Control Charts

A traditional control chart is a graphical tool which includes i) a centre line that represents the average value of the quality characteristics under in-control process conditions, and ii) one or two control limits chosen such that, if the process is in-control, only 100 x a% of the measured values of the quality characteristics will fall outside these limits. Every measured value of the quality characteristic is plotted onto the control chart in temporal sequence. An alarm is signalled whenever one limit is violated, but additional alarm rules can be defined to signal non-random patterns in the temporal sequence of quality characteristic values plotted onto the control chart, as such patterns may be symptoms of a change in the process behaviour. Indeed, if the process is in-control, all the plotted points should have a random distribution.

The use of control charts for statistical process monitoring involves two sequential phases, denoted by Phase I, also known as the training phase; and Phase II, also known as the monitoring phase. During the training phase, a dataset gathered under in-control process conditions, and representative of the natural process behaviour, is collected and used to estimate the parameters of the variables to be used for process monitoring and to design the control chart, i.e. choosing the control limits. During the training phase, the collected data can be plotted onto the designed control chart in order to determine whether the process was actually in-control or not during this phase. If out-of-control conditions are signalled and corresponding assignable causes are found by means of a retrospective investigation, an iterative procedure is applied to remove out-of-control data and finally design a control chart that is representative of the actual in-control behaviour of the process. During the following phase, additional measurements of the variable of interest are acquired and their values are plotted onto the previously designed control chart, and an alarm is signalled whenever an out-of-control shift is detected.

< Prev   CONTENTS   Source   Next >