.006" (0.152 mm) is by default a zone extending 0.003" (0.076 mm) to either side of the true surface. Software reports profile in more than one way, and the difference matters. Suppose the measured deviations on a surface referenced to datums range from −0.002" (−0.051 mm) to +0.004" (+0.102 mm). Evaluated as twice the largest deviation, the result is 0.008" (0.203 mm), which fails, correctly, because one point lies outside the zone. Evaluated as highest minus lowest, the result is 0.006" (0.152 mm), which appears to pass. Look for the minimum and maximum deviations on the report, not just the single profile figure.

Measurement uncertainty

Every actual value carries uncertainty. The performance of the CMM itself is specified and verified under ISO 10360, in terms such as a maximum permissible error for length measurement, usually a small constant plus a term that increases with the measured length. That figure describes the machine measuring calibrated artifacts. The uncertainty of a specific measurement on a real part is larger, because it also includes stylus length and configuration, point strategy, part surface and form, fixturing and temperature.

A long-standing rule of thumb is that measurement uncertainty should be no more than one-tenth of the tolerance being checked, with one-quarter often accepted where one-tenth is impractical. It is a rule of thumb, not a requirement of the drawing, and the right ratio depends on the consequence of a wrong decision. For a total tolerance band of 0.001" (0.025 mm), the one-tenth rule asks for an uncertainty of about 0.0001" (0.0025 mm).

A result at the edge of the tolerance is therefore not a confident pass: a value close to a limit could lie on either side of it. Decision rules such as those in ISO 14253-1 reduce the acceptance zone by the uncertainty when conformance has to be proven.

Temperature and fixturing

Dimensions on a drawing apply at the reference temperature of 20 °C (68 °F). A part measured warmer is larger. Aluminum alloys expand by roughly 13 µin/in/°F (23 µm/m/°C), about twice the rate of steel, so a 10" (254 mm) aluminum part that is 5 °C (9 °F) above reference is longer by about 0.0012" (0.03 mm). That is more than the entire tolerance on many precision features.

A part taken straight from a machine, warm from cutting and coolant, needs time to stabilize before it is measured. CMM software can compensate using temperature sensors and an expansion coefficient, but the correction is only as good as the coefficient assumed for the alloy. The report should record the temperature and whether compensation was applied.

Fixturing has a similar effect on thin or flexible parts. Clamping can pull a part flat on the CMM that is not flat when released, or distort a thin-walled bore out of round. Under ASME Y14.5, tolerances apply in the free state unless the drawing specifies a restrained condition, so measuring a part clamped is valid only if the drawing defines that restraint.

Tip: For thin-walled or large parts, state on the drawing how the part is to be supported or restrained during inspection.

Red flags in a report

• No statement of datum alignment, or a best-fit alignment used for located features.
• No temperature, machine identification, program revision or drawing revision in the header.
• Actual values at exactly nominal, or identical from part to part.
• A position result with bonus tolerance where the drawing has no MMC modifier.
• A profile result with no minimum and maximum deviations shown.
• Many results just inside a limit, with no mention of uncertainty.

Checklist

• Ask which datum features were used for alignment and how many points were taken on each.
• Ask whether features were scanned or touch-probed, and the point count on features with form or profile tolerances.
• Ask how profile was evaluated, and request minimum and maximum deviations.
• Ask for the part temperature at measurement, whether compensation was applied and how the part was fixtured.
• Ask for the estimated uncertainty on the tightest characteristics and the calibration status of the machine.
• State these expectations and the report format in the purchase order. For new parts, tie them to the first article inspection requirements.