Modeling dependency in inspection performance

Non-destructive evaluation of steel structures subject to flaws and fatigue cracks is considered. Hitherto it has been implicitly assumed that inspections of different locations in structural systems are statistically independent. This is, however, generally not the case due to common influencing factors such as environmental conditions and inspector characteristics. In the present paper the probabilistic modeling of inspection quality is discussed. Considering the different types of uncertainties in the modeling, it is demonstrated how the dependency in inspection performance from one location to another can be accounted for. Bayesian updating of the parameters of the inspection performance models is investigated and possible applications are discussed. Finally it is outlined how the dependency effects can be treated in the context of inspection and maintenance planning.