摘要
针对目前T/R组件外部测试点少,测量项目无法与内部模块一一对应,难以通过外部测量数据直接诊断到SRU级故障的问题,提出一种基于多分类SVM的SRU级故障诊断方法。首先介绍SVM的基本原理和多分类方法,结合T/R组件内部结构,介绍T/R组件SRU级模型,然后从理论层面分析接收通道当单个SRU发生故障时各个测试项目测量值的超标概率与程度,并以此为依据建立故障数据库,最后利用数据库中的数据对多分类SVM进行训练并预测,诊断准确率可达90%以上,表明该方法可准确有效地诊断出T/R组件SRU级故障。
In allusion to the problem that T/R modules have few external testing points,and the measuring items can not correspond to the internal modules one by one,as well as it is difficult to diagnose SRU-level faults directly according to the external measured data,a kind of SRU level fault diagnosis method based on multi-classification SVM is proposed. The basic principle and multi-classification method of SVM is introduced,and also the SRU level model of T/R module is introduced in combination with the internal structure of T/R modules. The probability and degree of the measured value of each test item when a single SRU fails in the receiving channel are analyzed theoretically. Based on this,a fault database is established. The multiclass SVM is trained and predicted by using the data in the database. The diagnostic accuracy of the model can reach above90%,which indicates that this method can diagnose SRU level faults of T/R modules accurately and effectively.
作者
王挺
盛文
蒋伟
WANG Ting;SHENG Wen;JIANG Wei(Department of Air Defense Early Warning Equipment,Air Force Early Warning Academy,Wuhan 430019,China)
出处
《现代电子技术》
北大核心
2019年第23期67-71,共5页
Modern Electronics Technique
基金
军内科研重点项目(KJ2012225)~~
