FCNN提升电大尺寸目标优化设计中RCS计算速度

FCNN improving the speed of RCS calculation in the optimization design of electrically large size target

摘要:

为在电大尺寸目标降低雷达散射截面积(RCS)的优化设计过程中提升计算目标RCS的速度，采用启发式算法对电大尺寸目标进行低RCS优化设计时，使用电磁仿真软件计算过的模型结果训练一个多层全连接神经网络(FCNN)。优化过程中，当已计算的模型数量足以完成对神经网络的训练后，使用训练好的神经网络改进电磁仿真计算。利用神经网络计算速度较电磁仿真更快的特点，实现提升电大尺寸目标低RCS优化设计速度。在本文所选的电大尺寸模型和使用模拟退火法进行优化设计的条件下，采用多层全连接神经网络改进电磁仿真计算，使模型的低RCS优化设计速度有明显提升，消耗时间从超过300 h缩减到约140 h。

Abstract:

To increase the speed of calculating the Radar Cross Section(RCS) in the optimization design process for reducing the RCS of electrically large targets, a multi-layer Fully Connected Neural Network(FCNN) is trained using the results of models calculated by electromagnetic simulation software when employing heuristic algorithms for low-RCS optimization design of electrically large targets. During the optimization process, once the number of calculated models is sufficient to complete the training of the neural network, the trained neural network is employed to improve electromagnetic simulation calculations. Leveraging the faster computational speed of neural networks compared to electromagnetic simulations, the optimization design speed for low-RCS of electrically large targets is enhanced. Under the conditions of the electrically large target model selected in this paper and the optimization design using the simulated annealing method, the use of a multi-layer fully connected neural network to improve electromagnetic simulation calculations significantly increases the speed of low-RCS optimization design, reducing the required time from over 300 h to approximately 140 h.