Abstract
Monitoring complex production systems is primordial to ensure management, reliability and safety as well as maintaining the desired product quality. Early detection of emergent abnormal behaviour in monitored systems allows pre-emptive action to prevent more serious consequences, to improve system operations and to reduce manufacturing and/or service costs. This study reports the design of a new methodology for the detection of abnormal situations based on the integration of time-series analysis models and statistical process control (SPC) tools for the joint development of a monitoring system to help supervising of the behaviour of emergency department services (EDs). The monitoring system developed is able to provide early alerts in the event of abnormal situations. The seasonal autoregressive moving average (SARMA)-based exponentially weighted moving average (EWMA) anomaly detection scheme proposed was successfully applied to the practical data collected from the database of the paediatric emergency department (PED) at Lille regional hospital centre, France. The method developed utilizes SARMA as a modelling framework and EWMA for anomaly detection. The EWMA control chart is applied to the uncorrelated residuals obtained from the SARMA model. The detection results of the EWMA chart are compared with two other commonly applied residual-based tests: a Shewhart individuals chart and a Cumulative Sum (CUSUM) control chart.
Original language | English (US) |
---|---|
Pages (from-to) | 2102-2114 |
Number of pages | 13 |
Journal | Neurocomputing |
Volume | 173 |
DOIs | |
State | Published - Oct 25 2015 |
Bibliographical note
KAUST Repository Item: Exported on 2020-10-01ASJC Scopus subject areas
- Artificial Intelligence
- Cognitive Neuroscience
- Computer Science Applications