SPC Control Charts: Principles, Chart Selection and Online Plotting
What is SPC?
SPC (Statistical Process Control) monitors the variation of a quality characteristic through control charts, separating common-cause (normal) variation from special-cause (assignable) variation. A control chart consists of a center line (CL), an upper control limit (UCL) and a lower control limit (LCL) set at +/-3 sigma: when the process is stable, about 99.73% of plotted points fall within the limits. A point beyond the limits or a systematic pattern signals a special cause that needs timely investigation.
How to Choose Among the 7 Charts
First decide the data type. For variable (continuous) data, choose by subgroup size: one value per subgroup (single pieces, destructive tests) uses I-MR; 2-8 values per subgroup use Xbar-R (the most common); more than 8 use Xbar-s. For attribute data, choose by what you count: nonconforming rate with variable sample size uses a P chart; nonconforming count with fixed sample size uses an NP chart; defect count per fixed inspection unit uses a C chart; defect rate per unit with variable unit size uses a U chart. The tool includes a histogram and an Anderson-Darling normality test to help you assess the data distribution.
The Eight Run Rules
The tool automatically applies the Western Electric / Nelson run rules and flags out-of-control points: one point beyond the 3-sigma limits, nine consecutive points on one side, six consecutive points increasing or decreasing, fourteen alternating points, and more. After a signal, stop and confirm the special cause (material, equipment, people, method change), remove it, then recalculate the control limits - do not simply delete the offending points.
How to Use It
After sign-in, paste measurement data, select the chart type, review the control limits and run-rule results, and judge whether the process is in control. Only after the process is judged stable does a CpK analysis make sense - SPC answers whether the process is stable and CpK answers whether it is capable, so use the two tools together.