Improved methods for prediction of electromagnetic noise radiated by electrical machines

Abstract

Improved methods for predicting the electromagnetic noise radiated by electrical machines are presented. The preferred method is based on the calculation of the sound-pressure and stator-vibration velocity distributions on a finite-cylindrical acoustic model. It results in a simple formula for calculating the acoustic power from the relative sound intensity and the stator-surface vibration, and is therefore eminently suitable for use during routine design. The method eliminates the need to calculate the sound-pressure and air-velocity distributions around a machine, as is required for existing methods. It can also account for the effect of an axial variation of the radial vibration. In an alternative method the relative sound intensity is calculated from a spherical acoustic model in which the radial vibration of a machine is expressed as a function of the circumferential mode order in the form of a Fourier series rather than a Legendre series, as in previous analyses. This makes it possible to use Fourier series throughout the various stages leading up to the prediction of electromagnetic noise. Predictions of the relative sound intensity from the different models are compared, and the sound-pressure distribution and the acoustic power are investigated both theoretically and experimentally on a TEFC 3-phase induction motor. It is shown that methods based on a spherical acoustic model are only geometrical and physical approximations, whilst a finite-cylindrical acoustic model is more generally applicable.