首页|一种基于元学习的大口径射电望远镜俯仰轴承故障辨识方法研究

一种基于元学习的大口径射电望远镜俯仰轴承故障辨识方法研究

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
在经过长期运行后大口径射电望远镜俯仰轴会出现微小扭曲,滚动轴承作为承载俯仰轴的核心部件,也会因长期承受交变载荷增加疲劳风险,导致轴承寿命以及望远镜指向精度的下降,极大影响望远镜的性能.以俯仰轴承为研究对象,开展故障辨识方法研究,可为望远镜天线的高性能运行提供重要支撑.为实现在有限数据和复杂工作条件下准确地辨识俯仰轴承故障,提出了一种小样本条件下基于元学习的故障辨识方法(Few-shot Meta-learning Fault Identification,FMFI).首先将不同工况下的原始信号转换为时频图像数据,之后按照元学习协议将数据样本随机采样到不同的学习任务中.在有限样本的条件下,FMFI可以通过训练任务中的样本信息获取通用的先验知识,在未知的测试任务下实现准确快速的故障辨识.选取了与望远镜俯仰轴承工况具有相似性的变负载轴承数据集进行实验,实验结果表明,FMFI方法具有很高的准确性和可靠性,为大口径射电望远镜俯仰轴承的主动运维和高质量服役提供了有力的技术支持.
Research on Fault Identification Method of Elevation Bearing for Large Aperture Radio Telescope Based on Meta Learning
The prolonged operation of the large aperture radio telescope will lead to slight distortion of elevation axis,increasing the risk of fatigue of the rolling bearing which is the core component of the elevation axis.This can lead to a decline in the bearing life and the pointing accuracy of the telescope,which will greatly affect the telescope's high performance service.Investigating the fault identification method for the elevation bearing can provide an important support for the high-performance operation of the telescope antenna.In this paper,a few-shot meta-learning fault identification(FMFI)method based on meta-learning is proposed in order to achieve accurate fault identification of elevation bearing under limited data and complex working conditions.The raw signals of different working conditions are first converted to time-frequency images data and then randomly sampled for different learning tasks according to the meta-learning protocol.Under limited sample data condition,the FMFI method can obtain universal prior knowledge from the sample in the training task to achieve accurate and fast fault identification in unknown testing tasks.The variable load bearing data set which is similar to the working condition of telescope elevation bearing is selected for experiment,and the experimental results show that the FMFI method is high accurate and reliable,providing strong technical support for the operation,maintenance,and high-quality service of large aperture radio telescope.

telescopeinstrumentation:elevation bearingmethods:data analysismethods:pattern recognition

朱金浩、许谦、薛飞、何飞龙、梁娟、许多祥

展开 >

中国科学院新疆天文台 乌鲁木齐 830011

中国科学院大学 北京 100049

中国科学院射电天文重点实验室 乌鲁木齐 830011

新疆射电天体物理重点实验室 乌鲁木齐 830011

展开 >

望远镜 仪器:俯仰轴承 方法:数据分析 方法:模式识别

国家自然科学基金中国科学院青年创新促进会项目国家重点研发计划新疆维吾尔自治区自然科学基金新疆维吾尔自治区自然科学基金中国科学院天文台站设备更新及重大仪器设备运行专项中国科学院科研仪器设备研制项目

12273102Y2020192021YFC22036012021D01B1112022D01B72PTYQ2022YZZD01

2024

10.15940/j.cnki.0001-5245.2024.03.007

天文学报
中国天文学会

天文学报

CSTPCD北大核心
影响因子：0.339
ISSN：0001-5245
年,卷(期)：2024.65(3)
  • 18