In high-quality processes, since defective rates can be very low it may not be possible to adopt normal distribution for the numbers of defectives per sample. In such situations time-between-event (TBE) charts have been found as more appropriate. Various types of TBE charts such as the cumulative count of conforming (CCC) Chart, the cumulative quantity control (CQC) chart, the exponential cumulative sum (CUSUM) chart and the exponential exponentially weighted moving average (EWMA) chart are developed and used (Ref. 1). The assumptions for these charts are Poisson occurrences of defectives and the time between occurrences following exponential distribution. Some of the above charts are found to be robust to the deviations from exponential TBE. These charts can be used in univariate situations. However, in practice, there may be more than one variable of interest and the variables may not be independent to use separate univariate charts. This paper proposes two TBE MEWMA charts based on Gumbel's bivariate TBE models (Gumbel Ref. 2) which are suitable for non-normal multivariate situations. (43 refs.)
展开▼
