Identification and clustering of alarm floods in a natural gas processing plant

Abstract

Alarm flood is a serious challenge in industrial processes. During an alarm flood, an operator receives a large number of alarms in a short period of time. As a result, in most cases the operator is unable to rectify critical situations using alarm information. In this paper, alarm floods in a natural gas processing plant (gas refinery) are identified using historical alarm data. Then, alarm floods are clustered, according to the Euclidean distance between them and based on a hierarchical algorithm. Finally, the dynamic time warping algorithm is used to extract patterns in each cluster. This analysis of alarm floods is useful for determining the dependency between alarms during a flood, and also finding patterns within the sequence of alarms. The patterns can be used for better management and improvement of alarm configuration.