Fatigue inspection and maintenance optimization: A comparison of information value, life cycle cost and reliability based approaches

Abstract

Fatigue cracks increase structural failure risk and timely maintenance is very important. Maintenance planning is often formulated as a probabilistic optimization problem, considering uncertainties in structural and load modelling, material properties, damage measurements, etc. A decision rule or strategy, e.g. condition based maintenance (CBM), needs to be set up, and then an optimal maintenance criterion or threshold is derived via solving the optimization problem. This paper develops a probabilistic maintenance optimization approach exploiting value of information (VoI) computation and Bayesian decision optimization. The VoI based approach explicitly quantifies added values from future inspections, and gives an optimal decision (or strategy) by direct modelling decision alternatives and evaluating their expected outcomes, rather than a pre-defined strategy. A comparative study on VoI, life cycle cost (LCC) and reliability based optimization approaches is conducted. It is shown that the VoI based approach takes all available maintenance strategies into account (both with and without involving inspections), and can reliably yield optimal maintenance strategies, whether the VoI is larger than or equal to zero. When the VoI is equal to zero, LCC and reliability based CBM optimization can lead to suboptimal maintenance strategies. The differences in the approaches are illustrated on fatigue-sensitive components in a marine structure.