Condition-based maintenance for complex systems based on current component status and Bayesian updating of component reliability

• Published in: Reliability engineering & system safety 2017-12, Vol.168, p.227-239
• Main Authors: Walter, Gero, Flapper, Simme Douwe
• Format: Article
• Language: English
• Subjects: Bayesian analysis; Block diagrams; Complex systems; Component reliability; Condition-based maintenance; Distribution; Maintenance; Mathematical models; Reliability aspects; Reliability engineering; Remaining useful life; Survival; Survival signature; System reliability; Unit time cost rate; Weibull distribution
• ISSN: 0951-8320; 1879-0836
• DOI: 10.1016/j.ress.2017.06.015

•Presents a new class of condition-based maintenance policies for complex systems.•Uses the reliability block diagram and the status (working/defective) of components.•Closed-form Bayesian update of system reliability using all available information.•Separate Weibull model for each component type using expert knowledge and test data.•Maintenance decision via minimization of the expected one-cycle cost rate. We propose a new condition-based maintenance policy for complex systems, based on the status (working, defective) of all components within a system, as well as the reliability block diagram of the system. By means of the survival signature, a generalization of the system signature allowing for multiple component types, we obtain a predictive distribution for the system survival time, also known as residual life distribution, based on which of the system’s components currently function or not, and the current age of the functioning components. The time to failure of the components of the system is modeled by a Weibull distribution with a fixed shape parameter. The scale parameter is iteratively updated in a Bayesian fashion using the current (censored and non-censored) component lifetimes. Each component type has a separate Weibull model that may also include test data. The cost-optimal moment of replacement for the system is obtained by minimizing the expected cost rate per unit of time. The unit cost rate is recalculated when components fail or at the end of every (very short) fixed inter-evaluation interval, leading to a dynamic maintenance policy, since the ageing of components and possible failures will change the cost-optimal moment of replacement in the course of time. Via numerical experiments, some insight into the performance of the policy is given.