一种电子设备电磁敏感性的仿真建模与优化设计方法  

Modeling and Optimization Method of Electromagnetic Susceptibility for Electronic Devices

作　　者：张爱民[1] 张杭[2] 谢美娟[1] 李宏[1] 陈德桂[2] 

机构地区：[1]西安交通大学电子与信息工程学院,西安710049 [2]西安交通大学电气工程学院,西安710049

出　　处：《西安交通大学学报》2008年第6期688-692,共5页Journal of Xi'an Jiaotong University

摘　　要：提出了一种电子设备内部电路的建模方法和基于该模型的优化设计方法.建模方法采用微粒群优化算法与误差反向传播算法相结合的网络训练方法,利用高频仿真软件构造的样本集,对3层神经网络进行训练.所建立的神经网络模型解决了电子设备在电磁敏感性仿真分析和优化设计中,仿真模型存在的计算量和存储量大的问题.优化设计方法采用微粒群算法的多目标优化方法,对内部电路参数进行优化设计,改变了以往寻优过程中穷尽式搜索的方式.仿真结果表明:所建立的神经网络仿真模型的仿真时间比高频仿真模型缩短了1000倍以上,优化设计后的电子设备的抗干扰能力也提高了6倍以上.A model for inner-circuit in electronic devices and an optimization method based on the model have been put forward. The training of neural networks with the combination of a particle swarm optimization （PSO） and the error back propagation （BP） was adopted in the model. Three layer networks were trained on the training data set up by a high-frequency electromagnetic simulation software. The problems of high calculation complexity and huge storage in existing simulation models on electromagnetic susceptibility and optimization for electronic devices are solved by the proposed simulation model. The parameters of inner circuits in electronic devices are optimized by PSO-based multi-objective optimization approach, and conventionally exhaustive searching way is changed at the same time. Simulation experiments show that time cost for simulation with the model is shortened more than 1 000 times compared with the existing high-frequency electromagnetic simulation models; and the capacity for immunity from interference of electronic devices is enhanced more than six times after optimization.

关 键 词：电子设备 神经网络 优化设计 电磁敏感性 微粒群优化算法 

分 类 号：TP29[自动化与计算机技术—检测技术与自动化装置] TN70[自动化与计算机技术—控制科学与工程]