Eventos Anais de eventos
MECSOL 2022
8th International Symposium on Solid Mechanics
Surrogate-Based Optimization of Functionally Graded Plates using Multi-Fidelity Models
Submission Author:
Evandro Parente Junior , CE
Co-Authors:
Leonardo Gonçalves Ribeiro, Evandro Parente Junior, Antônio Macario Cartaxo Melo
Presenter: Evandro Parente Junior
doi://10.26678/ABCM.MECSOL2022.MSL22-0022
Abstract
Optimization methods can be employed to find the optimal material designs in Functionally Graded (FG) structures. This is often performed by the use of bio-inspired algorithms, even though these may require a large number of function evaluations. For a more efficient process, surrogate models can be used to provide a cheaper estimate for the structural response. Also, on structural optimization of complex structures, analysis models with multiple levels of fidelity (via coarser mesh discretization or simplification of analysis theory) can be easily created. Thus, Multi-Fidelity models can be employed for a more accurate approximation. To further increase the optimization process effectiveness, the Sequential Approximate Optimization (SAO) can be employed, where the approximate surface is iteratively improved by the addition of new points in regions of interest. In this work, SAO will be employed in the optimization of Functionally Graded Plates. The multi-fidelity Hierarchical Kriging model will be employed. For comparison purposes, results using the single-fidelity Kriging model will be shown. These approaches will be compared in terms of efficiency and accuracy. Results show that the Hierarchical Kriging can greatly reduce the number of expensive evaluations required to find the optimal material gradation.
Keywords
structural optimization, Surrogate models, Composites, Functionally graded materials
