Abstract:
Individual and Moving Range (I-MR) charts are generally used for process
monitoring where we cannot group observations into rational subgroups or
when it‟s more convenient to observe individual measurement rather than
averages of the subgroups. However, the efficiency of I-MR charts is poor
when the underlying distribution of data deviates from normality. Process
monitoring is very important in rubber manufacturing where data on critical
to quality distributes non-normally. No previous work was found applying I MR charts especially on rubber data which are non-normal in nature. Hence
this study was carried out to investigate the effect of non-normal data on I MR charts and to develop a method to construct I-MR charts for non-normal
data. We suggest to construct I-MR charts with Jonson transformed data as a
solution for this issue. Performances of I-MR chart were compared with the
theoretical standards, under four different cases; simulated non-normal data,
real data, Johnson transformed simulated data and Jonson transformed real
data. While the simulated non-normal data and real data lead to high Type-I
error and low power, Johnson transformed data lead to very low Type I error
(<0.001) and power comparable to theoretical standards. Further
investigations are in progress with the objective of recommending this
methodology for process monitoring in rubber manufacturing.