Setting and revising control limits
Phase I and Phase II, how much data to use, when to recalculate, and handling out-of-control baseline data.
Phase I: establish a baseline
- Collect 20 to 25 subgroups (or about 100 individual readings) over typical conditions.
- Calculate trial limits and look for special causes.
- For each signal with a known, fixed cause, remove the point and recalculate. Keep points with no identified cause.
- Repeat until the baseline is stable.
Phase II: monitor
Lock the limits and plot new data against them. Do not recalculate limits every week with the newest data: a slowly drifting process would carry its limits along and never signal.
When to recalculate
- After a deliberate, verified process change (new tooling, new method). Start a new phase.
- When the chart shows a sustained shift that you have investigated and accepted as the new normal.
- After a long period, to confirm that the limits still reflect the process (for example yearly).
Phases on one chart
Showing before and after a change as separate phases, each with its own limits, makes the effect of an improvement visible.
Questions and answers
Can I use limits from another, similar machine?
Only as a temporary start. Each machine has its own variation. Replace borrowed limits once you have enough data.
Our customer gave us control limits. Should we use them?
Ask how they were calculated. Limits must come from your own process data. Customer-required capability targets belong in the capability study, not on the chart.
