基于CNN的卫星内部结构频域响应间接测量方法  

An indirect method for measuring frequency domain response of satellite internal structure based on convolutional neural network

作　　者：彭瑾坤 武文华[1,2] 顾乃建 胡奇 PENG Jinkun;WU Wenhua;GU Naijian;HU Qi(State Key Laboratory of Structural Analysis,Optimization and CAE Software for Industrial Equipment,Dalian University of Technology,Dalian 116024,China;Ningbo Research Institute of Dalian University of Technology,Ningbo 315000,China;Beijing Institute of Spacecraft Environment Engineering,Beijing 100029,China)

机构地区：[1]大连理工大学工业装备结构分析优化与CAE软件全国重点实验室,大连116024 [2]大连理工大学宁波研究院,宁波315000 [3]北京卫星环境工程研究所,北京100029

出　　处：《航天器环境工程》2023年第4期400-407,共8页Spacecraft Environment Engineering

基　　金：国家重点研发计划项目(编号:2021YFA1003501);深圳市自由探索类基础研究项目(编号:2021Szvup021);辽宁省兴辽英才计划项目(编号:XLYC2002108);大连市支持高层次人才创新创业项目(编号:2021RD16)。

摘　　要：针对生产线卫星脉动式快速生产的要求,为解决力学试验步骤烦琐且内部结构响应测量困难的问题,提出基于卷积神经网络(CNN)的卫星内部结构响应深度学习间接测量方法。分别对生产线I型卫星与成熟的Ⅱ型卫星进行正弦扫频试验,提取结构加速度频域响应信息,并利用基于CNN建立内部结构响应的间接测量模型验证所提出方法的可行性。结果表明该方法对两种型号卫星内部结构响应的间接测量总体精度分别达到了95.8%与96.9%,具有较强的工程应用潜力。In view of the requirement of pulsating rapid production for production line satellite,in order to solve the problems of tedious mechanical test steps and difficulty to measure the internal structural response,a deep-learning-based indirect measurement method based on convolutional neural network(CNN) was proposed.A sine sweep test was conducted on a production line satellite I and a mature satellite Ⅱ,respectively,to extract the frequency domain response of the structural acceleration.Then an indirect model based on CNN for measuring the internal response was established to verify the feasibility of the proposed method.The results show that the indirect measurement accuracy with the method reaches 95.8% and 96.9% for the two types of satellites,respectively,which indicates that the proposed method may have a strong potential for engineering applications.

关 键 词：生产线卫星 内部结构响应 间接测量 正弦扫频 卷积神经网络 

分 类 号：V416.2[航空宇航科学与技术—航空宇航推进理论与工程] TP183[自动化与计算机技术—控制理论与控制工程]