Iterative solutions of mildly nonlinear systems (Q433936)

The iterative solutions of large and sparse mildly nonlinear systems are considered, the systems are of type \[ V(x)+ Tx= b, \] where \(T\) is a Stieltjes matrix, or \(T\) is irreducible. After appropriate splitting of the diagonal nonlinear terms, the nonlinear contributions are linearized in sequence to derive a nested iterative method which is well-defined and converging. The splitting is based on a simple Jordan decomposition of functions with bounded variations. A dual algorithm with analogous properties is derived by reversing the linearization order. The proposed algorithms will always converge when the initial guess is chosen as suggested. Examples of mildly nonlinear systems arise from the numerical modelling of free-surface hydrodynamics. In the present paper, the algorithms are applied to simulate porous flows in a confined aquifer. Two figures illustrate the solution.