A Neural-Network Simulator for the 3-Lobes Stern Seal of a S.E.S.

摘要

In this article, the problem of modelling and simulating the 3-lobes stern seal of a Surface Effect Ship (S.E.S.) is considered. Recently it has been derived an algebraic model of the stern seal; unfortunately, the corresponding equation is an implicit, non-linear and stiff equation, which does not admit an explicit solution. On the other hand, the numerical solution of the equation turns out to be time-consuming and somewhat unreliable. In this article, we explore the possibility of replacing the stiff equation numerical simulator with a suitably trained neural network. The objective is to obtain a fast and reliable “black-box” model of the 3-lobes stem seal, suitable for an on-line use within the entire S.E.S. simulator. Several networks have been considered; among them, an innovative structured network revealed to provide the best performances in terms of approximation accuracy and parameters parsimony.