A Study on apllying statistical modelling techniques improvement and development

ABSTRACT The experiment design, which is one of the techniques of quality improvement and development has gained a great interest in the industrial area recently. The factors which have an important effect on the response variable which is a selected quality characteristic of a product are determined in the experiment design, and at what levels the factors in question will be determined is researched in order to obtain the desired value of this quality characteristic. Some of these factors (z1,z2,..., zq) are uncontrollable although some of them are controllable \xx,x2,...,xp). It is possible to take control of some the uncontrollable factors depending on the aim of experiment and the skills of the factory. The methods of experiment design are carried out to develop a process affected from the external variable ( from z's ) at a minimum level. Modelling techniques suggested by many scientists and modelling the mean and the variance of the product or process quality characteristics seperately were focused on in this study. There have been different approaches in modelling the mean and variance which are the response variable of the performance criterions, and the design and noise factors which affect them. In addition to these approaches for modelling, a new modelling approach was presented. In this approach, that a = yQ + y1zi model and noise factors have a direct effect on variance was supposed, and a new model was obtained by taking Y = Xfp +ae as a base. However, the method applied by Pledger was taken as a base in modelling the mean and variance for the response variable in the new model. In this method, the theorem used especially in modelling the variance is striking. Vll T.C. YÜKSEKÖ?RETİM KURDU» DOKÜMANTASYON MEBKGEt