Artificial neural networks -based prediction of mechanical properties in h-BN reinforced composites

In today's advanced engineering applications, fast and high-accuracy estimation of the mechanical performance of composite materials is of vital importance. This study aims to estimate the force and deformation values of hybrid structures containing hexagonal boron nitride (h-BN) nanoadditives in certain proportions of carbon and basalt fiber-reinforced composite materials. The research problem is to question the ability of artificial neural networks (ANNs) to model these complex force-deformation relationships and to provide reliable results using the data obtained from experimental three-point bending tests. The developed model was trained with the Levenberg-Marquardt backpropagation algorithm, and overlearning was prevented by the early stopping method. The obtained results showed that the model produced high accuracy estimations with the R2 value of 0.987, the mean error of 0.0021, and the maximum error of 0.015; the error value decreased from 0.08 to 0.0012 during the training process. These findings reveal that the proposed ANN-based approach provides a faster, cost-effective, and reliable estimation alternative compared to traditional experimental methods.