No abstract

No abstract

The paper describes an equation solver superstructure developed for a sequential modular dynamic process simulation system as part of a Eureka project with Norwegian and British participation. The purpose of the development was combining some of the advantages of equation based and purely sequential systems, enabling implicit treatment of key variables independent of module boundaries, and use of numerical integration techniques suitable for each individual type of variable. For training simulator applications the main advantages are gains in speed due to increased stability limits on time steps and improved consistency of simulation results. The system is split into an off-line analysis phase and an on-line equation solver. The off-line processing consists of automatic determination of the topological structure of the system connectivity from standard process description files and derivation of an optimized sparse matrix solution procedure for the resulting set of equations. The on-line routine collects equation coefficients from involved modules, solves the combined sets of structured equations, and stores the results appropriately. This method minimizes the processing cost during the actual simulation. The solver has been applied in the Veslefrikk training simulator project