基于SA-BP神经网络的直线电机优化设计  

Optimal Design of Linear Motor Based on SA-BP Neural Network

作　　者：郭凯[1] 李昊 李彪[1,3] 梁楠楠 GUO Kai;LI Hao;LI Biao;LIANG Nannan(School of Mechanical and Electrical Engineering,Suzhou University,Suzhou 234000,China;School of Information Engineering,Suzhou University,Suzhou 234000,China;School of Information and Control Engineering,China University of Mining and Technology,Xuzhou 221116,China)

机构地区：[1]宿州学院机械与电子工程学院,安徽宿州234000 [2]宿州学院信息工程学院,安徽宿州234000 [3]中国矿业大学信息与控制工程学院,江苏徐州221116

出　　处：《太原学院学报(自然科学版)》2025年第2期45-52,共8页Journal of TaiYuan University:Natural Science Edition

基　　金：安徽省科研编制计划重点项目(2023AH052233);宿州学院重点项目(2023yzd15,2023yzd13,2022yzd07);宿州学院质量工程结余项目(szxy2023jyjf78)。

摘　　要：针对永磁直线同步电机推力波动大、有限元仿真计算时间长等问题,提出了一种结合解析算法(SA)和BP神经网络算法的电机仿真优化模型:依据电机各部件的磁导率不同划分解析域,由解析算法算出电磁场分布等电机参数,利用解析获得的电机性能参数建立BP神经网络训练样本库,设计BP神经网络算法的训练周期、衰减率等参数后进行模型训练,拟合预测出电机尺寸参数与定位力之间的关系模型,最后利用多目标优化算法优化电机的尺寸参数。实验结果表明:基于SA-BP神经网络的电机模型的推力计算结果与有限元仿真结果的误差为2.35%,SA-BP神经网络算法不仅具有较高的计算精度,还能有效提升电机仿真计算速度。Aiming at the problems of large thrust fluctuation and long finite element simulation calculation time of permanent magnet linear synchronous motor,a motor simulation optimization model combining subdomain analysis(SA)and BP neural network algorithm was proposed.The analytical domain was divided according to the different magnetic permeability of each motor component,and the motor parameters such as electromagnetic field distribution were calculated by the analytical algorithm.The BP neural network training sample library was established by using the analytically obtained motor performance parameters.After the training period and attenuation rate of the BP neural network algorithm were designed,the model was trained,and the relationship model between the motor size parameters and detent force was fitted and predicted.Finally,a multi-objective optimization algorithm was used to optimize the size parameters of the motor.The experimental results showed that the error between the thrust calculation results of the motor model based on SA-BP neural network and the finite element simulation results was 2.35%.This indicates that the SA-BP neural network algorithm not only has high calculation accuracy,but also can effectively improve the motor simulation calculation speed.

关 键 词：永磁直线同步电机 解析算法 BP神经网络算法 定位力 多目标优化算法 

分 类 号：TM351[电气工程—电机]