Flexibility and Efficiency in Modular Dynamic Chemical Plant Simulation. (Volumes I and II) (Stiff, Simultaneous Numerical Integration).

Abstract

The research described here focused on flexible and efficient simulation of chemical plants that give rise to large, stiff ordinary differential equation sets. The primary goals of the research were to: (1) Design a single simulator structure capable of employing novel simultaneous and sequential-clustered integration approaches (previously thought incompatible with a general, modular simulator structure), as well as the traditional sequential approach, (2) Evaluate the merits of these various numerical integration approaches, and (3) Identify and employ data structures that are well-suited to this type of simulation. The research was accomplished by constructing and testing a prototype modular simulator that makes extensive use of highly-advantageous plex data structures, and that incorporates several unit model routines including dynamic distillation tray, reboiler, condenser, method-of-lines heat exchanger and controller models. The simulator was tested on controlled distillation and heat exchange systems, and accuracy, execution time and storage requirements were recorded and analyzed. It was concluded that the optimal integration strategy was the sequential-clustered approach (with clustering of units that display strong interdependence), where integration was performed simultaneously within each cluster (containing one or more pieces of equipment) while integration remained sequential from cluster to cluster. Other conclusions were drawn concerning the benefits of sparse matrix techniques, use of plex data structures, use of a preprocessor for run-time dimensioning of arrays, and applicability of the method of lines for treatment of distributed-parameter systems in modular dynamic simulation. Also, physical property evaluation was identified as a costly part of the simulation task, but initial testing of shortcut techniques for improving execution times was inconclusive. An extensive list of recommendations for future work is included.