基于天气雷达标准输出控制器的故障诊断方法研究  

Research on Fault Diagnosis Method based on Weather Radar Standard Output Controller

作　　者：徐颂捷 何建新[1] 黎志波[2] 张福贵[1] XU Songjie;HE Jianxin;LI Zhibo;ZHANG Fugui(Electronic Engineering College,Chengdu University of Information Technology,Chengdu 610225,China;Technical Center for AtmosphericSounding of Jiangxi Province,Nanchang 330096,China)

机构地区：[1]成都信息工程大学电子工程学院,四川成都610225 [2]江西省大气探测技术中心,江西南昌330096

出　　处：《成都信息工程大学学报》2019年第3期257-262,共6页Journal of Chengdu University of Information Technology

基　　金：国家重点研发计划资助项目(2018YFC1506100)

摘　　要：根据天气雷达标准输出控制器采集的雷达关键指标数据,提出了一种具有自学习且半监督作用的异常检测与支持向量机(SVM)的联合故障诊断方法,实现天气雷达运行状态评估与故障检测定位。针对采集数据,首先使用异常检测算法建立概率模型,计算样本落入正常范围的概率,实现非正常样本的识别;其次,以所得样本的概率值作为支持向量机模型的新增特征,建立SVM分类器,对故障进行诊断。经实验表明与传统的逻辑回归、神经网络的分类方法相比,在小样本且各类别训练数据正负偏离过大的情况下,此方法能够更准确、高效地诊断雷达故障。Based on the Weather Radar data Standard Output Controller(WRSOC), acombination fault detection algorithm is proposed in this paper,which is anomaly detection and support vector machine(SVM) with self-learning and semi-supervisory function, andit realizes the operation state evaluation and fault detection of weather radar. Firstly,building probability model with anomaly detection algorithm, calculate the probability of samples falling into the normal range, and realize the recognition of abnormal samples. Secondly, using the probability value of samples as the new feature of support vector machine model, the SVM classifier model is established to diagnose faults. Experiments show that this method can diagnose radar faults more accurately and efficiently than traditional logistic regression and neural network in the case of small samples and large deviation of training data.

关 键 词：雷达系统 故障诊断 异常检测 支持向量机 

分 类 号：TN956[电子电信—信号与信息处理]