Metamodel-assisted metaheuristic for structural optimization problems

Resumo

Optimization problems are common in many different areas, especially engineering. The complexity of modern problems has led to the development of increasingly complex mathematical models, resulting in expensive simulation models. An alternative for solving these problems is population-based metaheuristics, especially those of natural inspiration. However, they usually require many evaluations to obtain a feasible or even satisfactory solution. In this context, the application of metamodels, or surrogate models, together with metaheuristics, has received the growing attention of researchers in several areas. The metamodels generate a simpler computational model to be used in parts of the optimization process, replacing the original model. This work presents an application strategy of metamodels within metaheuristics, which allows for computational cost reduction. The methodology is applied to structural optimization problems, demonstrating its applicability and establishing it as an alternative to improving solutions in the context of fixed-budget simulations.