A new metamodel for predicting the nonlinear time-domain response of offshore structures subjected to stochastic wave current and wind loads

Offshore structures are subjected to stochastic loads from wave, current, and wind. A long-term response analysis is important for evaluating the structural safety; however, it is computationally demanding to perform the large number of nonlinear time-domain simulations. Metamodeling is an attractive prospect, but there are two existing obstacles to overcome. First, existing metamodels are mostly confined to a single sea state and are unable to predict the response for different wave directions. Second, for floating systems, the inputs are usually vessel loads/motions, which are a priori unknown. This paper develops a new metamodel framework for accurately predicting the time series response under short-term uncertainties from the stochastic process, and seven long-term variables, namely the significant wave height, spectral period, wave direction, and speeds and directions of current and wind. The exogenous input is based only on the environmental variables, which can be predefined. The proposed metamodel is implemented on a floating system and used to predict the time series of riser stress, the extreme stress response and fatigue damage. Four cases with increasing number of long-term variables are tested. Excellent agreement is found between the metamodel prediction and numerical simulation even for the most critical case with all variables included.