GlossaryKind of term: Manufacturing termProcess capability (Cpk)

What is process capability?

Process capability expresses how comfortably the variation of a stable process fits within the tolerance, in indices such as Cp and Cpk.

Process capability describes how well a stable process can make parts within tolerance. Cp compares the tolerance width with the spread: (USL − LSL) / 6σ, with USL and LSL the upper and lower limits and σ the standard deviation within subgroups of consecutive parts. Cpk also accounts for location: the distance from the mean to the nearest limit, divided by 3σ. At a Cpk of 1.33 that limit lies four standard deviations from the mean. Pp and Ppk use the same formulas with the overall standard deviation of all values, and are therefore known as process performance indices.

What does process capability depend on?

A Cpk is an estimate from a sample, not a property you read off. The figure only means something if the process is in statistical control, without drift or special causes, and if the distribution is roughly normal or handled with a suitable method. That matters for features like flatness, position or runout: they have only an upper limit, are bounded at zero and are seldom normally distributed. The measurement system's variation has to be small relative to the tolerance, or you mostly measure gauge variation. And an index from few parts carries a wide uncertainty.

What does the standard say about process capability?

Terms and calculation methods are set out in the ISO 22514 series, with a part for general concepts and a part for capability and performance. In automotive, the reference manuals for statistical process control and PPAP set the usual requirements; commonly cited are a Ppk of at least 1.67 for an initial process study and a Cpk of at least 1.33 for an ongoing process. Which index, which threshold, which sample size and which features apply is set by the customer in its quality requirements or on the drawing. A capability requirement without those conditions is ambiguous.

How does process capability work in practice?

A tolerance you hit with one good part is not yet a tolerance you hold at a Cpk of 1.33. To reduce the spread or centre the mean you can program to the middle of the tolerance, choose a stiffer setup or machine, add a finishing pass or change tools sooner, which can cost cycle time. Demonstrating it takes time too: a capability study, often a measurement system analysis, and measurements for a control chart during the run. A Cpk requirement is therefore a requirement on dimension, route and inspection together.

What is process capability often confused with?

Process capability is not variation: variation is what you measure, capability sets it against the tolerance. Nor is Cpk the same as Ppk: Cpk uses the spread within subgroups of consecutive parts, Ppk the overall spread of all values in the study, including shifts between subgroups.

Frequently asked questions

What does a Cpk of 1.33 mean?

That the nearest tolerance limit lies four standard deviations from the process mean. That only holds when the process is stable and the distribution roughly normal; otherwise the figure says less than it seems to.

What is the difference between Cp and Cpk?

Cp only says whether the spread fits within the tolerance. Cpk also accounts for where the mean sits. A process with a good Cp and its mean close to a limit has a low Cpk.

Can you calculate a Cpk for flatness or position?

Yes, but not simply with the usual formula. These features have only an upper limit and are bounded at zero, so there is no Cp and the distribution is seldom normal. Use a one-sided index with a method for non-normal distributions.

Your shop's terms. In every estimate.

Blake lays out the requirements from drawing and model as proposals and builds the cost per operation, with your own machines and rates. Your team decides.