Monitoring and prioritising alerts for exception analytics

Abstract

Statistical quality control based approaches provide tools such as control charts for monitoring process parameters (usually ‘physical layer’) as well as rules such as the Western Electric Rules for detecting anomalies. The design of these charts is based on the statistical properties of the monitored parameters and the anomaly notification is related to the statistical significance of the observed change, the practical significance of the change, i.e. the sensitivity of system to the deviation in the parameter values, is not captured. Monitoring and alerting in the ‘planning layers’, however, is driven by the practical significance of the detected change as against the statistical significance. Recent developments in the field of supply chain event management have been focusing on providing tools and solutions for addressing these needs of the planning layers – these studies have, to a large extent been ad hoc. In this paper we propose and present parametric charts – a post-optimality-based approach for the design and implementation of monitoring and prioritising strategies for logistical anomalies. Analogous to the quality control charts, the parametric charts have four automatically generated regions, namely, the insensitive, the perturbation, the re-optimisation, and the infeasible, with each region having specific priorities and alarm levels associated with them. The bounds for the regions are based on the practical significance of the detected change as against the statistical significance of the deviation. The proposed approach offers three benefits: (1) it removes the ad hoc-ness associated with the determination of the practical significance as well as the priority given to the different classes of impacts that the system user observes, (2) the time consuming process of entering and fine-tuning the alarm bands and the priorities associated with the different alarm types when the exception analytics systems are installed can be decreased and automated, and (3) it aligns the exception analytics with the specific plans that are being executed. The step-by-step implementation of the charting approach is elucidated through a case study for the production unit at the FAME lab, at the Pennsylvania State University.

Keywords:

• exception analytics
• alerts
• monitoring
• prioritisation
• control charts
• practical significance