神经网络法优化接触器电磁系统

将人工神经网络优化方法引入接触器电磁系统的优化设计中,利用人工神经网络强大的非线性逼近能力和高效率的并行计算功能进行优化.采用改进BP算法前馈神经网络法,建立了接触器反力特性中拐点位置的动态吸力和激磁电流为输入神经元,以电磁系统的尺寸作为输出神经元的前向三层的神经网络,并以CJX1型接触器为样本,通过变步长法训练神经网络.网络训练成熟后,可对接触器的外型尺寸进行优化.采用神经网络法减少了大量的三维电磁场计算,可以得到很高的优化速度,很快得到结果.经优化后的接触器可减小电磁系统的体积.

Optimization of the Electromagnetic System of Contactor by Neural Network

The electromagnetic system decides the size of contactors. So the size of electromagnetic system is often considered as optimization aim. However, the optimization of electromagnetic system is usually involved in the intricate electromagnetic computation. So in this paper, the electromagnetic system of contactor is optimized by neural network which has greatly nonlinear convergence and high efficient parallel computation. The three-layer forward network applies the improved BP arithmetic, in which the coil currents and attraction forces at four sport places of spring force are inputs nodes, the sizes of structure are output nodes and five nodes at layer 2 and layer 3. The neural network is trained by the CJX1 contactor in variational step. Once the train of neural network is finished, the size of electromagnetic system can be optimized quickly without the intricate electromagetic field computation. After optimization, the sizes of a contactor are reduced greatly.