Skip to Main content Skip to Navigation
New interface

Modèles Probabilistes de Séquences Temporelles et Fusion de Décisions. Application à la Classification de Défauts de Rails et à leur Maintenance

Abstract : Compared to the various activities of "Keeping the Operational Conditions" of an industrial system, these PhD, initiated in the framework of a partnership between the INRETS and CRAN, is focusing on the maintenance process with an application context dedicated to the maintenance of the railway. Railway firms, with the priority of improving the safety and welfare of passengers, are seeking to adjust their maintenance policy which is nowadays primarily corrective or executed at a predetermined time interval, to a more conditional or predictive planning with a minimum of costs. In this new context, rails maintenance should no longer be limited to the vision of the isolated component (portion of rail), but to the study of the whole system failure (N-components system). Therefore, maintenance decisions are no longer isolated from their context and exist on a continuum Surveillance - Diagnostic - Decision making. In response to these industrial needs and its scientific issues, our contribution focuses first on an original approach of diagnosis (hybrid approach), which is based on a fusion of two different information sources: Local Approach (sensor eddy currents) and Global Approach. RBD have been used to develop probabilistic models to for the classification of singular points of the track. The fusion between these models and the local approach was produced by naive Bayes fusion method. The result of this fusion constitutes the input of decision making process, for which we have proposed, secondly, a generic methodology for optimizing conditional based maintenance of N-components systems. Our approach is based on a combination of dynamic Bayesian networks and MDP (Markov Decision Processes) to be able to model N-components systems in a factored way. This new proposal was illustrated by a first instance academic to highlight its feasibility then it has been applied in the framework of the optimization of maintenance of the surface defects rail.
Document type :
Complete list of metadata

Cited literature [156 references]  Display  Hide  Download
Contributor : Thèses UL Connect in order to contact the contributor
Submitted on : Thursday, March 29, 2018 - 11:32:09 AM
Last modification on : Saturday, October 16, 2021 - 11:14:14 AM
Long-term archiving on: : Friday, September 14, 2018 - 12:23:57 AM


Files produced by the author(s)


  • HAL Id : tel-01748340, version 1



Abdeljabbar Ben Salem. Modèles Probabilistes de Séquences Temporelles et Fusion de Décisions. Application à la Classification de Défauts de Rails et à leur Maintenance. Autre. Université Henri Poincaré - Nancy 1, 2008. Français. ⟨NNT : 2008NAN10008⟩. ⟨tel-01748340⟩



Record views


Files downloads