Controle estatístico multivariado de processos através do gráfico T2 de Hotelling: um estudo de caso no setor de embalagens

Abstract

Multivariate analysis has been used lately as organizations have modern equipment that creates correlated data and requires simultaneous monitoring. One of the areas that deals with such problems is determined by multivariate quality control. The present work it’s a proposal to introduce in a process a technique of statistical control of multivariate process (Graph T2 of Hotelling) which was applied in a productive process of plastic packaging. The variables studied were: Maximum Thickness, Minimum Thickness and Weight. With the support of bibliographical review, the work used the technique proposed by Hoteling to collect the data, analyze the productive process and identify the variables causing instability. For data processing and analysis were used the software SPSS, and Minitab, in addition to the worksheet DEC_Hotelling. The results showed that the studied variables (thickness and weight) are strongly correlated ad therefore suitable for multivariate analysis. Through the statistic T2 of Hotelling was possible to establish appropriate control limits for the process, taking a point out of control by means of its decomposition. It was also verified that when the variables are under control in the univariate, this does not mean that the process is multivariate is under control. Finally, pulled of the point out of control, were established new limits and, with the new sample data entered, has been verified a stable process, under multivariate control.