Analysis of Industrial Alarm Floods

This chapter is on the analysis of industrial alarm floodsAlarm flood, mainly involving the detection, similarity analysis, pattern mining, and prediction of alarm floods. New criteria are designed for detecting the presence of alarm floods in both online and offline manners. For offline analysis of industrial alarm floods, a local alignment algorithm for efficient similarity analysis is proposed to find clusters of similar alarm flood sequences, and a pattern mining approach is developed to find interesting alarm patterns in alarm floods from historical alarm data. For online analysis, a maximum-entropy-based method is presented to predict upcoming alarms for an occurring alarm flood. These methods provide effective solutions to improve operators’ awareness to critical alarms in alarm flood periods, reduce irrelevant alarm flood sequences, and achieve early warning and prediction of alarm floods.