Minimisation of NVH of a Permanent Magnet Synchronous Machine with the Aid of Lookup Tables and Neural Networks in Stationary and Dynamic Operating Points

Minimisation of NVH (noise, vibration, harshness) is an important issue in many areas. In both electric vehicles and household devices, any noise caused by electrical machines is not desirable. NVH can be reduced by designing electrical machines appropriately. Such designs are usually expensive and complex. In this paper, a method is introduced to minimize electric machine NVH by injecting flux-linkage harmonics. With the help of the injection method, the NVH of any electrical machine can be reduced. The reduction of NVH is shown with the example of a permanent magnet synchronous machine (PMSM). Two concepts are presented: the reduction of NVH by using lookup tables (LUTs) and a neural network (NN). Both methods can be used to reduce and minimise NVH at stationary operating points and during acceleration and deacceleration. With regard to several input variables, the NN performs significantly better than LUTs.