Neuronal Networks to Determine Air Gap Flux Densities for High-Speed Electric Motors

High-Speed electric machines are used for electric compressors, for example in fuel cell applications. These machine designs are facing a couple of challenging design issues. High rotational speeds lead to high mechanical stresses and require sleeves to prevent rotor breakage. Further, the length of the design is limited by its eigenfrequency, which makes the design as compact as possible. From the magnetic point of view, the rotor losses are a major challenge which is depending on the electric frequency, the conductivity of the selected materials and the harmonics in the air gap. To fulfill the requirements permanent magnet synchronous machines (PMSM) are usually used. The authors of this work present a fast way to calculate the electric performance of a PMSM. Therefore, the focus is on the calculation of air gap fields for high speed machines considering the slot opening geometry, air gap length and rotor radius. Neuronal networks are trained with static finite element analysis (FEA) results, to predict the air gap flux amplitude and the slot influence on the magnetic field. The results show a very good accordance between the calculated air gap flux density and the FEA results. It is possible to get a detailed look on the air gap flux at every radial position and therefore the results are further used to determine rotor losses.