The aeroacoustics of a simplified alternator are analysed within the framework of this thesis, with the objective being the development of a tool capable of numerically predicting aeroacoustic noise. Focus has been placed on tonal noise and only dipole, rotating noise sources have been modelled. Input for these dipole sources are calculated using instationary CFD computations using the SST turbulence model. The acoustic tool developed is based on Lowson's equation, derived from the Ffowcs-Williams and Hawkings equation. Modifications to the formulation are included in order to model the reflective plane present during measurements and to ensure that sound propagates solely away from source surfaces (a half-dipole propagation). Numerical results are compared with experimental measurements, and show discrepancies in total sound power level. This is due to the inability of the methods used to model broadband noise sources that produce energy between the main orders (harmonics) of the alternator. Determining the total energetic content of the sound orders numerically calculated and comparing this to the energetic content of the measured orders yielded positive results; sound power level results were within 2 dB(A) of measurement. Dominant frequencies/orders were calculated generally within an accuracy of 6 dB(A); issues with the periodic boundary conditions and the rotor/stator interface produced spurious fluctuations that, for one tested case, produced strong 8th order noise. The developed tool was used successfully in a real alternator model, and once again showed very promising results. Fan orders were calculated within very good accuracy, while clawpole orders showed too much energy at higher orders due to the simplified CFD modelling of the casing holes. Total sound power level was computed too low due, once again, to the lack of broadband noise computation. Total energy from the computed orders was however calculated well within the range of the measured samples. The modelling of the casing using the BEM LMS Sysnoise was not possible due to the inability within the software to specify both pressure gradient and pressure boundary conditions, however models simulated using the FEM tool COMSOL Multiphysics showed that the in uence of the casing on most orders, with the exception of the 40th and those above the 60th, was low, in addition to a low effect on the total sound energy.
展开▼
