Run Rules-Based EWMA Charts for Efficient Monitoring of Profile Parameters

Ali Yeganeh,Sandile Charles Shongwe,Saddam Akber Abbasi,Ali Reza Shadman,Ioannis S Triantafyllou

doi:10.1109/access.2021.3061990

Abstract

In usual quality control methods, the quality of a process or product is evaluated by monitoring one or more quality characteristics using their corresponding distributions. However, when the quality characteristic is defined through the relationship between one or more response and independent variables, the regime is referred to as profiles monitoring. In this article, we improve the performance of the Exponentially Weighted Moving Average Range (EWMAR) control charts, which are implemented for monitoring linear profiles (i.e., intercept, slope and average residual between sample and reference lines) by integrating them with run rules in order to quickly detect various magnitudes of shifts in profile parameters. The validation of the proposed control chart is accomplished by examining its performance using the average run length (ARL) criteria. The proposed EWMAR chart with run rules exhibits a much better performance in detecting small and decreasing shifts than the other competing charts. Finally, an example from multivariate manufacturing industry is employed to illustrate the superiority of the EWMAR chart with run rules.

Highlights

Statistical Process Control (SPC) is widely utilized to monitor industrial processes and most of the academic researchers in SPC focus on charting techniques; see the introductory chapters of Montgomery [1] or Chakraborti and Graham [2]
The study used an Exponentially Weighted Moving Average Range (EWMAR) control chart to monitor the linear profiles with a combination of run rules to enhance the performance of the chart in detecting OC conditions
In order to evaluate the performance of the proposed control chart in comparison with other competitive methods including the conventional EWMAR, EWMA3 and etc., the numerical example of Kang and Albin [6] was used in simple linear profiles

Summary

Introduction

Statistical Process Control (SPC) is widely utilized to monitor industrial processes and most of the academic researchers in SPC focus on charting techniques; see the introductory chapters of Montgomery [1] or Chakraborti and Graham [2]. The exact time for the origin of profile monitoring is not known in the literature and some researchers drew on different terms to describe it; for instance, the term "signature" was introduced by Gardner et al [4] while Jin and Shi [5] used the "waveform signals" term in monitoring functional relationships. These terms are equivalent to the profile monitoring method in a way. In various situations and applications, the profiles relationship can be represented by linear [6], multichannel profiles [7], Gaussian process [8], nonlinear [9] and even a complex relationship [10]

Objectives

Results

Conclusion