Multivariate statistical control charts for process dispersion

In multivariate statistical control of process dispersion usually used generalized variance control chart, but its sensitivity is often insufficient to detect a small increase of process dispersion. To increase the sensitivity of control proposed to use cumulative sum and exponentially weighted moving average charts for the generalized variance. The choice of type of a chart, the most effective for concrete process, is defined by results of statistical tests. The data flow of technological process on the basis of multidimensional normal distribution with a vector of averages and a covariate matrix corresponding to investigated process is generated. It is thus modeled dispersion increases. The obtained data are investigated by means of various types of charts of multidimensional dispersion: the average run length of series for each type of charts is estimated on set simulated samples.