地平式太阳望远镜库德焦面的指向跟踪误差建模

YNAO OpenIR > 天文技术实验室

	地平式太阳望远镜库德焦面的指向跟踪误差建模
其他题名	Pointing and Tracking Error Modeling of Coude Focal Plane of Altazimuth Solar Telescope
	刘荣辉
学位类型	工程硕士
导师	柳光乾 ; 曾远辉
	2022-07-01
学位授予单位	中国科学院大学
学位授予地点	北京
培养单位	中国科学院云南天文台
学位专业	光学工程
关键词	指向跟踪精度库德焦面地平式太阳望远镜指向模型机器学习
摘要	随着太阳观测进入高时空分辨率和全方位观测的时代，地基太阳望远镜不断向大口径方向发展，对控制系统要求也越来越高，指向精度和跟踪精度作为一台望远镜控制系统的核心性能指标，直接影响望远镜的整体性能、数据质量和运行效率。由于地平式太阳望远镜特殊的光机机构，光机系统制造和装调误差的不可避免，导致望远镜指向跟踪和图像消旋过程中会在库德焦面产生复杂的二次跟踪误差。为了获得一种通用的指向模型来提高库德焦面的指向跟踪精度，论文分析了二次跟踪误差的产生机制，仿真了二次跟踪误差变化特点，提出了基于支持向量回归的机器学习方法对库德焦面综合误差的建模方案，并在一米新真空太阳望远镜NVST上开展了实测建模工作，论证了该方法的可行性。论文第一章论述了本论文研究工作的背景及意义，调研了国内外提高望远镜指向跟踪精度常用的控制方法。第二章首先通过对比地平式太阳望远镜和夜天文望远镜光机结构的差异，阐述了地平式太阳望远镜库德焦面的指向跟踪误差产生机制，以及与夜天文望远镜卡焦指向跟踪误差的不同。以NVST太阳望远镜为例，仿真了高度轴、方位轴、主光轴和消旋轴轴系不同心情况下，在库德焦面引入的二次指向跟踪误差。根据仿真结果，说明库德焦面引入的二次指向跟踪误差是有规律可循的，但是由于难以在库德焦面实测这些轴系旋转中心，要数理推动建立指向模型是非常困难的。提出了机器学习的方法来建立库德焦面的指向跟踪误差模型。第三章分析了望远镜的指向跟踪误差建模是一个双输入双输出的过程，这是一个典型的回归问题。然后介绍了多项式回归、支持向量回归、神经网络三类机器学习方法的相关算法原理、参数设置，分析了机器学习在库德焦面建立指向跟踪误差模型的可行性，以实测的库德焦面指向误差数据为基础，分析基于三种方法的指向跟踪建模过程，并对比了它们的建模效果，得到支持向量回归建模为最优。第四章应用支持向量回归方法进行建模，在NVST等不同太阳望远镜的库德焦面以及其他焦面位置上进行了实测。其结果表明支持向量回归模型可以有效提高地平式太阳望远镜库德焦面的指向跟踪精度，指向误差的RMS值和30分钟的跟踪误差都能达到几个角秒量级，并且具有较好的通用性。第五章总结了本论文研究工作的不足之处和对下一步工作的展望。由于建模数据采集时间与模型验证数据的观测之间相隔时间都较短，在几天之内完成一次建模和测试工作，因此，没有验证模型在长时间内的稳定性，需要进行更多的测试。另外，从工程应用的角度讲，建模数据越少越好，但从数学的角度讲，数据越多越好。因此，采集多少颗星能达到模型性能和工作量的最优，也还有待深入分析和实验。
其他摘要	As the solar observation enters the era of high temporal and spatial resolution and all-round observation, the ground-based solar telescope continues to develop in the direction of large aperture, and the requirements for the control system are becoming higher and higher. The pointing accuracy and tracking accuracy, as the core performance indicators of a telescope control system, directly affect the overall performance, data quality and operation efficiency of the telescope. Due to the special optical and mechanical mechanism of the altazimuth solar telescope, the manufacturing and installation errors of the optical and mechanical system are inevitable, resulting in complex secondary tracking errors in the coude focal plane in the process of telescope pointing tracking and image racemization. In order to obtain a general pointing model to improve the pointing and tracking accuracy of coude focal plane, this paper analyzes the generation mechanism of secondary tracking error, simulates the variation characteristics of secondary tracking error, puts forward the modeling scheme of coude focal plane comprehensive error based on machine learning method of support vector regression, and carries out the measured modeling work on 1m new vacuum solar telescope NVST to demonstrate the feasibility of this method.The first chapter discusses the background and significance of the research work of this paper, and investigates the common control methods to improve the pointing and tracking accuracy of telescopes at home and abroad.In Chapter 2, firstly, by comparing the differences of optical and mechanical structures between the altazimuth solar telescope and the night telescope, the mechanism of pointing and tracking error of the coude focal plane of the altazimuth solar telescope and the difference between the altazimuth solar telescope and the night telescope are described. Taking the NVST solar telescope as an example, the secondary pointing tracking error introduced in the coude focal plane is simulated when the elevation axis, azimuth axis, main optical axis and racemic axis are not concentric. According to the simulation results, it shows that the secondary pointing tracking error introduced by the coude focal plane is regular, but because it is difficult to measure these shafting rotation centers in the coude focal plane, it is very difficult to mathematically promote the establishment of pointing model. A machine learning method is proposed to establish the pointing and tracking error model of coude focal plane.The third chapter analyzes that the pointing and tracking error modeling of telescope is a dual input and dual output process, which is a typical regression problem. Then it introduces the relevant algorithm principles and parameter settings of three kinds of machine learning methods: polynomial regression, support vector regression and neural network, analyzes the feasibility of establishing the pointing tracking error model in the coude focal plane by machine learning, analyzes the pointing tracking modeling process based on the three methods based on the measured coude focal plane pointing error data, compares their modeling effects, and obtains that the support vector regression modeling is the best.In Chapter 4, the support vector regression method is used to model and measure the coude focal plane and other focal plane positions of different solar telescopes such as NVST. The results show that the support vector regression model can effectively improve the pointing and tracking accuracy of the coude focal plane of the altazimuth solar telescope. The RMS value of the pointing error and the tracking error of 30 minutes can reach the order of several angular seconds, and has good universality.The fifth chapter summarizes the shortcomings of the research work of this paper and the prospect of the next work. Due to the short interval between the collection time of modeling data and the observation of model verification data, the modeling and testing work is completed within a few days. Therefore, there is no verification of the stability of the model in a long time, and more tests are needed. In addition, from the perspective of engineering application, the less modeling data, the better, but from the perspective of mathematics, the more data, the better. Therefore, how many stars can be collected to achieve the optimal performance and workload of the model still needs in-depth analysis and experiment.
学科领域	天文学 ; 太阳与太阳系 ; 太阳与太阳系其他学科 ; 机械工程 ; 仪器仪表技术 ; 电子、通信与自动控制技术
学科门类	工学 ; 工学::光学工程 ; 工学::控制科学与工程
页数	0
语种	中文
文献类型	学位论文
条目标识符	http://ir.ynao.ac.cn/handle/114a53/25799
专题	天文技术实验室
推荐引用方式 GB/T 7714	刘荣辉. 地平式太阳望远镜库德焦面的指向跟踪误差建模[D]. 北京. 中国科学院大学,2022.

条目包含的文件
文件名称/大小	文献类型	版本类型	开放类型	使用许可
地平式太阳望远镜库德焦面的指向跟踪误差建（9577KB）	学位论文		开放获取	CC BY-NC-SA	浏览请求全文