Alarm processing for predictive fault diagnosis

This work proposes a methodology based on online processing of alarms with the objective of identifying proactively possible failures of industrial processes operation. For this, a knowledge base that relates failure scenarios previously registered is used then the feasibility of each failure scenario is continuously assessed based on the possibility of activation of alarms that the scenarios in question, with the possibility of occurrence of a particular alarm is calculated from the trend of its associated analog variables. Preliminary results are presented and future work discussed.