A note on convergence conditions of waveform relaxation algorithms for nonlinear differential-algebraic equations (Q5928466): Difference between revisions

The paper is devoted to a weakening of the known assumptions for convergence of the waveform relaxation (WR) algorithm for resolving of nonlinear differential-algebraic equations, unresolved with respect to derivatives. This algorithm has proposed by \textit{Lerarasmee} et al. [IEEE Trans. CAD of IC and Systems 1, No. 3, 131-145 (1982)]. It is some generalization of the Picard algorithm that is based on the transition from the initial functions of the system to be solved to the equivalent splitting ones via their variables set expansion. The authors develop the spectral analysis technique applying to some \(2 \times 2\) matrix the elements of which consist of Lipschitz coefficients of the splitting functions. This technique was developed by \textit{Z. Jackiewicz} and \textit{M. Kwapisz} [SIAM J. Numer. Anal. 33, No. 6, 2303-2317 (1996; Zbl 0889.34064)]. The obtained convergence conditions for the Picard and WR algorithms are more fine then previously published ones. It is demonstrated via a comparative analysis and on an example of a fifth-order system presenting an electric circuit. Numerical results also presented.