Failure load prediction of adhesively bonded pultruded composites using artificial neural network

Abstract

Mechanical joining and adhesive bonding provide convenience for manufacturing of complex structures, which made of composite materials. Failure load is directly related with process parameters of mechanical joining or adhesive bonding. In this study, the effects of bonding angle, patching type (single side and double side) and patching structure on the failure load were investigated in the pultruded composite specimens. For this aim, the pultruded composite specimens, which bonded with five different bonding angles (45°, 51°, 59°, 68° and 90°) and five different bonding types as unpatched, single-side woven patch, single-side knitting patch, double-side woven patch and double-side knitting patch were exposed to tensile loads at room temperature. In the view of experimental results, the failure loads of bonded pultruded composite specimens were predicted by training six different artificial neural network algorithms. The only three best prediction results of Bayesian regularization, Levenberg–Marquardt and scaled conjugate gradient were given in the figures for better understanding.