Fast iterative algorithms based on optimized explicit time stepping (Q1086986)

The generalized nonlinear minimal residual (GNLMR) method is shown to consistently accelerate and stabilize iterative algorithms for solving nonlinear problems by using the optimized explicit multistepping. The examples presented in this paper illustrate the beneficial effects of the optimized multistep algorithm on the computational efficiency and the convergence rate as applied to several nonlinear problems in fluid dynamics. The significant reduction in computing time when using the multiple optimized acceleration factors is negligibly weighed down by the computation costs due to the requirements for additional computer storage.