Briefly about control charts

The control chart is a proven tool to continuously monitor the stability of a process that produces anything that is measurable. The purpose of the control chart is 1) to alarm as early as possible if the process starts to behave irregularly or unpredictably, and 2) to avoid false alarms that disrupt production. Prompted by a control chart alarm the process operator should immediately investigate the process, diagnose the root cause of the problem, and try to fix it. Then the operator resumes the process and continues with the control charting. The goal is to stepwise improve the process until it works reliably and predictably, and within specification. Diligently done, control charting saves money and boosts customer satisfaction.

A basic Shewhart type control chart for individual data points may look like this:

A Shewhart control chart for individual data points showing control limits and one out-of-statistical-control data point

This I chart (I for individual data points) is a graph where the individual data points are plotted one by one as they are sampled from the monito­red process. The chart has three horizontal lines; one center line (green) around which two control limits, the upper control limit and the lower cont­rol limit (red lines), are symmetrically spaced. In the case that a newly added data point falls outside of a control limit, the control chart sounds the alarm (the red data point in the above image) alerting the operator that the process doesn’t behave predictably anymore. The distance between the center line and the two control lines is calculated from the actual variation of the data points ensuring that the control chart is well balanced, which means that the control chart should be enough sensitive to detect unpredictable process behavior, and still not trigger too many false alarms.

A process that for some time has behaved predictably, as indicated by an absence of control chart alarms, is said to run under statistical control.

ControlFreak List of Features

Control Charts: Individuals (I) Control Charts (Shewhart) for location of individual values
Moving Range (MR) Control charts for scale (variation) of individual values
J (Zone) Control Chart for location
Cusum control chart for location
Cusum control chart for scale
Cusum control chart with overlaid I chart
Self-starting Cusum control chart for location
Self-starting Cusum control chart for scale
EWMA control chart for location
EWMA control chart for scale (M-EWMV chart)
EWMA control chart with overlaid I chart

The QuickChart button for quick and dirty control charting button to make a quick and dirty Shewhart control chart

ARL calculations: For I and MR control charts
For Cusum control charts
For EWMA control charts
Versatile data point selection: Control charted data points
Data points used for statistical calculations
Robustification of data: By trimming
By winsorization
Box-Cox transform: Transforming non-normal data
Detecting of outliers: According to Dixon
According to Grubbs
According to Rosner
According to Iglewicz and Hoaglin
Goodness-of-fit tests: According to Kolmogorov-Smirnov
According to Anderson-Darling
According to Shapiro-Wilk
According to Jarque-Bera
According to Lilliefors-van Soest
According to Cramér-von Mises
According to Ryan-Joiner
According to d'Agostino-Pearson
Capability Indices: 14 different capability indices
Point estimates, confidence intervals, lower confidence bounds
Two different stability indices
Statistical plots: Histogram
Normal probability plot
Scatterplot
Box-and-Whisker plot
"Lag 1" plot for correlation
Autocorrelation plot
Pearson distribution diagram
Graphs illustrating the Box-Cox transform (if performed)
Reports: Distribution
Statistical point estimates
Process capability indices
Non-conforming output