LAMOST有限元模型的径向基网络修正法

发布时间：2018-01-08 11:13

本文关键词：LAMOST有限元模型的径向基网络修正法　出处：《中国科学:物理学力学天文学》2013年05期 　论文类型：期刊论文

【摘要】：天文学的发展要求望远镜的口径越来越大,对其建立有限元模型进行动力学分析已成为必要步骤.为了保证理论分析结果符合实际,必须对有限元模型进行修正.针对大望远镜模型修正的难点,提出一种基于径向基神经网络的含有隐藏变量的改进迭代法.首先建立了LAMOST空间桁架和平衡系统的有限元模型.其次从实际出发,讨论了神经网络输入输出的选择方法.用有限的数据信息修正误差大的参数,而小误差参数作为神经网络的隐藏变量,可看作噪声而不予修正.采用均匀设计法获得神经网络的训练样本集,减小了计算量.然后在网络迭代求解过程中,每次在输出参数的附近增加均匀分布的若干个样本,而不是输出参数的单独一个样本,可防止网络性能突然变差,改善泛化特性.最后,悬臂梁和望远镜桁架仿真结果表明,含有隐藏变量的神经网络仍然能够收敛.用实测数据修正了LAMOST的平衡系统,表明了该方法的可行性和有效性.
[Abstract]:With the development of astronomy, the aperture of telescopes is becoming larger and larger, so it is necessary to establish the finite element model of the telescope, so as to ensure that the theoretical analysis results are in line with the practice. The finite element model must be modified. An improved iterative method with hidden variables based on radial basis function neural network is proposed. Firstly, the finite element model of LAMOST spatial truss and balance system is established. The selection method of input and output of neural network is discussed. The parameter with large error is corrected with limited data information and the parameter of small error is used as the hidden variable of neural network. The uniform design method is used to obtain the training sample set of the neural network, which reduces the computational complexity. Then, in the process of iterative solution of the network, the proposed method can be regarded as noise without modification. Adding several samples with uniform distribution in the vicinity of output parameters each time, instead of a single sample of output parameters, can prevent the network performance from suddenly becoming worse, and improve the generalization characteristics. The simulation results of cantilever beam and telescope truss show that the neural network with hidden variables can still converge. The balance system of LAMOST is modified with measured data, which shows the feasibility and effectiveness of the method.
【作者单位】：中国科学院国家天文台南京天文光学技术研究所;中国科学院天文光学技术重点实验室;中国科学院大学;
【分类号】：P111
【正文快照】： 1引言为了探测出更暗的遥远天体,大型天文望远镜除了要有更高的空间分辨率,还要有更强大的聚光本领.增大望远镜的口径面积是增大望远镜聚光本领的根本途径.随着望远镜口径的增强,对其建立有限元模型进行动力学分析已成为必要步骤[1-3].我们不仅要求结构完成设计指标,比如对静

【相似文献】