Μη παραμετρικά διαγράμματα ελέγχου και εφαρμογές

Abstract

Statistical Quality Control is a very important part of the production of products and services. Part of this is the Statistical Process Control which uses various tools to monitor a process, one of them and the subject of study of this dissertation is the Control Charts. More specifically, the Non-Parametric Control Charts. Classical control charts (e.g. Shewhart type control chart, EWMA, CUSUM etc) can be applied to data that follow a specific distribution and are quite efficient in identifying a medium or large shift in some aspect of the production process. However, when the data do not follow a specific distribution or their distribution is unknown, the only solution for monitoring the process is to construct non-parametric control charts. One interesting feature of the non-parametric control charts is that they are quite sensitive and can detect even small shifts in a production process. In this dissertation we shall first present an introduction to the Statistical Quality Control and Statistical Process Control, a brief historical flashback to quality, the definition of non-parametric control charts as well as the advantages and disadvantages. Next, we shall present some of the most important non-parametric methodologies of non-parametric control charts as well as a comparison between some of them. At the same time, we shall proceed to a brief presentation and comparison of four new non-parametric control charts methodologies that have appeared in recent literature using code in the programming language R. A brief bibliographical presentation will be made to multivariate control charts as well as to two multivariate control charts for simultaneous monitoring of the medium and dispersion. Finally, we shall discuss about areas where the non-parametric control charts can be applied such as medicine, frequency of sequential events and 3D printing.