Degradation in presence of different types of maintenance: Modelling, inference and decision making
Speciality : Mathématiques Appliquées
12/06/2024 - 09:00 Margaux Leroy (Université Grenoble Alpes) Auditorium de l'IMAG, Campus de l'Université Grenoble Alpes, Saint Martin d'Hères
Keywords :
- Maintenance modeling and optimization
- statistical estimation
- deterioration modeling
This thesis is part of a project of the équipe action AMORE MIO, financed by PERSYVAL-lab, involving both the LJK and the GISPA-lab. Within the framework of this work, we are considering technological and industrial equipment or structural components (e.g. sewers, dams, ponds, etc.) that are subject to degradation due to intrinsic wear, use imposed by operating conditions or exposure to environmental factors. For these repairable industrial installations, it is essential to maintain the system operational, taking into account safety, availability and cost constraints. The maintenance actions can have different effects and different inspection policies can be considered. To represent this degradation over time and to anticipate eventual failures, two probabilistic models are proposed. For each of these two models, degradation is described by a Wiener process with drift, and the maintenance effect is of the ARD1 type (arithmetic reduction of degradation). Statistical inference is studied using various observation schemes. Based on one of these degradation models, the decision making is studied under two inspection policies. The maintenance cost is assessed and maintenance strategies are optimised. The optimal maintenance policies obtained are then compared, in particular between the two proposed inspection policies.
Directors:
- Laurent Doyen
- Christophe Bérenguer