Prediction of acoustic noise from variable-speed induction motors: deterministic versus statistical approaches

The traditional approach for the prediction of noise radiated from induction motors is usually based on a deterministic approach in which the motor is modeled using finite-element and/or boundary-element methods (FEM/BEM). A great deal of structural details has to be modeled using the deterministic approach. While these deterministic methods have been shown to provide reasonable estimates of the radiated acoustic noise at low frequencies (usually, well below 2 kHz) and for small motors, the demands on computing memory and time render calculations at high frequencies and for large motors almost impossible to make. In this paper, a statistical method especially suited for calculations at high frequencies is introduced. The statistical approach does not require the same amount of details to be modeled as in the FEM/BEM because only the averaged noise within a given frequency band is sought. Furthermore, it utilizes the analytical results obtained from simple structural elements that make up the motor structure. Results demonstrate the feasibility of the method, its ability to discern the contribution of various components of a motor to the overall radiated noise, and substantial savings in the computational effort compared with FEM/BEM.