基于BP神经网络的永磁直线同步电机齿槽力预估器

为了减小永磁直线同步电机（PMLSM）齿槽力的波动，采用二维电磁场有限元方法，结合动态分网来描述PMLSM定、动子间的相对运动.考虑非线性磁饱和的影响,对槽形尺寸、斜槽、闭口槽、气隙、分数槽等影响下的齿槽力进行了分析.计算结果表明，分数槽是有效减小PMLSM齿槽力波动的重要措施.将模拟计算得到的槽形、气隙对齿槽力波动的影响作为神经网络的训练样本，结合动量法和自调节学习规则，构建了基于BP神经网络的齿槽力预估器.通过该预估器，可以在PMLSM设计阶段对槽形尺寸、气隙大小进行合理的选择.

Cogging force estimator based on BP neural network of PMLSM

To reduce the fluctuation of cogging force in permanent magnet linear synchronous motors(PMLSM),a 2D-field finite element model of PMLSM was used to describe the relative movements between primary and secondary,combined with automatic meshing.The cogging force ripple was analyzed considering nonlinear magnetic saturation.In order to investigate the mechanism of cogging force,the force ripples affected by slot shapes,skew effects with various angles,close slots,air gaps and fractional slots were compared respectively.The results showed that the fractional slot was the most efficient way to diminish the cogging force.The suggested estimator based on BP neural network was set up using momentum method and self-adapting learning rate rule,where the training sets were obtained from FEM calculations.The estimator will evaluate the selections of slot shape and air gap in the design.