The Xbar-R chart is commonly used for variable data collected in small subgroups. The Xbar chart follows the subgroup averages, while the R chart follows the subgroup ranges. Together, they help separate changes in the process centre from changes in short-term variation.
What is a rational subgroup?
A rational subgroup contains observations produced under conditions that are as similar as practical. The variation within the subgroup should represent short-term common-cause variation. The sequence of subgroup averages should then reveal longer-term changes such as drift, setup changes, material changes, or tool wear.
Examples depend on the process. Five consecutive pieces from a stable cycle may form a useful subgroup. Five values selected across an entire shift may not, because the range now includes time-related changes that the Xbar chart was meant to detect between subgroups.
Why both charts are needed
The Xbar limits depend on the estimate of within-subgroup variation. If the R chart is unstable, the estimate used for the Xbar limits is questionable. Review the range chart first, investigate special-cause variation, and then interpret the subgroup averages.
Control limits versus specification limits
| Control limits | Specification limits |
|---|---|
| Calculated from process data and subgroup structure. | Defined by design, customer, or functional requirements. |
| Used to evaluate process stability. | Used to evaluate conformance and capability. |
| Change when the stable process system changes. | Should not change merely because the process changed. |
A point outside a control limit is a statistical signal, not automatically a nonconforming product. A point inside the control limits can still be outside the specification limits if the stable process is not capable.
How much data is enough?
Initial limits based on only a few subgroups are provisional. Many practical procedures begin with approximately 20 to 25 rational subgroups, but the approved standard and process risk should guide the decision. The important point is to collect enough representative data to estimate the stable process behaviour and to revise limits after removing documented special causes.
Signals beyond one point outside the limits
Run, trend, zone, and mixture rules can detect non-random patterns before a point crosses a limit. The exact rule set should be defined in the control plan or procedure. Applying many rules without a response strategy can create excessive false alarms; applying none can miss meaningful process changes.
Stability before capability
Once the process is reasonably stable and the chart structure is appropriate, capability can be reviewed against the specification limits. Keep the two conclusions separate: one describes predictability, the other describes how the process distribution fits the requirements.
A practical initial-study workflow
- Define the process stream and characteristic.
- Choose the subgroup size and collection frequency based on process knowledge.
- Collect data in true production order.
- Plot Xbar and R charts with provisional limits.
- Investigate and document special-cause signals.
- Recalculate limits from the stable baseline when justified.
- Define the ongoing sampling and reaction plan.
A control chart is a process-management system, not only a graph. The sampling plan, interpretation rules, ownership, and reaction plan determine whether it changes the process.