Problem Of Non-Normality In Statistical Quality Control: A Case Study In A Surface Mine
Organization: The Southern African Institute of Mining and Metallurgy
Jan 1, 2004
A quality characteristic analysed through construction of the Shewhart control chart is required to be normally distributed. In mining applications, this assumption is often violated as many quality characteristics follow skewed or lognormal distribution. In this paper, two methods were investigated using a heuristic approach for setting up the control limits for a skewed population. Based on these approaches a Mode chart, and a Weighted Variance (WV) control charts were constructed. The performance of the proposed methods was compared with the normal based Shewhart method through a case study conducted in a surface mine supplying iron ore to a steel plant. A simulation study was also conducted to evaluate the effectiveness of the proposed methods over the Shewhart method. The study revealed that the Shewhart control charts generated more out of control alarms compared with the Mode and Weighted Variance control charts for a skewed population of Fe%. The simulation results also revealed that for extremely high skewed population, the WV control chart and Mode control chart provided higher probability of coverage than the Shewhart method; although the detection rate (probability of detection) of out of control condition for these methods was slightly reduced compared with Shewhart method.