Optimizing processes based on censored data obtained in repetitious experiments using grey prediction

Abstract

The design of experiment (DOE) has been extensively adopted to increase the efficiency of designing new products and developing manufacturing processes in industry. However, some designed experiments cannot be completed for some uncontrollable reasons, such as cost and time restrictions or power damage during the experiment. Under such circumstances, incomplete data obtained in the experiment are referred to as censored data. Conventional approaches to analyzing censored data are computationally complex and frequently depend on assumptions of the normality of data. This study presents a procedure for analyzing the censored data obtained in repetitious experiments using the grey system theory. The proposed procedure does not make any statistical assumption and is less conceptual and computationally complex than current methods. Two experiments - one conventional experiment with type II censoring and one Taguchi experiment with type I censoring - are performed to demonstrate the effectiveness of the proposed procedure.