System health monitoring through Bayesian networks

Condition based maintenance is an advanced maintenance strategy able to achieve both cost reduction and high availability and reliability performance. It is based on the system health monitoring whose issues are di-agnostic and prognostic capabilities. The aim of this work is to develop and validate an expert system, by means of Bayesian networks, able to support the health monitoring process of a chiller unit, suitably manufactured for this research scope. We collected the data directly from the running chiller by a condition monitoring system, in order to build the belief network. The goal is the evaluation of the system residual life through the condition monitoring of the system compo-nents, in order to manage the on-condition maintenance activities according to spare parts’ cost, mission demand and magnitude of the potential damage and saving coming from the decision to postpone the operations