Replacement policies for a complex system with unobservable components using dynamic Bayesian networks

Abstract

We study maintenance of a complex dynamic system consisting of ageing and unobservable components under a predetermined threshold reliability level. Our aim is to construct an optimum replacement policy for the components of the system by minimizing total number of replacements or total replacement cost. We represent the problem with dynamic Bayesian networks (DBNs). We prove that under the existence of a predetermined threshold reliability, performing replacements at periods when the system reliability just falls below the threshold assures optimum replacement times. Four component selection approaches and their cost focused versions are proposed to choose the component to replace and are tested on a complex dynamic problem. Their performances are analyzed under various threshold and cost levels.

Evaluation of proactive maintenance policies on a stochastically dependent hidden multi-component system using DBNs

“…Based on DBNs, two preventive maintenance policies, CIPM and DIPM [34] inspired by block-based and age-based strategies and a threshold based proactive strategy (ThPM) inspired by [35] are designed and evaluated for multi-component dependable systems.…”