现代应用物理2025，Vol.16Issue(1)：158-164,7.DOI:10.12061/j.issn.2095-6223.202412010
基于多层次深度神经网络的相对论返波管优化技术
Optimization Method of Relativistic Backward Wave Oscillator Based on Multi-Level Deep Neural Network
摘要
Abstract
An optimization method of relativistic backward wave oscillator(RBWO)based on data-driven deep neural network is proposed.The structure or electrical parameters of RBWO are selected as the parameters to be optimized,and the corresponding working characteristic parameters under different parameters are generated by the fully electromagnetic particle-in-cell code,and the low-dimensional training data set and the high-dimensional training data set are generated.A multi-level deep neural network is developed,and the output of the low-level deep neural network is adopted as the input of the high-level deep neural network to realize the interconnection of the neural networks.Numerical examples show that the relative deviation between the predicted result of multi-level deep neural network and simulated results of fully electromagnetic particle simulation is less than 2％,indicating that they are in good agreement.This method addresses the challenge of achieving high prediction accuracy with limited sample data in deep neural networks,enabling high-precision optimization results.关键词
数据驱动/相对论返波管/全电磁粒子模拟算法/多层次深度神经网络Key words
data-driven/relativistic backward wave oscillator/fully electromagnetic particle simulation/multi-level deep neural network分类
信息技术与安全科学引用本文复制引用
陈再高,史雪婷,王建国,梁闪闪,唐泽华,陈柯,杨超..基于多层次深度神经网络的相对论返波管优化技术[J].现代应用物理,2025,16(1):158-164,7.基金项目
国家重点研发计划资助项目(2020YFA0709800) （2020YFA0709800）
