Surrogate-based evolutionary optimization of aerodynamic performance and emitted sound for an impeller fan
Eng. Appl. Comput. Fluid Mech., vol. 19, pp. 2585337 (2025)
@article{zoka-2025,
title = {Surrogate-based evolutionary optimization of aerodynamic performance and emitted sound for an impeller fan},
author = {Zoka, H. M. and Moxey, D. and Tabor, G.},
journal = eacfm,
year = {2025},
volume = {19},
number = {1},
pages = {2585337},
doi = {10.1080/19942060.2025.2585337},
url = {https://www.tandfonline.com/doi/full/10.1080/19942060.2025.2585337},
abstract = {Centrifugal fans find widespread application in building ventilation and air-conditioning systems. Nonetheless, they face challenges in meeting the growing need for energy efficiency, better performance, and reduced noise levels. In this paper, we examine simultaneous aerodynamics and aeroacoustics optimization of a backward-curved centrifugal fan. The study begins with the validation of 3D computational fluid dynamics (CFD) models for both aerodynamic and aeroacoustic simulations against experimental data. Using these validated models, two separate simulation domains are employed for aerodynamic and aeroacoustics analyses, and a selection of CFD simulations performed using a Latin HyperCube sampling. Based on this data, surrogate models were trained independently to predict aerodynamic and aeroacoustics behaviour based on five geometric design parameters, including the blade flow angle distribution from leading edge (LE) to trailing edge (TE). We then perform three optimization scenarios—purely aerodynamics, purely acoustics, and combined aerodynamics-aeroacoustics using the surrogate-based NSGA-II optimization algorithm and compare the optimal results against the baseline geometry. We show that aerodynamics optimization demonstrates up to an 8\% increase in efficiency within the operating range, while the acoustics optimization achieved a reduction of up to 4 dB in the maximum sound pressure level, at the expense of significantly reduced flowrates. The aeroacoustics optimization yielded a balanced outcome, maintaining baseline flowrates while delivering a 3.9\% increase in efficiency and a 3.5 dB reduction in maximum sound pressure level. These results underscore the trade-offs between aerodynamics performance and acoustics behaviour and highlight the potential of optimization strategies tailored to specific operational goals.}
}
Centrifugal fans used in building ventilation have to become more efficient and quieter at the same time. CFD models for both the aerodynamics and the aeroacoustics of a backward-curved fan are first validated against experimental data, then used to train surrogate models so that the two objectives can be optimised together.
Abstract
Centrifugal fans find widespread application in building ventilation and air-conditioning systems. Nonetheless, they face challenges in meeting the growing need for energy efficiency, better performance, and reduced noise levels. In this paper, we examine simultaneous aerodynamics and aeroacoustics optimization of a backward-curved centrifugal fan. The study begins with the validation of 3D computational fluid dynamics (CFD) models for both aerodynamic and aeroacoustic simulations against experimental data. Using these validated models, two separate simulation domains are employed for aerodynamic and aeroacoustics analyses, and a selection of CFD simulations performed using a Latin HyperCube sampling. Based on this data, surrogate models were trained independently to predict aerodynamic and aeroacoustics behaviour based on five geometric design parameters, including the blade flow angle distribution from leading edge (LE) to trailing edge (TE). We then perform three optimization scenarios—purely aerodynamics, purely acoustics, and combined aerodynamics-aeroacoustics using the surrogate-based NSGA-II optimization algorithm and compare the optimal results against the baseline geometry. We show that aerodynamics optimization demonstrates up to an 8% increase in efficiency within the operating range, while the acoustics optimization achieved a reduction of up to 4 dB in the maximum sound pressure level, at the expense of significantly reduced flowrates. The aeroacoustics optimization yielded a balanced outcome, maintaining baseline flowrates while delivering a 3.9% increase in efficiency and a 3.5 dB reduction in maximum sound pressure level. These results underscore the trade-offs between aerodynamics performance and acoustics behaviour and highlight the potential of optimization strategies tailored to specific operational goals.