HPPN-based Prognosis for Hybrid Systems

Pauline Ribot (1) , Elodie Chanthery (1) , Quentin Gaudel (1)

1 LAAS-DISCO - Équipe DIagnostic, Supervision et COnduite

This paper presents a model-based prognosis method for hybrid systems i.e. that have both discrete and continuous behaviors. The current state of the hybrid system is estimated by a diagnosis process and the prognosis process uses this state estimation to predict the future states and to determine the end of life (EOL) or the remaining useful life (RUL) of the system. The Hybrid Particle Petri Nets (HPPN) formalism is used to model the hybrid system behavior and degradation. A HPPN-based diagnoser has already been defined to provide a current state estimation that takes uncertainty about the system model and observations into account. We propose to generate a prognoser from the HPPN model of the system. This prognoser is initialized and updated with the result of the HPPN-based diagnoser. It computes a distribution of beliefs over the future mode trajectories of the system and predicts the system RUL/EOL. The prognosis methodology is demonstrated on a three tanks example.

Mots clés: prognosis; hybrid systems; HPPN

Domaines: Automatique / Robotique

Pauline Ribot, Elodie Chanthery, Quentin Gaudel. HPPN-based Prognosis for Hybrid Systems. Annual Conference of the Prognostics and Health Management Society 2017, Oct 2017, St. Petersburg, United States. ⟨hal-01579483⟩