Reliability assessment from degradation models

The key of a successful intelligent maintenance consists of replacing a worn equipment just before a failure occurs. In this context predictive maintenance policy can be implemented if a degradation indicator is available. This paper proposes some degradation models that are fitted to degradation measurements. The re-liability is then obtained from those models and mean residual lifetimes are also calculated. The strong point of this approach is that no failure is needed to retrieve the reliability information compared to classical methods. A discussion is proposed to know which model is the most adequate for degradation data. Further-more, physic based methods are also taken into account when the degradation evolution can be explained by a physical process.