FAST馈源支撑系统的天文轨迹优化与智能容错控制

FAST is the largest single-aperture spherical radio telescope in the world built by China. Due to the short commissioning time, there are still some optimization problems need to be studied in depth. In the long-term joint debugging of the feed support system of FAST, the applicant found that the feed support system has several typical problems such as low effective utilization rate of astronomical trajectory and weak fault tolerance. To this end, this project intends to study the astronomical trajectory optimization, intelligent fault diagnosis and fault-tolerant control of the FAST feed support system. The main research contents are listed as follows. First, a cascade-level vibration suppression trajectory optimization algorithm with the largest effective observation time will be designed by combining the time-space-aligned Bezier curve trajectory optimization method and input shaping method. Second, a fault diagnosis algorithm based on deep autoencoder network and XGBoost is studied to achieve fast and accurate diagnosis of faults. Third, an intelligent fault-tolerant control algorithm based on deep neural network signal back-propagation and fault-tolerant control set is studied to improve system fault tolerance. At last, a scaled down model of the FAST cable-driven parallel manipulator will be built to verify the applicability of the proposed methods. The outcome of this proposal will not only increase the effective utilization rate of the astronomical trajectory of FAST and enhance the fault tolerance, but also provide important theoretical and technical support for the safe and high-efficiency operation of the FAST feed support system.

FAST是我国建造的世界最大单口径球面射电望远镜，由于投入试运行时间短，仍存在部分优化问题，亟待有针对性的深入研究。申请人在长期的FAST馈源支撑系统联调中发现，馈源支撑系统存在着天文轨迹有效利用率低和故障容忍度不强等问题。为此，本项目拟针对FAST馈源支撑系统的天文轨迹优化、智能故障诊断及容错控制问题展开研究。主要研究内容包括：综合时空对齐Bezier曲线轨迹优化方法和输入整形方法，研究具有最大有效观测时长的串级抑振轨迹优化算法；研究基于深度自编码网络和XGBoost的故障诊断算法，实现故障的快速、精准诊断；研究基于深度神经网络信号逆推和容错控制集的智能容错控制算法，提高系统故障容忍度；最后，搭建FAST索牵引并联机构缩比模型，验证所提算法的适用性。本项目的研究成果将提升FAST望远镜天文轨迹有效利用率、增强系统故障容忍度，为FAST馈源支撑系统安全、高效率运行提供理论基础及技术支撑。

FAST是我国建造的世界最大单口径球面射电望远镜，由于投入运行时间不长，仍存在部分优化问题亟待科研人员进行针对性的深入研究。本项目以FAST馈源支撑六索牵引并联机构为原型，深入分析了其缩比模型的机械尺寸，同时构建了具备绳索导向功能的导向滑轮机构，研制开发了一种具有完全自主知识产权的索牵引并联机构实验装置；针对FAST馈源支撑系统天文轨迹的优化问题，面向天文跟踪模式，提出了基于多项式的轨迹规划方法，平滑处理了FAST从静止到正常观测阶段的运动轨迹，降低了因急速加减速运动引起的冲击；面向运动中扫描模式，提出了基于时空贝塞尔曲线的轨迹规划方法，缩短了运行中扫描模式下频繁的加减速时间，提高了扫描轨迹的有效利用率；综合最优双S形轨迹规划法和输入整形法的优点，提出了串联抑振轨迹规划方法，有效抑制匀速运动时段的振动；针对故障诊断问题，提出一种基于XGBoost算法的索牵引并联机构故障诊断方法，解决了数据特征选择困难和单个分类器方法精度低的问题；针对FAST馈源支撑系统的容错问题，提出了用于FAST馈源支撑系统AB轴机构故障的容错位姿分配方法，克服了当前AB轴旋转机构故障后FAST无法运转的缺陷，增强了系统对AB轴旋转机构故障的容错能力。本项目的研究成果不仅可以提升FAST望远镜天文轨迹的有效利用率，而且可增强系统对故障的容忍度，为FAST馈源支撑系统安全、高效率运行提供理论基础及技术支撑。本项目共发表学术论文7篇，其中SCI期刊论文6篇；授权发明专利3项；登记软件著作权2项。