Accelerating the computational process for analyzing dynamic thermal behavior of an innovatively designed longitudinally-graded sandwich plate: Deep-learning approach

Abstract

Employing the solutions through artificial intelligence (AI) to solve mechanical engineering problems can pave the way toward accelerating the computation process, especially for those engaged with the dynamic nature of the time-variable systems. In this regard, this article attempts to employ deep neural networks (DNN) for the coupled thermoelasticity assessment of the longitudinally sandwiched plates exposed to intensive thermal shock. The left and right face sheets of the sandwich plate are enriched with graphene platelets (GPLs) along the longitudinal direction. The thermal shock is considered as the instantaneous heat flux applied in the axial direction. Elasticity theory is employed for deriving the governing equations in an exact template and differential quadrature approach (DQA) in conjunction with the Laplace transform plays the role of solution approaches to attain the thermoelastic reaction of the sandwich system at training points. Then, the DNN utilizes these points to adjust a predictor mechanism for analyzing the behavior of the sandwich system in every load-exertion circumstances. Additionally, the DNN would be optimized by a brand-new metaheuristic algorithm according to some error estimation criteria to minimize the error of regression. The mentioned solution is validated by comparing its performance with the outcomes of the published literature. This study provides a novel platform to use the great potential of AI-based methods in solving dynamic thermoelastic problems in the world of solid structures.