Improvement Of Multivariate Statistical Control Model Based On Projection Transformation

As the quality features of the products become more and more divers ified,it is easy to make a wrong judgment just using a univariate statistical control chart for monitoring,so the multivariate statistical control chart arouses people's concern.In this paper,we use projection transformation to propose a multivariate statistical control chart of the variab leM?28?max?|y ₁|,|y ₂|,,|y _p|?,named M control chart.In theory,we compare the confidence region withT² and M control chart for two dimensional case,and find that the M control chart is more sensitive for the variation of single variab le in a particular d irection.We also use Monte Carlo simulation to calculate the ARL1 respectively to testify the previous theoretical results related to the goodness of M control chart.In addition,we also calculate the value ofP(ARL1_T2?29?ARL1 _M)to further verify the better performance of the M control chart.Finally,a real data from auto industry is analyzed to demonstrate the practical significance of the M control chart.