Search

Abstract: Alarm floods are a crucial problem in the process industry. An alarm flood makes it difficult for an operator to react and take necessary actions, which often can lead to risking an emergency shutdown or a major upset. In many cases, alarm floods are caused by interrelated process variables, which can be identified via similar patterns in alarm annunciations. This similarity can be investigated through alarm pattern analysis of industrial alarm flood data. In this work, alarm floods are discussed based on the standards presented in the new ISA 18.2 guidelines and the discussion given in EEMUA 191. A new analysis method is proposed to identify alarm floods that are similar from the historical alarm data and group them on the basis of patterns of alarm occurrences. Patterns in alarm sequences can be investigated through different distance measures. To calculate a distance between alarm patterns in two different sequences, preprocessing of industrial alarm data and effective flood period isolation are required. Hence, definitions of alarm floods and alarm flood periods are given based on the new ISA 18.2 standards. The effect of chattering alarms on alarm floods is also discussed. Three different distance scores, suitable for capturing alarm patterns in alarm flood sequences, are introduced. Finally, a case study on real industrial alarm data is presented to demonstrate the utility of the proposed analysis.