Application of Pattern Recognition Method for Estimating Wind Loads on Ships and Marine Objects

Abstract

Wind loads on ships and marine objects are a complicated phenomenon because of the complex configuration of the above-water part of the structure. This study presents an extension of application capabilities of elliptic Fourier descriptors (EFDs) from the usual pattern recognition and classification problems to problems of very complex nonlinear multivariable approximations of multi-input and multi-output (MIMO) functions, where EFDs are used for ship frontal and lateral closed contour representation. This approach takes into account all aspects of the variability of the above-water frontal and lateral ship profile. It is very suitable for assessing wind loads on marine structures wherever we have a wind load database for a group of similar vessels. In this way the cheaper and faster calculation can bridge the gap between ship shapes for which calculations or experiments have already been made. The Generalized Regression Neural Network (GRNN) is trained by EFDs of closed contours as inputs and wind load data derived from wind tunnel tests for a group of ships are used as targets. The trained neural network is used for the estimation of non-dimensional wind load coefficients and results for a group of offshore supply vessels, car carriers and container ships are presented and compared with the experimental data. Finally, sensitivity analysis is performed with respect to the variability of lateral container vessel contours in order to investigate how small changes in the contour geometry affect the overall estimation of wind load coefficients.