中国空间科学技术

中国空间科学技术 ›› 2022, Vol. 42 ›› Issue (5): 57-64.doi: 10.16708/j.cnki.1000-758X.2022.0067

• 论文 • 上一篇    下一篇

基于神经网络及深层充电的电子通量反演模型

周宏涛，方美华   

  1. 南京航空航天大学 航天学院，南京211100
  • 出版日期:2022-09-09 发布日期:2022-09-09

ANN inversion model for electron flux based on deep charging

ZHOU Hongtao，FANG Meihua   

  1. Institute of Aerospace，Nanjing University of Aeronautics and Astronautics，Nanjing 211100，China
  • Online:2022-09-09 Published:2022-09-09

PDF (PC)

85

摘要： 为了实现对空间高能电子通量的估计及航天器深层充放电的风险评估，基于深层充电和空间电子环境的关联性，利用人工神经网络（ANN）建立了一种由深层充电反演空间高能电子环境的模型。以深层充电探测电流密度及电子能量作为模型输入，电子通量作为模型输出，使用AE9模型对神经网络进行训练，将神经网络的MSE降低到了0.04122，并使用GioveA卫星的深层充电探测数据及GOES卫星的电子通量探测数据验证了模型反演电子环境的准确性。同时对由探测电流计算航天器典型介质材料最大内电场的神经网络模型进行了研究，以实现对航天器内充电风险实时评估。

关键词: 深层充电, 神经网络, 反演模型, 电子环境, 风险评估

Abstract:  To realize the estimation of high-energy electron flux and the risk assessment of spacecraft deep charging and discharging，an artificial neural network （ANN）for the electron flux inversion with deep charging was built based on the relation between deep charging and electron flux.The detect currents of a deep charging detector and the electron energy were taken as the model inputs，while the electron fluxes were taken as the output.AE9 was used to train the network，and the MSE of this model was reduced to 0.04122.The deep charging data from Giove-A and electron flux from GOES were used to verify the model′s accuracy.Based on this model，another ANN model was built to calculate the maximum internal electric field of the typical dielectric of spacecraft from the detection current to realize the real-time assessment of charging risk in spacecraft.

Key words: deep charging, neural network, inversion model, electronic environment, risk assessment