Fault detection sensitivity of T2 and MEWMA Control Charts by using local grouping of variables

The drawback of multidimensional control charts is that with increasing number of variables the fault detection sensitivity decreases. A new two-stage hierarchical multidimensional control chart is presented which works with less dimensionality than the number of the monitored variables. The basic idea is not neglecting but grouping some of the less important variables. The effectiveness of the new method is shown by simulating the average run length for different types of faults. The proposed method is illustrated by the fault detection of an industrial CO2 sensor.