| A Shewhart control chart is a tool used in statistical process control to measure the performance of the process and to give a signal for improvement. The normality assumption and the independence of observations are two important limitations. These assumptions may not be applied to many manufacturing processes. Recent research introduces the bootstrap resampling method to manage those situations with nonnormal or correlated data.; This research studies the violation of both assumptions in three proposed conditions: in-control process, out-of-control process, and process with autocorrelated data. Design of experiment (DOE) is used to discover key factors influencing the suitability of bootstrap and Shewhart charts in violation of these assumptions.; Experimental results confirm that the bootstrap control chart is superior to the Shewhart control chart under particular circumstances. Coefficient of skewness and subgroup size are the most common significant factors for both in-control and out-of-control processes. The magnitude of the process shift also affects control chart selection. The Moving Block Bootstrap, suggested for any process with correlated data, is not a good alternative due to the difficulty in selecting the optimal block length in each specific sample size and its failure to manage the easiest case (stationary process).; A bootstrap control chart is not feasible in practice because we should know our process very well, including such details as the kind of process distribution and the size of the subgroups. It is difficult to collect all this information in order to apply the bootstrap control chart to a process. Therefore, the bootstrap control chart has limited use; implying that it would be well understood with a limited number of possibilities for assignable causes. As a result, the bootstrap control chart is the superior choice only under a limited number of very specific conditions. |
