This paper presents three probabilistic lifing methods, Recursive Probability Integration (RPI), particle filtering (PF)-based probabilistic lifing method and a hybrid lifing method for fatigue management with physics-based damage propagation model and health state information obtained from health monitoring systems. RPI is suitable for fleet risk management if probability of detection (POD) is the only information provided by the system. Particle filtering approach can be used for individual risk tracking if a relationship between features from a health monitoring system and damage extent can be reliably developed. However, both RPI and PF have limitations for fatigue management. A hybrid method which utilizes RPI and PF is proposed for risk tracking in three stages. PRI is used in the first stage when the damage is small and signal-to-noise ratio from the health monitoring system is low. PF is applied in stage two in which reliable health information can be obtained for individual risk tracking. In stage three, RPI is applied again for risk predictions. In this paper, we will also provide numerical examples to discuss individual risk tracking using particle filtering approach with a typical crack growth model and a generic measurement function representing the relationship between features and damages. Effect of prior probability distributions of unknown parameters using PF on the predictions of fracture-based remaining useful life as well as calculations of posterior probability distributions will be presented and discussed.
展开▼
