基于仿真及神经网络的大型电磁脉冲模拟器近区场计算

Calculation of Near-Field of Large-Scale Electromagnetic Pulse Simulator Based on Simulation and Neural Network

本文针对大型电磁脉冲模拟器试验区域电场分布测量,设计了一种基于传递函数的近区场波形预估方法.该方法使用全波仿真获取初始传递函数,研究了基于频域数据的神经网络训练方法,建立了可以计算特定区域任意测点传递函数的神经网络模型.利用两套测量系统同时进行测量验证,结果表明:在40 m×40 m×10 m的范围内,该计算方法可以基于单一测点的实测结果计算任意点的电场波形,且基于该方法计算的电场波形与实测波形主波形幅值差异均小于3%,实测验证波形结果与预估结果基本一致.文中方法在满足试验区域为线性时不变(Linear Time Invariant,LTI)系统的情况下,可以应用于大型电磁脉冲模拟器的近区场快速预估计算.

Aiming at the large-scale electromagnetic pulse simulator,this paper designs a near-field waveform calcu⁃lation method based on the transfer function.This paper uses full-wave simulation to obtain the initial transfer function,studies the neural network training method based on frequency domain data,and establishes a transfer function neural net⁃work model that can calculate any measurement point in a specific area.The results of simultaneous measurement using two sets of measurement systems show that within a range of 40 m×40 m×10 m,this calculation method can calculate the elec⁃tric field waveform at any point based on the actual measurement result of a single measurement point.The difference be⁃tween the electric field waveform calculated based on this method and the measured waveform amplitude is less than 3%,so the measured verification results are consistent with the theory.In the case where the method meets the area as a linear time invariant system,the method can be applied to a near-field fast calculation of a large-scale electromagnetic pulse simulator.