Dataset for analysis of computational fluid dynamics results from an industrial centrifugal fan.

Abstract

This dataset comprises 504 observations obtained from numerical simulation using Computational Fluid Dynamics for a real industrial application involving a turbomachine, specifically a centrifugal fan responsible for air recirculation in ovens operating at high angular velocity and temperature. Prior to geometry construction, a Design of Experiments strategy was employed, based on a Central Composite Design to generate a response surface. Four geometric input variables related to the centrifugal fan blades were considered: number of blades, blade entry angle, blade opening angle, and blade length. The design comprised 24 simulation trials, which, combined with 20 replicates generated by small mesh variations, resulted in 504 numerical simulations. The entire process was executed in Ansys®, with Workbench used to organize and manage the project workflow from pre-processing to post-processing. Geometry modeling was carried out in SpaceClaim, while mesh generation was performed in Fluent Meshing using polyhedral elements, local refinements for blade throat regions and high-gradient zones, and quality control ensuring orthogonal quality above 0.150 and skewness below 0.950. Simulations were conducted under steady-state conditions with a Multiple Reference Frame at 5600 revolutions per minute, employing the k-ω Shear Stress Transport turbulence model and a Coupled solver scheme. Final analyses were conducted in Fluent and CFD-Post. The simulations produced 35 output variables, encompassing static, dynamic, and total pressures; velocity magnitude and its axial, radial, and tangential components; turbulence quantities; turbulent Reynolds number; Large Eddy Simulation resolution estimate; torque; mass flow rate; blade mass; material cost; and a dimensionless performance index. For certain variables, minimum, mean, and maximum values are reported. The structured workflow and dataset documentation promote reuse and reproducibility. This dataset is suitable for statistical analysis, response surface modeling and simulation, sensitivity studies, comparison of turbulence models and mesh configurations, design optimization, data science applications, and replication of the Design of Experiments methodology in other Computational Fluid Dynamics contexts.