An Efficient Identification Method for Orthotropic Material Parameters of Stator Cores With Arbitrary Structures Based on General Surrogate Model

Accurate orthotropic material parameters (OMPs) of stator cores are fundamental for their modal analysis, as well as crucial for improving the vibration and noise performance of motors. However, there is still no effective method to rapidly obtain the accurate OMPs of several stator cores with different structural characteristics. In this paper, a method based on a general surrogate model is proposed. First, the structure-material conceptual model for the natural modal features of stator cores is established by extracting effective structural parameters (ESPs) and material parameters. Then a general surrogate model for the natural modal features is established using convolutional neural network based on genetic algorithm optimization (GA-CNN). Subsequently, the identification of OMPs for stator cores with any structures is achieved using parallel multi-population particle swarm optimization (PMPSO). Moreover, it allows for the simultaneous identification for several stator cores. Finally, the effectiveness of this method is verified by FEM and modal experiments. Compared with experimental results, the maximum relative error in natural frequencies of the four prototypes is within 3%, and the identification process for these four prototypes takes less than one minute. The proposed method, with its high accuracy and efficiency, contributes to the precise prediction and control of electromagnetic noise in electric motors.