Search

Control charts are widely used to monitor the quality of industrial processes. It is quite common to assume that the random variable associated to the quality characteristic has a Normal distribution, and the control limits are defined so that the probability of obtaining a false alarm is 0.0027. However, the quality characteristic could follow a different distribution in practice, and this fact could have an impact on the efficiency of the control chart. In this paper, a Monte Carlo simulation study is carried out to evaluate empirically the impact of the lack of the normality assumption on the control chart for the mean. Different probabilistic distributions are considered. In addition, under control and out of control processes are considered. The results derived from the simulation study suggest that control charts are an effective tool when the distribution of the quality characteristic is slightly asymmetric. However, a large number of samples or larger sample sizes are required to obtain similar results to the case of symmetric distributions. In the case of asymmetric distributions, it is necessary to increase the sample sizes to obtain acceptable results. Finally, control charts are not recommended under evident cases of non-normality. [ABSTRACT FROM AUTHOR]