AI-based real-time anomaly detection in industrial engineering: A structured literature review, taxonomy, and research agenda

• Literature on real-time anomaly detection consists of idiosyncratic approaches. • This research synthesizes data from 91 publications across fields. • Findings are systematized in a taxonomy with 13 dimensions and 42 characteristics. • Findings provide starting points to develop theory and implementations. • Three focus points for future research are identified. Smart sensor technology enables capturing and communicating real-time data. Resulting data streams represent the fibers that responsive, interconnected industrial engineering networks, such as the Internet of Things, are made of. Establishing these real-time networks is no end in itself but requires advanced data analytics capabilities to leverage valuable insights. In this regard, real-time detection of anomalies has gained momentum in various applications as it enables timely reactions to emerging situations. However, literature on real-time anomaly detection is scattered over applications and mostly engages with idiosyncratic problems. In response, we have conducted a structured literature review. We have analyzed 90 publications from Information Systems and related fields and integrated these into a taxonomy that is useful for practice and academia. Based on our findings, we identified a future research agenda to gather starting points and guide further development of the field. Our research contributes to structuring decision alternatives for anomaly detection systems. The results can be used as guidelines to develop or customize anomaly detection software.