Skip to Main content Skip to Navigation
New interface
Conference papers

Model Transformation from CBM to EPL Rules to Detect Failure Symptoms

Abstract : The increasing complexity of modern systems, cost reduction policies and ever increasing safety requirements are bringing new challenges to the maintenance domain. In many fields, periodic maintenance actions become either insufficient or too expensive. In this context, Condition-Based Maintenance (CBM) strategies, and Prognostics and Health Management (PHM) in particular, are offering an interesting alternative by allowing systems to be maintained only when needed. These strategies rely on a constant monitoring and analysis of the systems operating conditions in order to detect and identify a failure when it occurs and even sometimes beforehand. Nowadays, two main approaches are explored to detect failures in PHM solutions: one based on machine learning, the other based on expertise and capitalised system knowledge. This work proposes to combine a Complex Event Processing (CEP), to manage incoming data’s volumetry and velocity, with an Expert System (ES) in charge of exploiting the capitalized knowledge. This paper focuses on the configuration of a CEP from rules contained in a CBM ES using a Model Driven Architecture (MDA). This configuration is a challenge, especially regarding the management of rules with temporal parameters and the need for intermediate results to deal with the rule’s complexity.
Document type :
Conference papers
Complete list of metadata
Contributor : IMT Mines Albi IMT Mines Albi Connect in order to contact the contributor
Submitted on : Thursday, April 15, 2021 - 9:09:41 AM
Last modification on : Thursday, November 10, 2022 - 1:33:52 PM




Alexandre Sarazin, Sébastien Truptil, Aurelie Montarnal, Jérémy Bascans, Xavier Lorca. Model Transformation from CBM to EPL Rules to Detect Failure Symptoms. MODELSWARD 2020 - 8th International Conference on Model-Driven Engineering and Software Development, Feb 2020, La Valette, Malta. pp.200-224, ⟨10.1007/978-3-030-67445-8_9⟩. ⟨hal-03198701⟩



Record views