Multiscale analysis of carbon nanotube-reinforced curved beams: A finite element approach coupled with multilayer perceptron neural network

This paper presents a comprehensive investigation into the structural response of curved composite beams enhanced with carbon nanotube (CNT). Employing a multiscale framework, our analysis leverages the finite element method (FEM) to account for both bending and shear deformations across six degrees of freedom. The inquiry encompasses diverse mechanical, geometrical, and boundary configurations to assess these composite beams' natural vibration features. Moreover, we introduce a multilayer perceptron (MLP) neural network architecture designed to forecast such beams' dimensionless first natural frequency. Trained on a meticulously curated dataset derived from FEM simulations, the neural network model exhibits promising predictive capabilities concerning the free vibration frequency. To ascertain the efficacy and precision of our proposed methodology, we conduct a comparative analysis between FEM results and employ statistical metrics to evaluate the neural network's predictive performance. The findings of this study reveal an impressive predictive accuracy of over 95 % with regards to the initial natural frequency of the composite beams, thereby emphasizing the potential effectiveness of neural network methodologies in engineering analyses. This study significantly contributes to advancing our comprehension of the vibrational dynamics inherent in carbon nanotube-reinforced composite beams, while concurrently underscoring the potential efficacy of neural networks in forecasting their dynamic attributes.