Data Analysis and Related Applications 3

16An Action-based Monitoring Tool for Processes Subject to Multiple Quality Shifts

Two usual key assumptions in the statistical process control (SPC) literature are: (i) the implementation of a process investigation whenever an out-of-control signal occurs and (ii) the process investigation is perfect and defines with certainty the process state. Nevertheless, in many real cases, a thorough process investigation may not be error-free and is not always a cost-efficient option. To this effect, a process monitoring tool, designed under an action-oriented framework, is developed. The proposed model is based on Markov chain theory and considers processes subject to multiple assignable causes. A benchmark of examples has been generated to demonstrate the performance of the monitoring scheme and a comparison to simpler approaches is presented.

16.1. Introduction

In the manufacturing sector, the need for viable, acceptable and stable levels of quality, aligned with the modern quality standards requirements, is ever-increasing, thereby, necessitating new, effective approaches to fault diagnosis. It is well known that many intangible and poor-quality costs, including inter alias, inventory, lost production and personnel costs, are hidden in unplanned downtime, and their elimination is a milestone commitment and challenge for quality (Woodall and Faltin 2019). The use of appropriate machine performance metrics to develop statistical models could eliminate the unplanned downtime and ...