Semantic correlation for alarms classification in maintenance process

Abstract

In the scope of maintenance process, alarms triggering and management is an important issue that diagnosis assistance system must take into account. Alarms floods have always been serious risks in monitoring problem, where is overwhelm operators with large amount of alarm messages and different parameters in a short period of time. Consequently, without a useful alarm’s management, diagnosticate the alarm on fly become a very hard task for maintenance operators. In order to improve alarms’ management, we elaborate in this work an approach allowing the classification of alarms according to the failed equipment and using past event traces. In this context, we studied various methodologies of problem discovering and classification. Hence, we provide in this work a comparative analysis of Meta-heuristic optimization techniques using knowledge capitalization extracted thanks to an ontology. The proposed approach has been tested on a real industrial data. The obtained results are encouraging.