基于支持向量机的卫星姿控系统的故障诊断研究

作者：李翔宇，李帅（沈阳理工大学，辽宁 沈阳 110158）

 

摘　要：【目的】解决卫星姿态控制系统中反作用飞轮执行机构中的故障定位问题。【方法】本研究提出了一种基于支持向量机的卫星反作用飞轮故障诊断方法。首先，利用半物理仿真平台模拟出卫星真实在轨运行数据，数据分为反作用飞轮卡死故障、空转故障、力矩衰减累积故障、正常数据四类，每类数据提取约70 h（周期为0.25 s）；其次，对卫星反飞轮注入故障，选取反作用飞轮的转速数据进行时域分析，选取其时域特征作为故障特征，并将正常数据集与反作用飞轮衰减数据集在时域中划分成13种特征，得到每类50份数据；最后，利用支持向量机算法完成对卫星反作用飞轮中出现的卡死、空转、衰减故障的识别。【结果】正常数据、反作用飞轮卡死故障与空转故障的分类结果，正常数据与反作用飞轮衰减故障的分类结果，识别准确率均为100%。【结论】本研究可以为卫星的故障定位问题提供一种更好的解决方案。

关键词：支持向量机；卫星姿态控制系统；故障诊断

 

Abstract: [Objective] Solving the problem of fault location of reaction wheel actuator in satellite attitude control system. [Method] This study proposes a satellite reaction flywheel fault diagnosis method based on support vector machine. Firstly, the semi physical simulation platform is used to simulate the real in orbit operation data of the satellite. The data is divided into four categories: reaction flywheel jamming fault, idle fault, torque attenuation cumulative fault, and normal data. Each type of data is extracted for about 70 hours (with a period of 0.25 seconds). Secondly, for the injection fault of the satellite anti flywheel, the speed data of the reaction flywheel is selected for time-domain analysis, and its time-domain characteristics are selected as the fault features. The normal dataset and the reaction flywheel attenuation dataset are divided into 13 types of features in the time-domain, resulting in 50 data for each type. Finally, the support vector machine algorithm is used to identify stuck, idle, and decaying faults in the satellite reaction flywheel. [Result] The classification results of normal data, reaction flywheel stuck fault, and idle fault, as well as the classification results of normal data and reaction flywheel attenuation fault, have a recognition accuracy of 100%. [Conclusion] This study can provide a better solution for the problem of satellite fault localization.

Keywords: support vector machine; satellite posture control system; fault diagnosis