基于卷积神经网络的雷达微波组件的故障诊断  

Fault Diagnosis Method for Radar Microwave Modules Based on Convolutional Neural Network

作　　者：袁坤 邓威[1] 戴敏[1] 邵子金 YUAN Kun;DENG Wei;DAI Min;SHAO Zijin(Nanjing Research Institute of Electronics Technology,Nanjing 210039,China)

机构地区：[1]南京电子技术研究所,江苏南京210039

出　　处：《电子工艺技术》2022年第4期219-222,共4页Electronics Process Technology

摘　　要：雷达微波组件作为复杂电子系统的典型代表,其电路密度高、制造工艺路线长、指标要求严苛,实际生产环节工艺问题引入的性能缺陷时有发生,因此微波组件生产过程中故障诊断至关重要。但诊断过程对调试、制造工艺方面工程经验依赖度较强,为降低人工依赖度、提升故障诊断效率及准确率,提出了基于卷积神经网络的雷达微波组件故障诊断方法,根据微波组件实际工艺流程,通过将缺陷故障记录信息分组、标准化、代码化,卷积处理,实现了雷达微波组件及套装模块智能诊断。诊断试验结果表明,雷达微波组件故障诊断准确率达到95%以上,套装件的诊断准确率达到86%以上,大幅度提升了诊断效率。As a typical representative of complex electronic systems,radar microwave modules have high circuit density,long manufacturing process routes,strict index requirements,and performance defects introduced by process problems occur from time to time in actual production.Therefore,fault diagnosis is crucial important in the production process of microwave modules.However,the diagnosis process is highly dependent on engineering experience in debugging and manufacturing process.In order to reduce the manual dependence and improve the efficiency and accuracy of fault diagnosis,a fault diagnosis method for radar microwave module based on convolutional neural network is proposed,which is according to the actual process flow of microwave modules,through grouping,standardization,coding and convolution processing of defect fault record information,the intelligent diagnosis of radar microwave module and package modules is realized.The diagnostic test results show that the fault diagnosis accuracy rate of radar microwave module is over 95%,and the diagnostic accuracy rate of the package modules is over 86%,which greatly improves the diagnostic efficiency.

关 键 词：卷积神经网络 故障诊断 雷达微波组件 故障模式标准化 智能故障诊断 

分 类 号：TN407[电子电信—微电子学与固体电子学]