A distributed architecture to implement a prognostic function for complex systems

Abstract

The proactivity in maintenance management is improved by the implementation of CBM (Condition-Based Maintenance) principles and of PHM (Prognostic and Health Management). These implementations use data about the health status of the systems. Among them, prognostic data make it possible to evaluate the future health of the systems. The Remaining Useful Lifetimes (RULs) of the components is frequently required to prognose systems. However, the availability of complex systems for productive tasks is often expressed in terms of RULs of functions and/or subsystems; those RULs have to bring information about the components. Indeed, the maintenance operators must know what components need maintenance actions in order to increase the RULs of the functions or subsystems, and consequently the availability of the complex systems for longer tasks or more productive tasks. This paper aims at defining a generic prognostic function of complex systems aiming at prognosing its functions and at enabling the isolation of components that needs maintenance actions. The proposed function requires knowledge about the system to be prognosed. The corresponding models are detailed. The proposed prognostic function contains graph traversal so its distribution is proposed to speed it up. It is carried out by generic agents.