Output prediction about crosstalk coupling of representative multi-cable bundle on aircraft platform based on neural network

A method of output prediction about crosstalk coupling of multi-cable bundle on aircraft is presented without solving the coupling matrix. BP neural network is used to simulate the coupling relation between input and output. The network of input and output are trained with LMBP algorithm. MOM is applied to calculate the output value of multi-cable bundle when inputs in different frequency point are given through establishing 3D model of transmission lines. It´s shown that the predicted values of neural network in different frequency points coincide with values of MOM simulative outputs well. Otherwise a concept of relative coupling error is put forward to estimating the accuracy of prediction algorithm. The predicted values are much closer to the simulative ones after the inputs of neural network have been optimized properly. This method can be used to predict the output of nonlinear multi-cables coupling in complex system structure.