Artificial Neural Network Prediction of Ultimate Tensile Strength of Randomly Oriented Short Glass Fibre-Epoxy Composite Specimen using Acoustic Emission Parameters

Abstract

Acoustic emission (AE) data have been collected from 20 randomly oriented short E-glass fibre – unsaturated polyester tensile specimens, while loading up to failure in a tensile testing machine. Peak amplitude and cumulative energy data from AE response of each specimen were classified and segregated by understanding the failure mechanism and data acquired up to 50% of the failure load was utilized for analysis. An optimized feed-forward back-propagation (FFBP) type artificial neural network (ANN) was designed and the segregated data of amplitude hits and cumulative energy was processed using it. Even though the accuracy of both networks were satisfactory, amplitude hit based network gave better predictions of the ultimate tensile strength (UTS) than the energy based network. Also the performance of various training algorithms in the designed network was analysed and the results were compared.