Also called control chart, X-bar and R chart, MSP
What is statistical process control?
Measuring a process while it runs and acting only when the numbers show something has genuinely changed.
What else it is called
- Other names
- control chart, X-bar and R chart, MSP
- In French
- maîtrise statistique des procédés
Related entries
Understanding statistical process control
Samples are taken as the job runs and plotted on a chart with limits calculated from the process’s own behaviour. While points fall inside those limits and show no pattern, the process is doing what it always does and should be left alone. A point outside them, or a run of points trending one way, means something has changed and is worth investigating.
The hardest idea for a plant to accept is that adjusting in response to ordinary variation makes things worse. An operator who nudges the setting after every slightly high reading is adding their own correction on top of the process’s natural swing, and the output gets wider rather than tighter. Deming demonstrated this repeatedly and it still surprises people.
The second thing worth knowing is that control limits and tolerance limits are different animals. Control limits say what the process does. Tolerance says what the drawing allows. Putting the tolerance on the chart as if it were a control limit is the commonest way these charts get misread.
In practice
If an operator is adjusting the machine several times a shift and quality is still drifting, the adjusting is very likely the cause. Chart it for a week and leave it alone.
Frequently asked
What is the difference between control limits and tolerance limits?
Control limits are calculated from how the process actually behaves and tell you whether it has changed. Tolerance limits come from the drawing and tell you whether the part is acceptable. A process can sit comfortably inside its control limits and still be making scrap.
Do we need software for this?
No. A paper chart at the machine, filled in by the operator, works and has the advantage that the person who can act on it is the person looking at it. Software helps when you want to analyse across machines and shifts.