The use of experimental design in the control of continuous processes

Abstract

This dissertation examines a new method of controlling sources of variance in the area of continuous production via the use of control charts based upon operating characteristic curves developed from the following experimental designs: the Completely Random Two-way Classification, the Latin Square, the Graeco Latin Square, and the Hierarchical. The term "continuous process" as used here-in refers to those daily production line functions required to manufacture a product "steadily" (as opposed to batch and lot/time period production) given all the needed, and previously inspected, raw materials. The models are developed and the theory of control charts based upon them is proposed. Through simulation, on a computer, probability of acceptance curves are produced and plotted to show each design's ability to estimate the variances of interest in the production process. Finally, a discussion of the applications of these control charts along with a summary of this work and suggestions for continued research are presented.