Gabriel Mattos Gonzalez
◽
Marcos Queija de Siqueira
◽
Marina Leivas Simão
◽
Paulo Maurício Videiro
◽
Luis Volnei Sudati Sagrilo
Abstract
Nowadays, artificial intelligence algorithms are regaining visibility mainly due to the increase in computational capability. Among those, artificial neural networks (ANN) are very useful for the regression of highly nonlinear phenomena, such as the dynamic response of offshore structures. Due to the escalating demand in the oil and gas industry, offshore fields have been explored in deeper waters, which leads to more severe environmental conditions. A reliable and efficient evaluation of the long-term response of mooring systems, a crucial element of floating offshore structures, is then imperative. The estimation of the mooring long-term response is usually obtained numerically through the convolution of the short-term responses, based on short-term stationary environmental conditions (typically 3-h). Each of these short-term responses is obtained through a time-domain dynamic structural analysis from which statistical parameters of interest are calculated, such as the mean of the tension maxima sample or the maxima frequency. Such analyses tend to be quite time-consuming and a reliable estimator of these short-term statistical parameters may be of great help. In this paper, an ANN is trained to predict the short-term extreme peak response statistical parameters. The used training datasets include the wave significant height and spectral peak period for both wind sea and swell waves, generated by Importance Sampling Monte Carlo Simulation (ISMCS) method. Fixed directions of wind sea and swell are considered. It is shown that the ANN successfully predicts the short-term response statistical parameters for both cases, which are later used for the evaluation of the long-term N-year mooring response.