基于进化卷积神经网络的外壳屏蔽效能预测  

Prediction of Shielding Effectiveness of Mechanical Shell Based on Evolutionary Convolutional Neural Network Algorithm

作　　者：肖保明 丁爱飞 陈泽南 丁佳飞 张武 颜伟[3] Xiao Baoming;Ding Aifei;Chen Zenan;Ding Jiafei;Zhang Wu;Yan Wei

机构地区：[1]国网电力科学研究院有限公司 [2]国电南瑞吉电新能源有限公司 [3]南京师范大学电气与自动化工程学院

出　　处：《安全与电磁兼容》2023年第1期41-45,共5页Safety & EMC

基　　金：国家自然科学基金(52107005)资助项目。

摘　　要：随着设备功率密度和可靠性要求的不断提升,设备电气外壳应设计孔缝以解决散热和电磁屏蔽问题。为了优化孔缝结构和位置,需在设计过程中计算和预测电气外壳屏蔽效能。针对现有屏蔽效能分析方法误差较大等问题,先建立了基于电磁拓扑理论BLT方程的带孔外壳屏蔽效能等效模型,并提出了基于进化卷积神经网络的屏蔽效能参数提取方法,再以外壳屏蔽效能部分频点结果作为模型输入,提取屏蔽效能模型参数,最后根据模型计算其它频点的外壳屏蔽效能。对比分析结果表明,文中方法能够实现带孔缝外壳屏蔽效能预测,误差小于10 dB且不受外壳结构、孔缝位置和屏蔽材料参数影响。With the improvement of equipment power density and reliability requirements,the electrical shell of equipment shall be designed with holes to solve the problems of heat dissipation and electromagnetic shielding.In order to optimize the slot structure and location,it is necessary to calculate and predict the shielding effectiveness of the electrical enclosure in the design process.Aiming at the problems of large error of existing shielding effectiveness analysis methods,an equivalent model of shielding efficiency of perforated shells based on BLT equation of electromagnetic topology theory is first established,and a shielding efficiency parameter extraction method based on evolutionary convolutional neural network is proposed,and then the shielding efficiency model parameters are extracted by taking the frequency points of shell shielding effectiveness as the model input,and finally the shell shielding efficiency of other frequency points is calculated according to the model.The comparative analysis results show that the method in this paper can realize the prediction of the shielding effectiveness of the shell with holes and slots,and the error is less than 10 dB,which is not affected by the shell structure,the location of the holes and the shielding material parameters.

关 键 词：电磁干扰 电磁屏蔽 BLT算法 ECNN算法 

分 类 号：TN03[电子电信—物理电子学] TP183[自动化与计算机技术—控制理论与控制工程]