Multistage process monitoring using survival analysis regression models

In this paper, we consider the monitoring of a multistage process in which the outgoing quality attribute is a reliability-related characteristic. To measure the tensile strength of such quality characteristics, the process output is commonly inspected under limited load conditions. As a result, we may encounter censored observations which demand a certain monitoring procedure. The monitoring scheme becomes more complicated when the censoring occurs due to variable competing risks as well. To address this critical issue, we propose a regression-adjusted CUSUM chart to effectively monitor a quality characteristic that may be right censored because of both fixed and variable competing risks. The result of a simulation study confirms the superiority of the proposed method in comparison with the existing one. The proposed control chart is applicable not only to reliability measures in manufacturing processes but also to similar quality indices in service operations such as survivability measures in healthcare services.