A Thorough Comparison of the Variable-Sample-Size Weighted-Loss CUSUM and ABS-SPRT Control Charts

It is frequently documented that concurrent shifts in the mean and dispersion of a process quality characteristic carry serious consequences to a manufacturing enterprise. It is, therefore, crucial to consider the detection ability of a control scheme when monitoring simultaneous changes in the process mean and dispersion. In this article, we conduct a thorough study on the detection performances of two dynamic control schemes, i.e., the variable-sample-size weighted-loss cumulative sum (VSS WLC) chart and the absolute-value sequential probability ratio test (ABS-SPRT) chart. This article reveals that the optimal ABS-SPRT chart outperforms the optimal VSS WLC chart in terms of the average extra quadratic loss and the average time to signal over a range of shift sizes. For small process shifts, the optimal VSS WLC chart is a slightly better performer than the optimal ABSSPRT chart in terms of the average number of observations to signal. However, for moderate and large process shifts, the optimal ABS-SPRT chart remains the most powerful scheme in all aspects. The ABS-SPRT chart is also favored due to its smaller long-run expected sample size compared to the VSS WLC chart, making it extremely promising in many industrial applications.