Development of reduced structural theories for composite plates and shells via machine learning

This paper presents a new approach for the development of structural models via three well-established frameworks, namely, the Carrera Unified Formulation (CUF)[1], the Axiomatic/Asymptotic Method (AAM)[2], and Artificial Neural Networks (NN)[3]. CUF and AAM provide the finite element arrays and measure the relevance of any given generalized displacement variable. The NN training makes use of the data from CUF-AAM and the outputs are the Best Theory Diagrams [4] curves providing the minimum number of nodal degrees of freedom required to satisfy a given accuracy requirement and the accuracy of any structural theory. The main governing equations for plate and shell finite elements via CUF are the following and lead to the implementation of any order theory, u(x, y, z) = FτNi(z)uτi(x, y) => ∫