If you have access to OneMine as part of a member benefit, log in through your member association website for a seamless user experience.
|The Shewhart [X]?and R charts are still the most popular control charts in use today for continuous monitoring of quality data. However, for these charts to function properly, the underlying assumption of normality and independence of data must be satisfied. When either independence and/or normality of data are not present, which is a common feature of ore quality data, an application of the conventional Shewhart [X]?and R charts may introduce false alarms in the analysis of the data. To address these issues, a guideline is proposed for construction of appropriate univariate and multivariate control charts for a variety of situations. A use of this guideline to a bauxite mine suggested the construction of special cause control chart for the analysis of individual variables namely, Al2O3% and SiO2%; whereas, the multivariate T2 chart based on residuals is proposed for the bivariate analysis of the variables. The case study clearly revealed that without considering the data correlation, one may be in a state of false impression about an out of control condition while using the Shewhart [X]?and R charts for univariate and the Hotelling T2chart for multivariate analysis. To compare the effectiveness of the control charts, specifically the Shewhart [X]?and the special cause control charts, a simulation study was conducted. It was found that the conventional Shewhart chart provided lower probability of coverage than the special cause control chart. It was also revealed that the quality specification of Al2O3% is met as the variation of Al2O3% is well within the specification limits; whereas, it is difficult to meet the quality specification of SiO2% on a regular basis.|