Reliability-Based Analysis of a Caisson Breakwater with the Application of Bayesian Inference

Abstract

Caisson breakwaters are mainly constructed in deep waters to protect an area against waves. These breakwaters are conventionally designed based on the concept of the safety factor. However, the wave loads and resistance of structures have epistemic or aleatory uncertainties. Furthermore, sliding failure is one of the most important failure modes of caisson breakwaters. In most previous studies, for assessment purposes, uncertainties, such as wave and wave period variation, were ignored. Therefore, in this study, Bayesian reliability analysis is implemented to assess the failure probability of the sliding of Tombak port breakwater in the Persian Gulf. The mean and standard deviations were taken as random variables to consider dismissed uncertainties. For this purpose, the first-order reliability method (FORM) and the first principal curvature correction in FORM are used to calculate the reliability index. The performances of these methods are verified by importance sampling through Monte Carlo simulation (MCS). In addition, the reliability index sensitivities of each random variable are calculated to evaluate the importance of different random variables while calculating the caisson sliding. The results show that the reliability index is most sensitive to the coefficients of friction, wave height, and caisson weight (or concrete density). The sensitivity of the failure probability of each of the random variables and their uncertainties are calculated by the derivative method. Finally, the Bayesian regression is implemented to predict the statistical properties of breakwater sliding with non-informative priors, which are compared to Goda’s formulation, used in breakwater design standards. The analysis shows that the model posterior for the sliding of a caisson breakwater has a mean and standard deviation of 0.039 and 0.022, respectively. A normal quantile analysis and residual analysis are also performed to evaluate the correctness of the model responses.

Keywords

Breakwater sliding, First-order reliability method (FORM), Aleatory and epistemic uncertainty, Monte Carlo simulation, Sensitivity analyses, Bayesian linear regression (BLR)