A control chart helps you decide whether a process is behaving as it normally does or whether something has changed. You plot a statistic, such as the average of a small group of measurements, over time. If the process is stable, almost all points fall between two control limits set around the center line. A point outside them is a signal worth investigating. This calculator computes the limits for an X-bar chart, which tracks subgroup means.
How the calculation works
- Standard error of a subgroup mean = process standard deviation ÷ √n
- Upper control limit = center line + k × standard error
- Lower control limit = center line − k × standard error
- False alarm rate = 2 × (1 − Φ(k)), the chance that a stable process is flagged
- Average run length = 1 ÷ false alarm rate
Example
A process has a mean of 50 and a standard deviation of 2, measured in subgroups of 5. The standard error is 2 ÷ √5 = 0.894, so the 3-sigma limits are 47.32 and 52.68. A subgroup mean of 51.2 is inside the limits. The false alarm rate is 0.27%, or about one point in 370 if the process is stable.
Using the limits
Control limits describe what the process does, not what the customer wants. They are different from specification limits, which come from the requirements of the product. A process can be in control and still produce unacceptable parts, and a process can be out of control while still meeting the specification. Use control limits to detect change, and specification limits to judge capability.
Setting the limits correctly
Estimate the center line and the standard deviation from a period when the process was stable, ideally 20 to 25 subgroups. Do not use the specification limits as control limits, and do not recompute them every time a point looks odd. Recalculate only after a deliberate change to the process, so that the limits keep their meaning.