Convergence criteria for waveform iteration methods applied to partitioned DAE systems in chemical process simulation

In this paper we consider the application of block waveform iteration methods to initial value problems for implicit DAE systems of index 1 arising in chemical process simulation. Block waveform iteration methods permit the concurrent treatment of blocks of subsystems of the entire system with multirate integration techniques gaining a coarse granular parallelism. Their convergence properties strongly depend on the assignment of variables to equations and the partitioning of the system into subsystem blocks. First we proof convergence for waveform iteration methods applied to semiexplicit DAE systems of index 1. The convergence conditions are given in a blocksystem oriented manner, i.e. only the blocksystems have to satisfy some corresponding conditions. Then we show that the convergence conditions are fulfilled for a simplified modeling of distillation columns. An assignment and partitioning algorithm is given, which takes the requirements of the convergence theorems into account. Based on this investigations a prototypes of a waveform-iteration code has been implemented and tested by means of examples included in the user package of the chemical process simulator SPEEDUP [PS1]. (orig.)SIGLEAvailable from TIB Hannover: RR 5549(262)+a / FIZ - Fachinformationszzentrum Karlsruhe / TIB - Technische InformationsbibliothekDEGerman