Session: 04-11-01 Posters

Paper Number: 194413

194413 - Multi-Fidelity Design of 3d Backward Centrifugal Fan Using Vorticity Correlation 

Abstract: 

In the research a multi-fidelity strategy for generating 3D centrifugal fan blade geometry in the design space is implemented, and an integrated design software is developed to achieve optimal design configurations for various specific speeds. Our approach is based on combining selected high-fidelity CFD simulations with a large number of lower-fidelity computationally inexpensive models to achieve the maximum Merit Index based on both the circulation strength and the fan static efficiency near an operation point. While the high-fidelity CFD for surrogate models can provide detailed physics of unsteady vortex flows and aero-acoustic noise generation sources in detail, it is too computationally demanding for performing optimization or uncertainty quantification (UQ) studies over the range of parameters. This is because performing optimization and/or UQ requires a large number (often in the order of 1,000-10,000) of simulations for the fan operating at a single specific speed. In order to generate low-fidelity CFD models satisfying the proposed performances given as inputs, a 2D meanline design method is combined with total pressure and sweep-angle distribution functions set by the Taguchi method allowing 3D streamlines, which is equivalent to a 2.5D design method. The outcome model from the low-fidelity optimization is found to have a multi-curvature blade surface resulting in the high energy transfer accompanying with a strong vortex strength with the specified exit height, not near the backplate on the contrary. The high-fidelity optimization model is searched through a reinforced machine-learning A.I. algorithm with surrogate models. This strategy significantly reduces the overall design cost while maintaining design quality, thus providing a practical and robust framework for optimizing centrifugal fan performance in terms of both efficiency and noise reduction.

Keywords: Multi-fidelity Design, Centrifugal Fan, Surface Morphing, CFD  

Presenting Author: Jae-Kwon Shin Inha Univ.

Presenting Author Biography: Master Degree Graduate Student
in Mech. Eng. Dept. at Inha Univ.

Authors:

Jae-Kwon Shin Inha Univ.
Tae-Young Yeom Inha Univ.
Kyung-Hak Lee Inha Univ.
Seungbae Lee Inha Univ.