Reliability assessment of marine structures considering multidimensional dependency of the variables

For an improved estimation of marine structural reliability, a consideration of random variable dependency is essential. With a limited study on dependence modelling of marine structures, this study proposes a framework for reliability assessment of ocean structural systems with multidimensional variables. This framework captures possible nonlinearity and tail dependence in the variables using vine copula. The proposed method develops a graphical structure of random variables consisting of nodes, edges, and trees using the Drawable Vine (D-vine) approach. This study demonstrates the developed framework on a jacket support structure subjected to the extreme environmental load conditions at Jeanne D′ Arc basin on Canada's east coast. The structure's reliability is evaluated with optimally selected copulas in the D-vine trees and associated marginal distributions. A comparison between the reliability result using the D-vine copula method, Gaussian coupling assumption, and statistical independence between variables proved its superiority in modelling variable dependence of complex marine systems. The probability of failure (Pf) using D-vine copula was closer to the reference Importance Sampling (IS) results than other methods. •A framework considering multidimensional non-linear dependent variables using vine copula is proposed.•Application of the framework is demonstrated on a jacket support structure subjected to the extreme load.•D-Vine copula effectiveness is demonstrated by comparing it with independent variables and Gaussian coupled variables.