Design of Nanocrystalline Alloy for Reducing Motor Hysteresis Loss by Machine Learning

Hysteresis loss, the main component of core loss, is directly related to the motor core material. A nanocrystalline alloy with low hysteresis loss induced by low coercivity is developed by machine learning (ML). Firstly, the eXtreme Gradient Boosting (XGBoost) decision tree algorithm is employed to design amorphous alloys with good soft magnetic properties. Subsequently, the alloy Fe77.5Co3.8B15P1.7Si1Nb1designed is prepared by melt spinning and annealed at 470 °C for 180 s. The obtained coercivity (Hc) of the nanocrystalline alloy is 2.84 A/m, and the saturation magnetic flux density (Bs) is 1.88 T. Finally, hysteresis loss is investigated by the finite-element analysis of the motor with this nanocrystalline alloy stator core and compared with the results of the DW315-50 stator core motor. It is revealed that the hysteresis loss of the motor by the annealed alloy has been reduced by around 88% compared with the silicon steel stator core. It is indicated that the lowHcnanocrystalline alloy designed by machine learning can effectively reduce the hysteresis loss of the motor, which is of great significance for industrial energy conservation.