Evaluation of Analytical Aeroacoustic Models for a Low-Speed Axial Ventilation System

Noise Prediction by Analytical or Numerical Models

Several analytical aeroacoustic models are applied for a compact ducted ventilation system with high rotational speed with uniformly distributed blades and vanes. RANS simulations are performed at several mass-flow rates and provide the parameters to calibrate the noise sources for the rotor and stator acoustic models. Several blade response and acoustic propagation models have been evaluated. Due to the limited numbers of blades and vanes, the cascade effects are shown to be negligible. The propagation assuming infinite duct allows capturing the cut-off frequencies but does not modify the global spectral shape of the results. The flow parameters which define the strength of interaction mechanisms are the crucial parameters for the acoustic results. For tonal noise, the rotor wake deficit at the stator leading edge can be modeled using a Gaussian shape evolution model fitted from RANS extracted data. This model provides a satisfactory estimation and a good trend for flow rates above design but is not robust for the lowest flow rates investigated. The third blade passing frequency is overestimated by 8 dB with the present wake evolution model and with an isolated airfoil response with in-duct propagation. For broadband noise, turbulence ingestion by the rotor, turbulent rotor wake interaction with the stator and scattering of the turbulent eddies in the boundary layer at the trailing-edges of the rotor and the stator rows are investigated. The latter mechanism on the rotor is found to be dominant as the high rotational speed of the machine and the strong loadings induce high shear stress and pressure gradient on the blade surfaces. The global trend of the acoustic power evaluations with mass flow variation for the broadband noise levels estimated from the four contributions agree with current experimental investigations. For Strouhal numbers, based on fan diameter and tip velocity, above 3, the broadband noise spectrum is predicted within 3 dB accuracy with isolated response and free-field propagation.