DEEP LEARNING-BASED PREDICTION AND EXPERIMENTAL EVALUATION OF MECHANICAL PROPERTIES IN BASALT-FIBER-SiC-REINFORCED HYBRID EPOXY COMPOSITES

This study investigates the predictive capabilities of deep neural network (DNN) models in estimating the mechanical properties of basalt-fiber-reinforced hybrid composites (BFRHCs). A total of 27 composite samples were fabricated using the compression-molding technique, incorporating varying numbers of basalt-fiber layers, fiber orientation angles (0°, 45°, 90°), and silicon carbide (SiC) filler contents (0, 3 and 6 w/%). The fabricated specimens were subjected to mechanical characterization by ASTM standards to evaluate their tensile strength, flexural strength, impact strength, hardness, shear strength, and interdelamination resistance. A DNN model was developed and trained on the experimental dataset to capture the complex, nonlinear relationships between input fabrication parameters and output mechanical properties. Model performance was assessed using the coefficient of determination (R²), mean absolute error (MAE), and root-mean-square error (RMSE). The DNN achieved high predictive accuracy with R² values exceeding 0.95 for most properties, demonstrating its effectiveness in forecasting composite performance. The results confirm that deep-learning frameworks such as DNNs offer a powerful and reliable approach to predicting the behavior of hybrid composites, reducing the need for extensive experimental trials and supporting efficient material design and optimization in structural applications.