Obtaining G12 and Xt Using Mixed ANNs Based on Matrix and Fiber Properties

This study aimed to create ANN architectures capable of modeling the shear modulus (G12) and longitudinal ultimate tensile strength (Xt) of unidirectional composites. The trained and tested ANNs are used to obtain these mechanical properties. Based on the properties of the fiber, matrix, and their volumetric fraction, this analysis will use the lowest possible number of fiber- and matrix-related properties. To that end, data were collected from the literature and divided into a training and test set for cross-validation. Seven different architecture types were developed, changing their inputs and adding analytical models parallel to the analysis so that three were used to obtain G12 and four to obtain Xt. Classic analytical models, such as the Halpin-Tsai and rule of mixtures, were used to construct the mixed ANNs and serve as a reference for comparison purposes.