Acoustic noise due to electromagnetic excitations in synchronous machines

Just like their efficiency or recyclability, the acoustic noise emitted by electrical machines is part of their environmental impact. The ever-increasing proximity of human beings to this type of machine, whether in industrial applications or in the transport sector with electric vehicles or urban tramways, is prompting engineers to integrate this issue upstream in the design process.

At the same time, minimizing the cost of raw materials suggests that the mechanical structures of electrical machines are more deformable, and therefore noisier.

The overall noise level of an electrical machine comes from three main sources: mechanical noise (bearings, gears, etc.), aerodynamic or hydrodynamic noise (ventilation, turbulence, etc.) and electromagnetic noise. The latter can generally dominate other noise sources, especially at low and medium speeds where airborne noise is still low. Moreover, acoustic noise of magnetic origin is characterized by pure frequency emergences perceived as annoying, in the most sensitive frequency range of the human ear. Understanding the phenomena behind magnetic acoustic noise is therefore of prime importance in designing silent machines, or diagnosing and solving vibro-acoustic problems on existing machines.

Magnetic noise is mainly due to the creation of magnetic forces within the electrical machine, which tend to deform its magnetic parts. The deformations of the machine's magnetic parts will generate acoustic radiation perceptible to the human ear. Predicting magnetic noise therefore requires multiphysics modeling: both an electromagnetic model of the machine, to determine the excitation forces, and a vibro-acoustic model of the excited structure. In addition, the noise must be simulated in variable regime, to take account of resonance phenomena when the excitation frequencies cross those of the machine structure's eigenmodes.

The use of numerical tools such as electromagnetic and vibro-acoustic finite elements during the design phase enables very appreciable levels of accuracy to be achieved. However, this type of modeling raises difficulties, both in terms of calculation time and digital coupling between the various models. Analytical electromagnetic, vibratory and acoustic models have therefore been developed. Their speed enables them to be coupled with optimization algorithms to find the optimum trade-offs between objectives such as minimizing magnetic noise and maximizing electromechanical performance.

We will begin by characterizing all the phenomena likely to produce audible magnetic noise in synchronous machines, fractional or not, with permanent magnets, surface or buried, or with wound rotor. Variable reluctance machines, which have their own electrotechnical...