Investigation on accuracy of numerical simulation of aerodynamic noise of single-stage axial fan

The prediction accuracy of turbomachinery aerodynamic noise, particularly in relation to broadband noise with uncertain factors, has long been a challenging issue. Previous studies have not fully comprehended the factors influencing its prediction accuracy, lacking an objective and comprehensive evaluation method. An improved approach combining orthogonal experiment design and principal component analysis is employed to address these limitations. The evaluation method expands the noise metrics and provides a comprehensive assessment of the accuracy of numerical simulation for aerodynamic noise. The evaluation method is utilized to optimize and quantitatively analyze the impact of the refinement size of the core area on noise prediction for single-stage axial fans. Subsequently, the three metrics, namely, Z(1), Z(2), and broadband noise Z(3), are integrated using PCA to form a new integrated optimal metric Z(total). The influence of different refinement sizes, particularly on Z(total), is quantitatively examined. The findings reveal that the mesh size of the stator wake (D area) exhibits the most significant influence on noise prediction accuracy, with a calculated weight of 81.3% on noise accuracy. Furthermore, a comprehensive investigation is conducted on the influence of turbulence models and the wall Y+ value on aerodynamic noise. Detached-eddy simulation and large eddy simulation demonstrate effective capabilities in simulating both upstream and downstream turbulent flow characteristics of the stator, enabling accurate prediction of broadband noise. This study presents a set of numerical simulation schemes that achieve precise prediction of turbomachinery aerodynamic noise.