一种在线数据驱动的永磁直线电机优化设计

针对电机优化设计是一个非线性度高、强耦合、计算成本高的问题，文中介绍了一种在线数据驱动的电机优化设计方法。该设计方法采用kriging代理模型快速获得电机的近似数学模型，通过期望提高的在线加点策略实现高精度模型，结合多目标粒子群算法最终获得电机多目标优化的Pareto前沿。在线数据驱动的优化设计原始样本少，且能以最少数量的样本加点来改善模型精度，大大减少了计算成本。该法用于一台12槽10极永磁直线电机，结果显示优化的电机较原始其推力波动降低了64%且平均转矩提高了6%，计算周期缩短了60%。此优化设计方法同样适用于其它优化问题，具有一定的参考价值。

An Optimization Design of Permanent Magnet Linear Motor Driven by Online Data

For the problem of optimization design of permanent magnet synchronous motor (PMSM) is high nonlinearity, strong coupling and high calculation cost, this paper introduces an online data-driven optimization design method of PMSM. In this design method, the kriging surrogate model is used to quickly obtain the approximate mathematical model of the motor, and high-precision model is achieved through the expected improved (EI) online adding point strategy. Combined with multi-objective particle swarm optimization algorithm, Pareto frontier of multi-objective optimization of motor is finally obtained. The online data-driven optimization design has fewer original samples, and can improve the accuracy of the model with the least number of samples, which greatly reduces the calculation cost. The method is applied to a 12-slot 10-pole permanent magnet linear motor. The results show that the thrust ripple of optimized motor was reduced by 64%, and the average torque was increased by 6%, and the calculation cycle was shortened by 60%. This optimization design method is also suitable for other types of problems, and has a certain reference value.