飞轮系统的对应性分析故障检测方法  

Correspondence Analysis of Flywheel System for Fault Detection

作　　者：龚学兵 姚保江 于雷 杨灿 GONG Xue-bing;YAO Bao-jiang;YU Lei;YANG Can(Xi an Institute of Aerospace Power Technology,Xi an 710025,China)

机构地区：[1]西安航天动力技术研究所,西安710025

出　　处：《科学技术与工程》2018年第22期157-164,共8页Science Technology and Engineering

基　　金：国家"973"计划(2012CB720003)资助

摘　　要：为了实现复杂系统故障的简单、快速检测问题,提出一种基于对应性分析的检测方法,研究了闭环飞轮的故障检测问题。首先,利用飞轮系统的健康数据与故障数据描述飞轮系统的状态,数据向量由飞轮的物理状态参数构成。其次,利用对应性分析方法提取健康数据的故障检测指标及其相应的阈值。最终,利用对应性分析方法提取故障数据的故障检测指标值;并与故障检测指标阈值对比,检测飞轮系统的故障。与扩展卡尔曼滤波故障检测方法相比,不需要复杂的物理建模,利用数据向量之间的关联性即可实现飞轮系统的故障检测。飞轮仿真模型的故障检测结果的对比分析表明:本文算法利用数据关联性避免复杂物理建模;算法能利用实时数据实现在线监测,通用性强。可见,对应性分析故障检测方法具有简单、快速的优点。In order to quickly and easily detect the faults of complex systems,the correspondence analysis is proposed and used to investigate the fault detection problems of flywheel system.First,the normal and abnormal states of flywheel are described by the data collected from the physical variables.Second,the fault detection indices of correspondence analysis are extracted from the normal data,and the thresholds of fault detection indices are defined by the correspondence analysis.Finally,the value of fault detection indices is extracted from the faulty data and compared with the thresholds,which are used to detect the faults of flywheel system.Compared with the extended Kalman filter fault detection method,the proposed algorithm does not need the complicated physical modeling and is easy to detect the faults of flywheel by the correlation of physical variables.The comparative analysis of the fault detection results given by the flywheel simulation model shows that the algorithm can avoid the complex physical modeling by data correlation,what s more,it can achieve online monitoring by the real-time data,and it is of universality.It is concluded that the proposed method has the merit of fast and simple.

关 键 词：飞轮系统 对应性分析 扩展卡尔曼滤波 故障检测 

分 类 号：V474[航空宇航科学与技术—飞行器设计]