Criteria and Algorithms for Online and Offline Detections of Industrial Alarm Floods

Alarm floods are referred to the phenomena that there are too many alarms occurring in a short time period to be promptly handled by industrial plant operators. This paper is on the criteria and algorithms to detect the presence of industrial alarm floods in both online and offline manners. Two basic criteria are analyzed for detecting alarm floods, based on the number of alarm occurrences and the number of alarm variables in the alarm state. Due to chattering alarms and long-standing alarms, the two basic criteria have a drawback of resulting in false detected alarm floods. In order to alleviate the drawback, a new criterion is formulated based on the number of alarm variables newly appeared in the alarm state. An algorithm is proposed based on the new criterion to detect an occurring alarm flood or the presence of alarm floods in historical data sets. Industrial examples are provided to illustrate the obtained results.