Rapid assessment of out-of-plane nonlinear shear stress–strain response for thick-section composites using artificial neural networks and DIC

Abstract

Thick-section fiber-reinforced polymer matrix composites (TSCs) have been increasingly used in primary and secondary load-bearing structures. Due to complex deformation mechanisms, accurate out-of-plane constitutive relations, including nonlinear shear stress–strain response, are required for structural analysis and failure prediction of TSC structures. A plate torsion test is proposed to assess the nonlinear out-of-plane shear stress–strain response using an artificial neural network (ANN) and digital image correlation (DIC). The ANN model was developed using three-dimensional finite element numerical simulations, and the out-of-plane shear stress was obtained with good accuracy. The efficiency of the shear stress calculation has been improved. The usability of the test method has been refined substantially. A comparison of the shear stress with the FEM-calculated result showed that the average error of the ANN-predicted shear stress was less than 0.5%. Accordingly, the constitutive parameters of a 15-ply thick S6C10/AC318 glass/epoxy unidirectional tape panel have been extracted from the minimization of the normalized error between ANN-calculated shear stress and DIC-measured shear strain iteratively. The efficiency of stress updating for the plate-torsion specimen has been improved by 105 times to extract the material constitutive parameters. Not limited to plate-torsion specimens, the proposed ANN-DIC method can be extended further for rapid assessment of material constitutive parameters for other anisotropic polymer-based composite materials using complex, statically indeterminate tests.