【作者】 姚亮； 韩雪； 蒋瑞琦； 张银伟； 栾天； 陈隆； 连培园； 王从思；

【Author】 YAO Liang;HAN Xue;JIANG Ruiqi;ZHANG Yinwei;LUAN Tian;CHEN Long;LIAN Peiyuan;WANG Congsi;Key Laboratory of Electronic Equipment Structure Design,Ministry of Education;China Electronics Technology Group Corporation No.54 Research Institute;Guangzhou Institute of Technology,Xidian University;

【机构】 西安电子科技大学机电工程学院； 中国电子科技集团公司第五十四研究所； 西安电子科技大学广州研究院；

【摘要】 大型反射面天线受复杂环境载荷影响会发生结构变形影响观测性能,需要将副面移动到和变形主面相匹配的位置,针对大型天线副面初始位姿实时确定问题,本文提出了一种基于深度学习和应变测量数据的副面初始位姿快速预测方法。在撑腿上安装应变传感器采集副面撑腿应变数据,建立基于应变数据的深度神经网络副面初始位姿预测模型,结果表明,9 0%以上的数据预测误差低于5%,实现了对副面初始位姿的快速预测。同时考虑传感器测量误差问题,基于去除极值的滑动平均法对所采集数据进行滤波处理后使用副面初始位姿快速预测方法进行预测分析,最大预测误差低于8%,验证了去除极值的滑动平均法的有效性以及所提方法的工程可行性。更多还原

【Abstract】 Large reflector antennas may undergo structural deformation due to complex environmental loads,which affects their observation performance.Therefore,it is necessary to move the subreflector to a position that matches the deformed main reflector.In response to the problem of real-time determination of the initial position and attitude of the subreflector of large antennas,this paper proposes a fast prediction method for the initial position and attitude of the subreflector based on deep learning and strain measurement data.Install strain sensors on the support legs to collect strain data from the support legs of the subreflector,and establish a deep neural network model for predicting the initial position and attitude of the subreflector based on strain data.The results show that more than 90% of the data has a prediction error of less than 5%,achieving rapid prediction of the initial position and attitude of the subreflector.At the same time,considering the issue of sensors measurement error,a sliding average method based on removing extreme values was used to filter the collected data,and then a rapid prediction method for the initial position and attitude of the subreflector was used for prediction analysis.The maximum prediction error was less than 8%,verifying the effectiveness of the sliding average method removing extreme values and the engineering feasibility of the proposed method.更多还原