电大尺寸矩形谐振腔耦合场的神经网络预测

以电大尺寸的矩形谐振腔局部点的前后门耦合场的计算,通过神经网络方法实现其他点耦合场的预测,判定矩形谐振腔的电磁敏感点.由于电大尺寸的矩形腔很难通过全波分析或小波分析获得特定条件下的耦合场,而神经网络方法不需考虑内部模型的复杂性便可实现非线性预测,因此将人工神经网络方法应用于电磁预测中可实现矩形腔耦合场的计算.通过电大尺寸矩形腔前后门耦合场实验方案,提取了目标参数,创建了BP神经网络的预测模型.即在平面波照射下,以入射波的功率,极化方向,预测点的位置坐标作为BP网络的输入参数,相应点的功率(电压)作为输出参数,经过适当的训练,建立耦合场的预测模型,并以此模型预测了腔体内探测点的耦合场.预测结果与实测结果相比较显示了该方法的有效性和准确性,为电大金属腔耦合场的计算提供了一种有效的方法.

Prediction Based on a Neural Network for Coupling Fields of Electrically Large Rectangular Metal Cavity

Electromagnetic sensitive points of electrical-large rectangular metal cavity have been calculated in the case of front and back door coupling fields.Neural network method is used to predict coupling fields by measurement some local fields of the cavity.It is difficult for electrical-large rectangle metal cavity to obtain coupling fields by full wave and wavelet analysis.The neural network method need not consider complexities of internal structure and can realize nonlinear prediction.Therefore,it can be applied in predicting the front and back door coupling fields of the cavity.We extract target parameters and create the prediction model of BP neural network by experiment design.That is,under irradiation of the TEM wave,input parameters of BP network are incident wave power,polarized direction and position coordinates of predicted points.Output parameter is the voltage of the corresponding points.The prediction model of coupling fields are established after proper training and the coupling fields of the detected points are predicted by the model.Research results show that the method is effective and accurate.That is,an effective method is provided for calculation coupling fields of electrical-large metal cavity.