Um estudo do desempenho dos gráficos de controle por grupos na presença de autocorrelação no processo

Abstract

Graphs of statistical process control has, among others, in order to monitor the variability in processes whose products derive from only one production flow. However, there are processes characterized to contain multiple streams, so-calledd parallel processes. The control charts for groups (gg) allow control of various production flows through a single pair of graphs. Another key feature for the construction of control charts is the assumption of no autocorrelation in the monitored characteristic. Therefore, the main objective of this article is to analyze the performance of gg or not when you consider the presence of autocorrelation in the observations, as the detection of process disturbances and compare them for performance. Was used as the research methodology, the simulation technique to obtain the results, simulating different scenarios of production. This work showed that the use of gg without considering the autocorrelation, when in reality it is present in the process directly affects the performance of gg, since the control limits are narrower and thus there is a higher incidence of points outside the control limits.