Tonal and Broadband sound prediction of a locomotive cooling unit

Abstract

The present work is dedicated to the understanding, modeling and eventually prediction of the noise generated by a laboratory-scale mock-up representative of a generic locomotive cooling unit. The geometry is relatively simple, but a difficulty in terms of modeling and simulation is the inclusion of the acoustic and aerodynamic installation effects taking place between a typical protection grid, a heat exchanger and a low-speed cooling fan combined in a parallelepipedic box setup. In our case the grid and heat exchanger, combined with a 90-degrees turning of the mean flow direction, are affecting the noise generated by the fan rotor through aerodynamic and acoustic installation effects. This study includes experimental and simulation work aimed at measuring and predicting such installation effects. On the experimental side, adding or removing components has permitted assessing the resulting changes quantified as delta-dB levels in the far-field. These far-field measurements reveal that the tonal component of the emitted spectrum, of relatively modest amplitude above the broadband noise, is the most strongly affected by the turning angle imposed by the mock-up geometry, but is not much altered by the presence of the grid and heat exchanger. The broadband component of the measured spectra shows some sensitivity to the presence of the grid and heat exchanger, presumably related to the alteration of the structure of the turbulence ingested by the fan, and due to acoustic absorption through the heat exchanger. On the simulation side, a first attempt is made to include the various components in a complete simulation chain. A semi-analytical approach is used to model the acoustic radiation, limited in this study to the contribution from the rotor blades subjected to turbulence interaction, consistently with the experimental observations. Some of the required input data include the incoming turbulence intensity and correlation length, provided in the present case by a RANS simulation of the mock-up. In this simulation, the heat exchanger is represented as a source term of the momentum equation, adjusted to yield the required pressure drop through this element. This model presents the advantage of simplicity, however comparisons between the predicted and measured turbulence intensities upstream of the rotor plane indicate that a more elaborated model would be necessary to correctly describe the alteration of turbulence through the narrow channels of such automotive heat exchanger. For the acoustic scattering part, the incident broadband acoustic field is scattered on the mock-up geometry using a specific numerical methodology based on a Boundary Element Method already validated by the present authors on simpler geometries. The heat exchanger is there represented by a lumped model characterized through transfer admittance matrix. The comparison between the measured and predicted acoustic fields indicate that a reasonable match is found as far as the spectral distribution of energy is concerned, however with overall levels that are substantially under-predicted. Likely explanations for the discrepancies stands in the fact that potentially important sources of noise have been neglected such as the rotor trailing-edge noise, stator noise or the noise associated to the complex vortical motion that develops between the shrouded rotor ring and the casing. More investigations will be required as well about the structure of the turbulence ingested by the rotor.