During plants’ operation a huge amount of data is generated and recorded, including alarm data. In general, this data is only used for warranty and traceability purposes.
Due to the increasing applicability of Big Data, Cyber-Physical Systems (CPS), Industrie 4.0 and Data Mining methods, it is increasingly possible to identify patterns within industrial data. One use case is the automatic identification of alarm floods by analyzing historical alarm logs, including recorded alarm data, to obtain information on how to increase the capacity of alarm management systems (i.e. by reducing alarm floods).
At the Institute of Automation and Information Systems, an algorithm has been developed, which allows automatic identification of alarm floods. This algorithm has been evaluated on 12 different machines of the manufacturing and process industry. The alarm flood identification has the capability to identify recurring patterns by consolidating causal dependent alarms within the alarm logs, based on statistical methods. The algorithm identified causal dependencies within the historical alarm logs and aided in identifying the reasons for the alarm floods. Furthermore, it revealed that mainly design mistakes are responsible for alarm floods as well as functional dependent actuators and sensors. These results have been evaluated by the participating companies and have been utilized to re-engineer their alarm management systems.