Towards alarm flood reduction

Abstract

Alarm systems play critically important role for the safe and efficient operation of modern industrial plants. However, most existing industrial alarm systems suffer from poor performance due to too many alarms needing to be handled by operators in control rooms. This paper proposes a method to reduce the alarm flood by detecting redundant alarms so that they can be filtered later before being transmitted to operators. To do that, an approach based on pattern mining is selected. That method is then applied on an actual dataset coming from a General Electric power plant. The results show that removing redundant alarms allow significantly reducing alarm flood, without loss of efficiency nor safety.