Introducing hybrid deep neural networks to predict the transient vibrations of the sandwich systems reinforced by multi-phase nanocomposites

Abstract

In advanced materials engineering, understanding transient vibrations in sandwich annular plates reinforced by multi-phase nanocomposites is essential. To adequately represent the geometry of these systems, a generic cylindrical coordination system is often used in modeling. The heterogeneous laminated plate is homogenized into an analogous uniform medium to facilitate analysis. In order to represent precise stress and strain distributions, taking into account rotating inertia and transverse shear deformation, a higher-order theory based on a hyperbolic function is used. The governing equations of motion are derived using Hamilton’s principle in polar coordinates, guaranteeing compliance with energy laws. The 2D generalized differential quadrature (GDQ) method effectively discretizes the governing equations for numerical solutions while retaining a high degree of accuracy in recording transient reactions. Furthermore, by solving the differential equations in the frequency domain, the Laplace transform approach sheds light on the dynamic behavior of the system. Using the deep neural networks (DNNs)-fuzzy approach improves the prediction of these composite structures’ vibrational activity. This hybrid approach uses fuzzy logic to control uncertainty and DNN for pattern identification. The DNN-fuzzy approach reliably predicts the intricate vibrational properties of multi-phase sandwich systems reinforced with nanocomposite using datasets from mathematical modeling. This combined use of mathematical techniques, theoretical models, and predictive algorithms improves our knowledge of complex composite structures’ transient vibrations and makes them more useful in a range of engineering domains.