On the modeling of wave propagation in the multi-curved inhomogeneous system under in-plane load via a non-model-based mechanism and fully connected deep neural networks

Abstract

The deep neural networks (DNNs) technique is one of the best computational methods with low error. However, due to their superior performance, such as reducing the cost of computational complexity, this kind of machine learning has attracted much attention. Therefore, in the present paper, for the first time, wave-dispersion results of an inhomogeneous system are compared with the outcomes of DNN techniques for obtaining the exact mean squared error (MSE) parameter for future research. In the mathematical modeling section of this report, an improved three-dimension higher-order theory (3D-HSDT) considering the effect of thickness stretching for obtaining the governing equations of the multi-phase hybrid nanocomposites reinforced rectangular microplate has been presented. Phase-velocity of the current micro structure has been obtained by employing the three-dimension modified couple stress theory (3D-MCST). This non-classical model is capable of capturing the small-size impact on the behavior of the rectangular plates. Also, with the aid of COMSOL multiphysics finite element simulation, the results are verified, and new outcomes for improving the stability of the presented structure are studied. Finally, the results show that the presented mathematical modeling and COMSOL multiphysics finite element software can be verified with the DNN techniques in the specific MSE parameter.