Process capability tells you whether a process can consistently produce parts inside the specification limits. A process can be stable and still make out-of-tolerance parts if its natural spread is wider than the tolerance, or if its average is off target. The indices Cp and Cpk summarize this in one number each, and are widely used in manufacturing and quality control.
How the calculation works
- Cp = (USL − LSL) ÷ (6 × standard deviation)
- Cpu = (USL − mean) ÷ (3 × standard deviation), and Cpl = (mean − LSL) ÷ (3 × standard deviation)
- Cpk = the smaller of Cpu and Cpl
- Sigma level = 3 × Cpk
- Parts out of specification = the share of a normal distribution beyond either limit, in parts per million
Example
A shaft diameter must be between 9.5 and 10.5 mm. The process averages 10.02 mm with a standard deviation of 0.12 mm. Cp is 1 ÷ 0.72 = 1.39. The mean sits closer to the upper limit: Cpu = 0.48 ÷ 0.36 = 1.33 and Cpl = 1.44, so Cpk is 1.33, a sigma level of 4.0. About 39 parts per million are expected to fall out of specification.
Reading Cp and Cpk
A large gap between Cp and Cpk means the process is capable but off center, so centering it on the target is the cheapest improvement. If Cp itself is low, the process spread is too wide and needs a reduction in variation: better materials, tooling, maintenance or method. Typical targets are 1.33 for ongoing production, 1.67 for new processes and safety characteristics.
Conditions for a valid result
Capability indices only make sense when the process is stable, meaning in statistical control. Check a control chart first, and use enough measurements, usually at least 30 and ideally 100 or more. If the data are clearly not normal, as with a measurement bounded at zero, the ppm figure can be far off and a non-normal method is better.