Failure prediction of ±55° glass/epoxy composite pipes using system identification modelling

Abstract

Black-box modelling using system identification method to predict the performance of glass fibre reinforced epoxy (GRE) composite pipe under multiaxial loading stress ratio is presented. In this study, both linear and nonlinear models were derived namely; linear time-invariant parametric model and artificial neural network model. The models derived are to approximate the pure hydrostatic loading performance using GRE pipes with winding angles of ±55°. Three different linear model structures were derived, and the best fit model achieved at 96.64% of best fit. On the other hand, the Artificial Neural Network (ANN) modelling showed better accuracy with the best fit of 99.82%. Finally, the point of failure at which first damage takes place predicted by the models derived was validated using experimental data.