I. Introduction
Alarm floods are critical situations in large-scale industrial operations and usually caused by various problems, such as the improper alarm system design, major process failures, and state transitions. During alarm flooding, a large number of alarm notifications appear, making it difficult for the industrial operators to identify the critical alarms and take corrective actions in time. As a consequence, the situation may get worse with time and develop to serious accidents. Industrial standards and guidelines [1]–[3] define an alarm flood as a situation where the number of alarms is more than that the operator can effectively manage; more specifically, 10 alarms per 10 min for each operator is used as the threshold to identify an alarm flood. In the past years, plentiful research has been carried out to handle and address alarm floods; especially, the similarity analysis of alarm floods has been proposed and recognized as an effective solution.