Statistical process control: watching a dimension part after part
Two caps from the same mould are never quite identical. The question is not to eliminate the difference, but to know whether it is still the one expected. That is what statistical process control measures.
By Tecmom editorial team · 2 min read
Two kinds of variation
All production varies. Part of this variation is permanent and small: slight fluctuations in temperature, material, pressure. It is called common-cause variation. The other kind appears suddenly and has a cause that can be found: a cavity wearing, a setting changed, a different batch of material. That is special-cause variation.
Statistical process control (SPC) is used to tell the two apart. It goes back to Walter Shewhart's work in the 1920s; the ISO 7870-2 standard, revised in 2023, describes its central tool, the control chart.
The control chart
A few parts are taken at regular intervals, a dimension is measured, and the average is plotted on a chart. A centre line and two control limits, usually placed three standard deviations on either side, frame the normal behaviour of the process.
A point outside the limits, or a run of points drifting in the same direction, signals a special cause: it is investigated and corrected before parts go out of tolerance. Control limits come from the process itself; they are not the same as the drawing's tolerances.
Capability, or the room the tolerance leaves
A stable process can still be too scattered for the required tolerance. Capability compares the two. The Cp index relates the tolerance interval to six standard deviations of the process; the Cpk index also accounts for centring, by looking at the limit closest to the mean.
A high Cpk means production holds the tolerance with a margin. Many customers set a minimum value in their specifications.
On an injection-moulded part
In injection moulding, SPC covers the dimensions that determine function: diameter and height of a thread, thickness of a tamper-evident band, part weight. In a multi-cavity mould, each cavity may have its own mean: they are often tracked separately.
It complements the customer's incoming inspection, which judges a finished batch; see AQL. And it relies on the FMEA, which identifies the dimensions to monitor.
Statistical control does not promise identical parts. It promises to know, at every moment, whether production is doing what is expected of it.
Glossary
- SPC
- Statistical process control: monitoring production with statistical methods.
- Control chart
- A chart that tracks a measurement over time between two control limits.
- Capability
- A process's ability to hold a tolerance, measured by Cp and Cpk.
- Standard deviation
- A measure of how spread out values are around their mean.