Most quality control and quality improvement procedures involve making assumptions about the distributional form of data it uses; usually that the data is normally distributed. It is common place to find processes that generate data which is non-normally distributed, e.g. Weibull, logistic or mixture data is increasingly encountered. Any method that seeks to avoid the use of transformation for non-normal data requires techniques for identification of the appropriate distributions. In cases where the appropriate distributions are known it is often intractable to implement. This research is concerned with statistical process control (SPC), where SPC can be apply for variable and attribute data. The objective of SPC is to control a process in an ideal situation with respect to a particular product specification. One of the several measurement tools of SPC is control chart. This research is mainly concerned with control chart which monitors process and quality improvement. We believe, it is a useful process monitoring technique when a source of variability is present. Here, control charts provides a signal that the process must be investigated. In general, Shewhart control charts assume that the data follows normal distribution. Hence, most of SPC techniques have been derived and constructed using the concept of quality which depends on normal distribution. In reality, often the set of data such as, chemical process data and lifetimes data, etc. are not normal. So when a control chart is constructed for x or R, assuming that the data is normal, if in reality, the data is nonnormal, then it will provide an inaccurate results. Schilling and Nelson has (1976) investigated under the central limit theory, the effect of non-normality on charts and concluded that the non-normality is usually not a problem for subgroup sizes of four or more. However, for smaller subgroup sizes, and especially for individual measurements, non-normality can be serious problem. The literature review indicates that there are real problems in dealing with statistical process control for non-normal distributions and mixture distributions. This thesis provides a quantile approach to deal with non-normal distributions, in order to construct median rankit control chart. Here, the quantile approach will also be used to calculate process capability index, average run length (ARL), multivariate control chart and control chart for mixture distribution for non-normal situations. This methodology can be easily adopted by the practitioner of statistical process control.
Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:325933 |
Date | January 2000 |
Creators | Arif, Osama Hasan |
Contributors | Kanji, Gopal G. ; Gilchrist, Warren |
Publisher | Sheffield Hallam University |
Source Sets | Ethos UK |
Detected Language | English |
Type | Electronic Thesis or Dissertation |
Source | http://shura.shu.ac.uk/19285/ |
Page generated in 0.0017 seconds