Process capability is a statistical measure that measures the inherent variability in a characteristic. A process-capability study can be used to determine the process’s ability to meet specifications.
A capability estimate is usually obtained during a quality improvement project, such as Six Sigma. It helps to show the improvement level.
There are many capability estimates that are widely used, including:
- Estimates of process capability include Potential capability (C, p), and actual production capability (C, pk). C, p, and C -pk indicate how capable a process is of meeting the limits of its specification when paired with continuous data. These are useful tools to evaluate the initial and ongoing capabilities of parts and processes.
- “Sigma” is a capability estimation that’s typically used in conjunction with attribute (i.e. with defect rates).
These capacity estimates are essentially used to calculate the process’s nonconformance rate. They can be expressed as a single number. This involves, in general, calculating a ratio between process spread and specification limits.
Understanding Process Capability
It is difficult to assess process capability. Some books tell users to wait for the process to reach equilibrium, then take approximately 30 samples and calculate the standard deviation. However, it can be difficult to determine when equilibrium has been reached and whether the recommended samples are representative. It is much more difficult to measure process capability.
Let’s say, for example, that your rotary tablet press produces 30 tablets. One tablet is made from each of the 30 pockets per rotation. You might be interested in the thickness of your tablet by using the standard deviation from 30 consecutive tablets to determine your process capability. You can even guarantee representation by repeatedly taking the 30 consecutive tablets over eight periods of time spaced equally throughout a production run (Table 1). The eight standard deviations would be combined to give a thickness capability estimate. This is based on (8 X 30 – 1) = 232 degrees freedom.
Capability Suggestions Concerns
Capability estimates have both positive and negative sides. C p and C estimate pk are sensitive to the assumption one is a sampling from normal. This means that most data points are concentrated around the mean (or average), creating a bell-shaped curve.
Additionally, sampling is crucial to obtain meaningful estimates of the process performance for future production.
Many quality practitioners only report the numerical values of capability estimates. However, others point out that capability estimates are statistics or point estimates of the true capabilities of a process. The confidence intervals used to determine the true capability values can also be reported.
Other strategies may be useful when sampling from non-normal but stable distributions to get meaningful capability estimates.
- To transform the data into a model that is approximately well-modeled, use a Normal distribution.
- You can use different probability distributions such as Weibull and lognormal.
