ABSTRACT
Understanding the remaining life, readiness, and future reliability of complex marine structures is one of the central challenges of marine lifecycle structural management. The authors have previously proposed a parametrically-encoded Bayesian network approach to combine load monitoring, structural deformations and crack inspection results into an updating framework to perform future fatigue life predictions. The framework uses a lognormal S-N crack initiation approach, assuming vessels start their service lives with uncertain fatigue strength and load spectral estimates. Additionally, extreme values from the same load spectra are assumed to cause visible deformations of the plating field in the structure. The network updates design-stage estimates of loading and fatigue strength based on in-service measurements. The updated parameters are then used for prognosis of future structural state. In our previous work, a fleet of five vessel, with four inspected and one un-inspected was assumed. In this case, we showed that either positive or negative evidence (e.g. the presence or absence of cracks) was useful to the network for updating. Thus, the value of differing types of evidence was not as critical as may be expected. However, in the previous work the impact of the number of ships inspected was not explored. Here we extend the study to explore how the prognosis capability degrades as fewer and fewer ships are inspected.
