Combining knowledge modeling and machine learning for alarm root cause analysis

Industrial alarm systems inform the operator of abnormal plant behavior and are required to guarantee safety, quality, and productivity of the plant. However, modern alarm systems often produce large amounts of false or nuisance alarms which leads to alarm floods. Operators receive far more alarms than they can handle. To reduce these alarm floods, we developped an alarm system that performs Root Cause Analysis (RCA) upon an alarm model constructed with Bayesian networks. In this paper, we present methods to construct Bayesian networks for RCA with a knowledge-based and a machine learning approach. Finally, we evaluated both approaches with an example of an industrial plant and propose an architecture to combine both approaches.