论文部分内容阅读
针对静电放电(ESD)电磁脉冲测试及其辐射场计算这一难点问题,提出了一种基于BP(back propaga-tion)神经网络的任意曲线拟合方法,并给出了任意曲线的数学解析表达式通式。根据IEC 61000-4-2标准的相关规定,采用本方法得出了理想接触式ESD电流波形的解析表达式,神经网络隐含层神经元个数选取为9时,拟合曲线均方误差(MSE)为4.491 7×10-4。提取本方法得到的理想接触式ESD电流曲线的波形参数(上升时间为0.81ns,峰值电流为14.938 5A,30ns时电流值为8.038 9A,60ns时电流值为4.037 7A),证明了该解析式满足IEC61000-4-2的相关规定。通过对6kV放电电压的工程实测ESD电流波形曲线拟合,得到工程实测ESD电流波形的解析表达式,隐含层神经元个数选取为20时,拟合曲线均方误差为0.093 5。本方法为工程实测ESD辐射场的相关计算提供了更准确的解析表达式。
Aiming at the difficult problem of electromagnetic pulse testing and radiation field calculation of electrostatic discharge (ESD), an arbitrary curve fitting method based on BP (back propaga-tion) neural network is proposed and the mathematical expression of arbitrary curve is given Formula. According to the relevant provisions of IEC 61000-4-2 standard, the analytical expression of the ideal contact ESD current waveform is obtained. When the number of neurons in the hidden layer of neural network is selected as 9, the mean square error of the fitted curve MSE) of 4.491 7 × 10-4. The waveform parameters of the ideal contact ESD current curve (0.81 ns rise time, 14.938 5A peak current, 8.038 9A current at 30ns and 4.037 7A current at 60ns) are extracted from this method. It is proved that the analytical expression satisfies IEC61000-4-2 the relevant provisions. The analytic expression of the measured ESD current waveform is obtained by curve fitting of the measured ESD current waveform of a 6kV discharge voltage. When the number of hidden layer neurons is selected as 20, the mean square error of the fitted curve is 0.093 5. The method provides a more accurate analytic expression for the correlation calculation of the measured ESD radiation field in engineering.