A Probabilistic Method for Quantifying Uncertainty in Crack Detection and Its Effects on Marine Structural Integrity Management

Non-destructive testing (NDT) methods have been widely used for damage examination and structural maintenance, e.g., detecting and repairing fatigue cracks. In-service inspections help increase fatigue reliability by providing new information for updating structural failure probability and making repair decisions. However, these benefits are often compromised by uncertainties associated with inspection methods. Sometimes, existing cracks may not be identified, and positive inspection indication may actually not exist. It is of great interest to consider the influence of inspection uncertainty in maintenance optimization, because the benefits and costs of maintenance are affected by inspection decisions (e.g., inspection times and methods), which are subjected to inspection uncertainty. The influence of inspection uncertainty on maintenance optimization has not been explicitly and adequately covered in the literature. In this paper, the problem has been investigated by probabilistic modelling of the qualities of inspection methods via probability of detection (PoD) functions. A new PoD function is proposed to characterize the inspection quality when inspection uncertainty is neglected. Optimum inspection decisions are derived by the objective of maximizing lifetime reliability index under two scenarios (considering and not considering inspection uncertainty). The effectiveness index of a planned inspection is defined based on the max reliability indexes under the two scenarios. It is shown that the max lifetime reliability index generally deceases when inspection uncertainty is considered. Inspection uncertainty may have little influence on the lifetime reliability index, depending on the planned inspection time. The effectiveness index of a planned inspection increases with the decrease in the mean detectable crack size. However, inspection uncertainty can result in significant increases in expected life cycle costs and maintenance costs.