Estimating Maneuvering and Seakeeping Characteristics with Neural Networks

Maneuvering and seakeeping are two very important naval architecture research areas. There are several methods to determine a vessel's behavior, but most of them are time-consuming, apply linear techniques or introduce several simplifications. This paper proposes to apply feed-forward neural networks to predict maneuvering behavior in the design phase or following changes on a new design. The feed-forward neural network is trained using sea maneuvering trials data of similar vessels. In order to prove this hypothesis, the method is applied to a set of 47 maneuvering trials from two different vessels, obtaining a standard error of 6.61%, which compares favorably with conventional methods.