Sampling Strategies to Reduce the Effect of the Autocorrelation on the Xbar Chart's Performance

• Activity Number: 235
• Type: Contributed
• Date/Time: Monday, August 1, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Section on Quality and Productivity
• Author(s): Marcela Machado*+ and Antonio Costa
• Companies: UNESP and UNESP
• Keywords: autocorrelation ; control chart ; sampling strategies

Abstract

In many fast production lines, the neighboring units, regarding to the time they were produced, are highly correlated, that is, the X values of the quality characteristic under surveillance from neighboring items are autocorrelated. According to the literature, the dependence among neighboring items fits pretty well to an autoregressive model AR (1). We adopt this model in our study and we build-up the samples with neighboring items considering the rational subgroup concept. The consequence of the autocorrelation among sample units is a drastic reduction in the Xbar chart's ability in signaling process mean changes. The first sampling strategy to reduce the effect of the autocorrelation consists in building-up the sample by taking one item from the production line and skipping one, two or more before selecting the next. The second strategy consists in working with mixed samples, that is, we build-up the mixed sample with units from the current sample plus items from the former or the formers samples. The main conclusion we take from this study is that the proposed sampling strategies counteract the undesired effect of the autocorrelation.