Prediction of the acoustic form function by neural network techniques for immersed tubes

Abstract

A new approach is used to predict the acoustic form function (FF) for an infinite length cylindrical shell excited perpendicularly to its axis using the artificial neural network (ANN) techniques. The Wigner-Ville distribution is used like a comparison tool between the FF calculated by the analytical method and that predicted by the ANN techniques for a stainless steel tube. During the development of the network, several configurations are evaluated for various radius ratios ba (a: outer radius: b: inner radius of the tube). The optimal model is a network with one hidden layer. It is able to predict the FF with a mean relative error about 1.61% for the cases studied in this paper.