Acoustic noise prediction in a vector controlled induction machine

The deterministic approach for the prediction of noise in electrical machines is generally based on the use of FEM/BEM models. Mechanical and electromagnetic properties included in these models, such as the damping or the BH curve, are difficult to estimate, compromising the accuracy in the noise prediction. Simulations showed that a 10% increase in the saturated part of the BH curve results in an increase of up to 3 dB in the sound power level, and that the sound power level could be reduced by up to 10 dB with a 5 times increase in the damping coefficient. The accuracy of the noise prediction was tested using a vector controlled drive with a very high PWM switching frequency. This system allowed the structure to be excited with the radial and tangential forces independently. Discrepancies between predictions and measurements of the sound power level at low frequencies have been attributed to the rotor and some stator/casing modes not being modeled, while at high frequencies, discrepancies are due to skewing of the rotor not being included in the model.