Decision support with a markovian approach for maintenance context activities

Pascal Vrignat (1) , Manuel Avila (1) , Florent Duculty (1) , Bernard Robles (1) , Frédéric Kratz (1)

1 PRISME - Laboratoire Pluridisciplinaire de Recherche en Ingénierie des Systèmes, Mécanique et Energétique

Today, maintenance strategies and their analyses remain a worrying problem for companies. Our study, presented in this paper, shows that our indicator can provide help for expert decision (maintenance manager) in the context of condition-based maintenance. The paper deals with the proposition of using Hidden Markov Models to track and estimate the degradation of a system, according to observations (maintenance activities registered in a database). In a first time, the degradation level of process was established by a "classical" degradation laws (statistical laws). In a second step, this level was established by Hidden Markov Model (probabilistic laws).Tests conducted on the synthesis model, for which degradation levels were known, allowed us to implement the method.

Domaines: Automatique

Pascal Vrignat, Manuel Avila, Florent Duculty, Bernard Robles, Frédéric Kratz. Decision support with a markovian approach for maintenance context activities. European Safety & Reliability International Conference, Sep 2011, Troyes, France. pp.66. ⟨hal-00626533⟩