A robust five-unknowns higher-order deformation theory optimized via machine learning for functionally graded plates

• Autores: Yarasca, J., Mantari, J.L., Monge, J. C., Hinostroza, M.A.
• Formato: artículo
• Fecha de Publicación: 2024
• Institución: Universidad Nacional de Ingeniería
• Repositorio: UNI-Tesis
• Lenguaje: inglés
• OAI Identifier: oai:cybertesis.uni.edu.pe:20.500.14076/29155
• Enlace del recurso: http://hdl.handle.net/20.500.14076/29155; https://doi.org/10.1080/15376494.2024.2344037
• Nivel de acceso: acceso abierto
• Materia: Higher-order deformation theory; Functionally graded plates; Genetic algorithms; Machine learning; Static analysis; https://purl.org/pe-repo/ocde/ford#1.01.02

This article presents a new kind of higher-order deformation theory, called Parametric Higher-order Deformation Theory (PHDT), for the static analysis of functionally graded plates (FGPs). The novelty of the PHDT is the use of strain shape functions that are calibrated by a set of tuning parameters to approximate 3D results along the plate thickness. The tuning parameters are assumed to vary with side-to-thickness ratios and power-law indexes. In contrast to higher-order shear deformation theories (HSDTs), the PHDT is not mathematically constrained to satisfy the traction-free boundary condition on the bottom plate’s surface. The proposed plate model is based on a 5-unknown HSDT previously presented by one of the authors. The governing equations are derived from the principle of virtual works, and Navier-type closed form solutions have been obtained for simply supported FGPs subjected to bisinuisoidal transverse pressure. A general methodology that uses genetic algorithms to determine the optimal tuning parameters of PHDTs for FGPs with various side-to-thickness ratios and power-law indexes is presented. The accuracy of the PHDT is assessed by comparing the results of numerical examples with a 3D elasticity solution, HSDTs reported in the literature, and the well-known Carrera Unified Formulation. The results show that quasi-3D displacement and stress distribution are obtained using a set of tuning parameters to form adaptable strain shape functions that are optimized for the given structural problem.