Modelling imperfect maintenance in an ARD1-type degradation model through partial maintenance effects - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2023

Modelling imperfect maintenance in an ARD1-type degradation model through partial maintenance effects

Résumé

Industrial components undergo degradation during their lifetime due to the intrinsic wear, environmental factors or operating conditions. To reduce this degradation, maintenance actions are performed on the system. In practice, these maintenance actions may have an effect on a part of the system only. For instance, on a vehicle’s braking system, a maintenance action usually only affects the brake pads but not the whole disc. In this case, maintenance effect is never perfect as a part of the system is not affected by the intervention and keeps deteriorating over time. This work focuses on the modelling of such a degradation process: we propose a new degradation model including imperfect maintenance actions and we develop the associated statistical inference procedures. The studied degradation model is based on the sum of two dependent underlying Wiener processes with drift. Maintenance effect is described by an arithmetic reduction of the degradation and only affects one of the two processes. Each time a maintenance is performed, the degradation of one out of the two processes is reduced of a certain quantity that is proportional to the degradation accumulated on this process since the last maintenance action. To simplify likelihood’s writing, two equivalent parameterizations are proposed. An identifiability problem is raised in the model requiring constraints on parameters to prevent this issue. Statistical inference is explored according to two different observation schemes, depending on whether degradation levels are observed at maintenance times or not. In the first observation scheme, degradation levels are observed just before and just after every maintenance and can also be observed between these maintenance actions whereas in the second (more generic) observation scheme, degradation levels are observed anytime between maintenance actions but not necessarily just before or after these actions. Finally, a simulation study is carried out in order to assess the quality of the estimators. The impact of different features of the model and data (values of the parameters, number of observations, number of maintenances) is also analyzed.
Fichier non déposé

Dates et versions

hal-04322351 , version 1 (04-12-2023)

Licence

Paternité - Pas d'utilisation commerciale - Pas de modification

Identifiants

  • HAL Id : hal-04322351 , version 1

Citer

Margaux Leroy, Laurent Doyen, Olivier Gaudoin, Christophe Bérenguer. Modelling imperfect maintenance in an ARD1-type degradation model through partial maintenance effects. MMR 2023 - 12th International Conference On Mathematical Methods In Reliability, Universidad de Murcia, May 2023, Murcia, Spain. ⟨hal-04322351⟩
33 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More