Data and settings
Chart
Range chart
Signals
Capability
What the checks found
The checks look at the data and the settings. They cannot tell whether the measurements were taken properly, whether the gauge is good enough, or whether the points are really in time order.
What a control chart tells you
A control chart separates ordinary variation, which the process produces all the time, from signals, which point to something that changed. The limits are not specification limits. They come from the process's own variation in the baseline you chose. A point inside the limits can still be out of specification, and a point outside them can still be inside it.
How the limits are worked out
- Individuals and moving range: centre is the baseline mean. Limits are the mean plus or minus 2.66 times the average moving range (the gap between each point and the one before). The moving range chart has an upper limit of 3.267 times that average.
- X-bar and range: centre is the mean of the baseline subgroup means. Limits are that mean plus or minus A2 times the average range. The range chart has limits of D4 and D3 times the average range. A2, D3 and D4 are the standard constants for the subgroup size, 2 to 10.
- Sigma for the zones is the distance from the centre line to the upper limit, divided by 3.
- Limits use the baseline only. Later points are judged against them, so a shift after the baseline shows up instead of moving the limits.
The signals it looks for
- Rule 1: a point beyond a control limit.
- Rule 2: nine points in a row on the same side of the centre line.
- Rule 3: six points in a row all rising or all falling.
- Rule 5: two of three points in a row beyond 2 sigma on the same side.
- Rule 6: four of five points in a row beyond 1 sigma on the same side.
- On the range chart only rule 1 is used.
- A signal is marked on the point that completes the pattern. Signals are a prompt to look for a cause. They are not proof of one. About 1 point in 370 falls outside the limits on a stable process by chance, and using several rules raises the overall false alarm rate.
Capability, and why it can be withheld
Cp and Cpk use the short-term spread (average range or moving range divided by d2). Pp and Ppk use the spread of all baseline values. They compare the process with your specification limits. They only mean something when the process is stable, so the tool shows them only when the baseline has no signals, has at least 20 points, and there are valid specification limits. They also assume the values are roughly bell shaped, which the tool does not test. Customer targets differ, so check yours. A good capability figure says nothing about the points after the baseline.
Taking points out of the limit calculation
If a point has a known cause that will not recur (a wrong part loaded, a gauge dropped), you can leave it out of the limit calculation by entering its number and the reason. It stays on the chart as a hollow marker. Do not leave points out only because they look odd. The tool asks for the reason and flags it if it is missing.
Related tools
Use this for the Control phase of the DMAIC project board. Find causes with the 5 Whys tool. Track fixes in the Corrective Action Tracker. See where line time is lost in the OEE and Downtime Log.
Limits
No gaps found does not mean the process is in control or the data is right. This is decision support only.