AN INVESTIGATION OF QUALITY CONTROL CHARTS FOR AUTOCORRELATED DATA

An application of the Shewhart control charts for quality monitoring and control requires an assumption that observations are independent and normally distributed. An assumption of independence of quality related data in mining operations is questionable, as autocorrelation amongst the observations becomes an inherent characteristic in mineral deposits where ore grades are spatially distributed. This phenomenon led to an examination of other types of control charts namely modified Shewhart chart, special cause control chart, and common cause control chart to capture the autocorrelation among observations while constructing control charts. An investigation of these charts was conducted in an iron ore mine and the behaviour of the charts was studied on three quality characteristics namely, Fe%, SiO2% and Al2O3%. The results suggest that the serial correlation of the observations has substantial effect on the performance of the conventional Shewhart chart. The effectiveness of the control charts was compared using the sliding simulation approach. It was revealed that the modified Shewhart chart and the special cause control chart provided a higher probability of coverage than the conventional Shewhart chart. Therefore, it was inferred that the conventional Shewhart chart generated false alarm of out of control situation, which in turn revealed that the modified Shewhart chart and special cause control chart are more appropriate in a correlated environment. For the case study mine, it was also revealed that the modified Shewhart chart and special cause control chart behaved in a similar way.