Reliability of marine structures under scheduled inspections using Bayesian analysis.

摘要

Fatigue is a dominant failure mode of modern offshore structures. Because of the great investment in each structure and concern over the consequences of structural failure, owners must periodically inspect and maintain their structures in order to ensure continued serviceability. A good inspection and record keeping system can provide feedback to revise the inspection and maintenance program to reduce its overall cost and increase its effectiveness. In this dissertation probabilistic methodologies are developed for reliability analyses of fatigue critical components under scheduled maintenance in service. Fatigue cracking is represented probabilistically in terms of (1) crack initiation, (2) crack growth, and (3) the ability to detect cracks during inspections. Bayesian analysis provides a tool for incorporating the results of inspections into the inspection program itself, in order to maintain the structure at or above a specified level of structural reliability. The occurrence and growth of fatigue cracks is modeled in terms of statistical distributions with uncertain parameters representing the three aspects identified above, and which are updated based upon inspection results. Possible inspection results for fatigue critical locations are expressed in terms of mutually exclusive events, which provide for calculating the structural reliability of a critical location at a specified time after each inspection. A numerical example demonstrating the procedures for determining inspection intervals incorporating previous inspection results has been carried out for cases with all three uncertain parameters. A study has been carried out for additional cases using any two of the three uncertain parameters to determine how well the uncertain parameters can be estimated under such limiting conditions.