Monitoring the covariance matrix of multivariate processes combining attribute and variable data

• Grant number: 17/02376-9
• Support Opportunities: Regular Research Grants
• Start date: November 01, 2017
• End date: April 30, 2020
• Field of knowledge: Engineering - Production Engineering
• Principal Investigator: Marcela Aparecida Guerreiro Machado de Freitas
• Grantee: Marcela Aparecida Guerreiro Machado de Freitas
• Host Institution: Faculdade de Engenharia (FEG). Universidade Estadual Paulista (UNESP). Campus de Guaratinguetá. Guaratinguetá , SP, Brazil
• Associated researchers: Antonio Fernando Branco Costa ; Linda Lee Ho ; Paloma Maria Silva Rocha Rizol

Abstract

Control charts by attributes have been widely used for monitoring univariate processes. Some authors have also studied the monitoring of the mean vector of multivariate processes by attributes. However, there are still few studies focusing at the monitoring by attributes of the covariance matrix of such processes. In this project, we will propose a control chart for monitoring the covariance matrix of multivariate processes combining attribute and variable data. Previous works have shown that the control chart based on the VMAX statistic is faster in detecting shifts in the covariance matrix than the chart based on the generalized sample variance |S|. In addition, the VMAX statistic has a better diagnostic feature, that is, with the VMAX statistic, it is easier to identify the variable that had its variability changed by the occurrence of the assignable cause, and it is easier to calculate than the |S| statistic. The existing VMAX control chart is restricted to variable inspection. Recent works have shown that the strategy of combining inspection by attributes and variables data (the MIX strategy) is economically more advantageous than attribute inspection only. Moreover, a control chart based on the MIX strategy may show performance similar to the control chart based only on inspection by variables. (AU)