Otimização estrutural de roda-gigante acessível por algoritmos genéticos com avaliação baseada em desempenho

Abstract

The development of accessible Ferris wheels with adequate structural performance and competitive cost remains a significant challenge in the Brazilian context, characterized by a scarcity of applied technical literature and reliance on imported equipment. This dissertation proposes an integrated computational methodology for the design, modeling, and structural optimization of an accessible ferris wheel, combining the finite element method with genetic algorithms in a Python environment. A parametric finite element model was implemented and coupled to the ANSYS solver through the PyMAPDL library, enabling automated geometry generation, meshing, loading application, and post-processing. In an initial deterministic stage, a proprietary genetic algorithm was adapted for discrete optimization of commercial sections, investigating genetic operators and execution parameters with the objective of minimizing structural mass while satisfying safety and integrity requirements. Complementarily, a performance-based probabilistic approach was adopted, in which uncertainties associated with wind, occupancy, and structural parameters were propagated via Monte Carlo simulations, structured according to intensity measures and system parameters and evaluated using probabilistic metrics. From the constrained optimization implemented in the genetic algorithm, a deterministic design with 14 cabins was obtained, approaching operational limits of peak displacement and comfort. As a contribution, this work delivers an automated, reproducible, and extensible workflow for the optimization and performance assessment of Ferris wheels, providing guidelines for lighter, safer designs aligned with normative and operational criteria, with the potential to support the national development of accessible recreational structures. In addition, the study advances the state of the art by integrating genetic algorithms and the finite element method in an automated manner, offering a replicable model for future applications in large-scale structures.