M‐ATTRIVAR: An attribute‐variable chart to monitor multivariate process means

• Published in: Quality and reliability engineering international Vol. 34; no. 2; pp. 214 - 228
• Main Authors: Aparisi, Francisco, Lee Ho, Linda
• Format: Journal Article
• Language: English
• Published: Bognor Regis Wiley Subscription Services, Inc 01-03-2018
• Subjects: Alarm systems; average run length; Control charts; gauge; Hotelling's T2; Monitoring; multivariate statistical process control; Sampling
• ISSN: 0748-8017; 1099-1638
• DOI: 10.1002/qre.2250

In this paper, an attribute‐variable control chart, namely, M‐ATTRIVAR, is introduced to monitor possible shifts in a vector of means. The monitoring starts using an attribute chart (classifying the units as approved or not using a gauge) and continues in such a way until a warning signal is given, shifting the control to a variable chart for the next sampling. If the variable chart does not confirm the warning, the monitoring returns to an attribute control. Otherwise, the monitoring remains with the variable chart. Whenever any of the charts (attribute or variable) signals an alarm, the control scheme triggers an alarm. The main advantage of this new proposal is the possibility of judging the state of the process only by the attribute chart most of the time (normally more economical and faster). The performance of the M‐ATTRIVAR control chart is compared versus the main competitor (T2 control chart) in terms of performance detection (out‐of‐control average run length) but also economically (average sampling cost). The M‐ATTRIVAR is always cheaper than T2, and in many scenarios, it detects quicker process shifts than the T2 control chart. A numerical example illustrates a practical situation.