CMM inspection: when it is needed and when it is not

What a CMM measures

A coordinate measuring machine moves a probe through three axes and records points on the part. Software fits lines, circles, planes and cylinders to those points, then evaluates dimensions and geometrical controls against the drawing.

It is not a camera that produces a verdict. The result depends on the program, the probe, the alignment, the fixturing and the point density. Two labs can measure the same part on the same machine and report different numbers if those choices differ.

The output looks precise because it comes with three decimal places and a software report. Precision is not accuracy. A well presented report can still rest on a program that measured the wrong datum or a probe that was not qualified.

When a CMM earns its place

A CMM is strongest where the drawing uses a datum reference frame and a geometrical control. Position, profile, perpendicularity, parallelism and flatness all depend on how the part is aligned and how many points are taken. A CMM can apply the same alignment and the same evaluation across every part.

It is also useful for first articles, for complex three dimensional geometry, and for resolving a dispute between a supplier and a customer. When two parties disagree about a feature that a caliper cannot capture, a shared CMM program gives them a common reference.

A drawing with a position tolerance at maximum material condition is a good example. The CMM can evaluate the position, apply the bonus tolerance from the actual size and report the result against the virtual condition. A functional gauge can also check it, but it will not show how much margin remains.

When it does not

A simple length or diameter is usually faster on a micrometer, and the measurement is easier to explain. A thread is better checked with a go and no-go gauge. Surface finish needs a stylus instrument, because a CMM cannot resolve the roughness profile.

High volume production often needs a hard gauge or a vision system at the line, not a CMM in a metrology room. A CMM can validate the gauge and audit the process, but it is a poor use of a shared machine to measure the same simple feature every hour.

Thin, flexible or soft parts are another limit. Clamping can distort the part, and the CMM will measure the distorted shape accurately. If the part must be measured free state, say so and use a support that does not load it.

Programming and fixturing

A CMM program is a measurement method. It defines the probe, the tip, the alignment, the point locations, the filters and the evaluation. Offline programming from the CAD model can save machine time, but the program still has to be proved on a real part before it is used for acceptance.

The fixture has to hold the part in a repeatable position without over-constraining it. For a machined part, three points on a primary datum, two on a secondary and one on a tertiary is a common approach. For a thin part, the support may need to match the free state rather than force it flat.

The probe tip radius has to be calibrated, and the program has to correct for it. A worn or dirty tip changes the effective radius and shifts the fitted feature, which can look like a process shift. Clean and qualify the tip on a regular schedule.

Environment and uncertainty

Most CMM accuracy statements assume a reference temperature of 20 degrees Celsius and a controlled environment. A shop floor that swings by ten degrees will move the part and the machine. Thermal compensation helps, but it does not remove a gradient through the part.

Machine acceptance is usually verified against the ISO 10360 series, which covers the performance of the machine and the probing system. The measurement uncertainty for a specific part is a separate question, and the ISO 15530 series gives methods for estimating it. A calibration certificate for the machine is not a statement of uncertainty for your feature.

Reading a CMM report

A useful report names the program and its revision, the software, the probe and tip, the alignment, the temperature and the number of points. It shows the nominal, the tolerance, the measured value and the deviation for each characteristic. Out of tolerance results should be visible, not buried in a table.

Check the alignment first. If the datum reference frame in the program does not match the drawing, every geometrical result is suspect. Then check the point density on form and profile features, and the filter settings on a scanned surface. A report with 400 points and no alignment note is hard to audit.

Cost and throughput

The cost of a CMM check has three parts: programming, setup and run time. Programming is front loaded, so the first part is expensive and later parts are cheaper. Setup includes the fixture, the probe change and the alignment. Run time depends on the number of points and the scanning speed.

A CMM is usually a shared resource, so queue time can matter more than the measurement itself. If a production cell waits two days for a CMM slot, the inspection is not controlling the process. In that case, a line side gauge with a daily CMM audit is the better arrangement.

Quoting an inspection cost per part without the programming and setup is misleading. For a new part, the first CMM report may carry most of the engineering effort. Spread that effort over the expected volume before comparing it with a gauge.

Choosing the method

Match the method to the characteristic and the volume. The table is a starting point, not a rule. The drawing, the tolerance and the customer requirement decide the final method.