基于GABP神经网络曲线拟合的快沿电磁脉冲信号源模型求解

Modeling of FREMP Source Based on GABP Neural Network

摘要:

为给电子设备的电磁脉冲效应仿真提供准确的快沿电磁脉冲(fast rise-time electromagnetic pulse,FREMP)信号源模型,提出一种基于遗传算法优化BP神经网络(GABP-NN)曲线拟合的信号源模型求解方法;该方法通过示波器对脉冲信号进行采集,利用GABP神经网络对波形曲线进行高精度拟合,提取网络参数建立信号源模型;为进一步获得BP神经网络拟合规律设置对比实验,采用隐含层神经元数为10的GABP神经网络对FREMP信号源进行建模,所得模型拟合度为91.64%;仿真结果表明该方法运算速度快、精度高。

Abstract:

In order to provide an accurate FREMP (fast rise-time electromagnetic pulse) source model for electronic devices EMP effect simulations, a curve fitting and modeling method based on GABP-NN (BP neural network optimized by genetic algorithms) is proposed. In this method, the FREMP signal collected by oscilloscope is used for high precision waveform curve fitting by GABP-NN and the network parameters is extracted to establish the source modeling. Comparative experiments are set for finding the fitting rule of BP neural network. Using GABP-NN for FREMP source modeling with 10 hidden layer neurons, the fit was 91.64%. The simulation results show that this method does modeling well and has a high computation speed.