In order to overcome the shortcomings of large time consuming in the near matrix filling and the slow convergence in iteration of adaptive integral method (AIM), the triangle filling strategy and the double-threshold criterion are combined to accelerate the near matrix filling speed in this paper. This method separates the near matrix filling procedure and the near correction during calculation. With the help of incomplete LU decomposition, the preconditioning matrix is constructed from the sparse near matrix before near correction, which could alleviate the ill-conditioned properties of the impedance matrix. Numerical simulation results show that, the triangle filling strategy with ILU preconditioning could improve the filling speed by 2 times and the iteration converges at most 20 times faster than traditional AIM without any accuracy reduction.
Matrix Reordering Using Multilevel Graph Coarsening for ILU Preconditioning