Acceleration of Iterative Solutions of Linear Algebraic Equation System by Neutralizing Larger Eigenvalues of Transition Matrix

A method is offered for accelerating the convergence of iterative solutions of a system of linear algebraic equations applied in the mathematical simulation of electrical circuit modes. The method is based on neutralizing the transition matrix eigenvalues with a larger modulus. The dependence of a number of the Gauss-Zeidel method iterations on a number of neutralized roots is illustrated by an example of designing the circuits described by matrices with complex components. Calculation results are compared to theoretical convergence estimations obtained using the characteristic equation roots. For the optimal variant in the sequential upper relaxation method a possibility is shown of the further convergence improvement at the expense of the application of the procedure to neutralize roots.