基于BP神经网络的X频段功率分配器优化设计

针对电大尺寸目标的高功率分配器电磁特性模拟难、对计算结果精度要求高等特点，利用BP(Back Propagation)神经网络良好的非线性模拟能力，优化设计了一种电大尺寸的X频段四路高功率分配器，与HFSS仿真相比较减少了运算时间。采用Agilent矢量网路分析仪N5244A测试表明，在92～9.6 GHz频率范围内，各路损耗均优于6.8 dB,相位差为2°，与BP神经网络模拟结果吻合较好，为电大尺寸的微波电路工程设计提供了参考。

Optimization Design of X-band Power Dividers Based on BP Neural Network

It is difficult to simulate the electromagnetic properties of high power dividers and high precision of calculation result is required for the electrically large objects.According to above mentioned characteristics,an X-band four-way electrically large high power divider is optimally designed based on BP(Back Propagation) neural network.The computation cycle has been greatly shortened compared with HFSS simulation.The test results measured by vector network analyser Agilent N5244A indicate that the loss of every way is better than 6.8 dB,phase difference is 2° at 9.2～9.6 GHz,which are in good accordance with simulation results of BP neural network.The method provides reference for the design of electrically large microwave circuits.