A method predicting crosstalk coupling between interconnecting wires with artificial neural networks is proposed.Parameters that influence crosstalk coupling are selected as input prediction factors.Back propagation neural network constructs a mapping relation between input prediction factors and output crosstalk coupling of interconnecting wires.Learning sample sets computed by MTL and FDTD method are used to train BP networks.A prediction BP model is obtained.BP predictions and test sample results are compared.

For boundary conditions with functions changing rapidly,especially in charged conductors,we combine compact feature of wavelet functions and high accuracy of RBF in a coupled method and solve ground metal trough/box problems. Rectangular,ridge and round waveguides are analyzed with RBF and FEM methods.Numerical experiments show that RBF method is efficient in solving electromagnetic boundary and eigenvalue problems.RBF method is easier and more accurate with less nodes.

Based on the distribution characteristic of fields in the neighbourhood of a sharp corner, the B spline finite element method in polar coordinates is exploited for solving the boundary-value problems with singularities at the sharp corner. The proposed method improves the overall accuracy and is convenient to implement. The Motz problem, cracked-beam problem, and eigenvalue problem for the sector waveguides are calculated and analyzed to verify the validity of the method.