On the Use of Inductive Biases for Semantic Characterization of Industrial Alarm Systems

Process Automation Systems (PAS) have a great benefit when including Probabilistic Web Ontology Language (PR-OWL) in intelligent systems and use automated reasoning of the underlying process data. It appears as a strategic element of modern Distributed Digital Control Systems (DCS). Inside industrial processes, the analysis of alarms in the process control rooms is not an easy operator task. Given the lack of an effective analytical formulation that is capable to provide adequate support to plant operators in their real-time decision-making processes by making explicit critical conditions using industrial alarm systems. Due to specificities for industrial alarm systems and the increasing number of alarms that operators must handle, the main challenge is assessing the meaning and relevance of event patterns during different situations. In this paper, a probabilistic model, which further specializes information flows in accordance with the environment adopted for control and monitoring systems of the discrete industrial domain, seems to be more effective in this task. Further, the proposed approach was evaluated in a real case study scenario.