Detection of Anomalies in Compressed Air Systems using Time of Occurrence and KPI Analysis

Compressed air systems are widely used in industry, but they represent an expensive form of energy and are prone to inefficiencies. This paper presents two novel, non-invasive data-driven approaches for anomaly detection in compressed air systems, focusing on analysing pressure drop curves extracted from pressure signals. The methods, based on time anomalies and a Key Performance Indicator (KPI), identify pressure drops occurring at irregular times or exhibiting unusual curve patterns. This study, utilizing eight datasets from various industrial sources, demonstrates how anomalies can be indicated by the presented methods. Our findings show that the time-based method successfully identifies unusual timing patterns, while the KPI-based method faced limitations, as some detected anomalies may be the result of manual interventions or operational changes. These methods provide a cost-effective solution for anomaly detection without additional interruptions or specialized equipment, as data can be collected during normal operation and the analysis can be conducted in real time, improving energy efficiency and sustainability in compressed air systems.