Dynamic Bending Moment Identification Using Neural Networks

Abstract

An alternative method for the prediction of a ship's dynamic bending stresses at sea is presented and examined in this paper. The method uses a ship's heave and pitch motion to determine the dynamic bending moment at a point along the ship's length. This can be combined with the known still water bending moment, and known ship sectional properties to determine deck and keel stresses. A combination of mathematical modeling, the random decrement, and neural network techniques have been used to determine the relationship between ship motion and bending moment, without any prior knowledge of the wave excitation level To test this method, two sets of model experiments have been used. One set from a Great Lakes bulk carrier, the other from a Canadian patrol frigate. In each experiment, the mean and variance of the bending moment have been successfully predicted, demonstrating this method as a valid approach.