Resumo:
This study presents the combination of computational fluid dynamics, design of experiments, and multi-objective optimization to assist decision-makers in selecting an ideal and feasible geometric configuration according to their preferences. This hybrid approach enables the automatic selection of Pareto-optimal solutions. The proposed method utilizes response surface models, classification metrics, and posterior optimization techniques to refine parameters and identify the most suitable solutions within the solution space. The methodology is exemplified through a case study of a centrifugal fan made of AISI 304 material, used for air circulation in an industrial oven operating at high temperatures. The approach covers individual optimization based on the original responses, followed by model validation through new simulations, comparing the results obtained using Ansys Fluent® and assessing the percentage gains relative to the original fan. The input variables refer to the fan blades, while the response variables are related to mass, mass flow rate, torque, performance, cost, speed, pressure, and turbulence. The most significant improvements were observed in blade mass, which was minimized by 65.8%, magnitude of maximum velocity, which was maximized by 23.7%, and mass flow rate, which was maximized by 19.9%. For comparison and assessment purposes, the results obtained are also compared with the non-dominated sorting genetic algorithm and machine learning techniques, revealing suitable performance.
