基于RBF神经网络的静电电位动态测试仪波形重建

Dynamic Tester of Electrostatic Potential Waveform Reconstruction Based on RBF Neural Network

为降低静电电位动态测试仪采集波形的失真,提出利用基于RBF(Radial Basic Function)神经网络的系统辨识方法进行波形重建;使用静电高压动态电位标准装置进行试验,将输入的标准方波脉冲信号和经过静电电位动态测试仪后的畸变信号作为黑箱系统的输出和输入信号;利用基于RBF网络的系统辨识方法进行建模,根据所得网络模型预测不同高压脉冲下的重建波形,并与实测波形对比;结果表明基于RBF神经网络的系统辨识方法较好地还原了输入的标准方波脉冲信号,为静电高压动态电位波形校准提供了新的方法。

In order to reduce waveform distortion of the dynamic tester of electrostatic potential, the method of system identification based on RBF （Radial Basic Function） neural network is used for waveform reconstruction. The standard equipment for high voltage generated electrostatic is used to collect data. The standard input pulse signal and the distortional output signal is viewed as output and input of black--box models. The RBF neural network is created to predict the reconstructed waveform under different high--voltage pulse. Then, reconstructed waveform and measured waveform is compared. It is proved that the method of system identification based on RBF neural net- work can restore the standard input pulse signal waveform. And a new method has been put forward for the waveform calibration of dynamic electrostatic potential.