CONVERGENCE ACCELERATION ALGORITHMS FOR SOLVING RECYCLE PROCESS PROBLEMS IN A SEQUENTIAL MODULAR APPROACH

The sequential modular approach, used in essentially all of today's industrial steady state process simulation and design software packages, was adopted in simulating steady state recycle processes. With the current industrial trend heading towards better utilization of material and energy, processes of the future are likely to be represented by more complex recycle configurations. The added complexity is expected to increase the computer time required to simulate these processes. A further increase in simulation time is anticipated when strongly nonideal systems are handled by the process. Despite the apparent limitations for design and optimization calculations, the sequential modular approach is expected to remain as the basis for solving chemical engineering process flowsheeting problems for a number of years. The development of an improved convergence acceleration algorithm, therefore, would help the user to reduce the simulation cost expenditure and appears to be particularly timely.;A new convergence acceleration technique for the successive substitution procedure, called the accelerated first derivative method, was developed along with an optimum computational strategy to promote the convergence of steady state process simulations. Its effectiveness was tested with simulation of three steady state recycle processes and compared to those tested in this work. Computer time reduction up to 20% over the next best acceleration algorithm was demonstrated in the strong variable interacting example solved using the new convergence acceleration approach. Moreover, the proposed new algorithm can be easily incorporated into the existing sequential modular software and is especially suitable for applications where the Jacobian matrix is not available.;Most of the researchers in the process flowsheeting field share a common experience--the mass balance iterations in sequential modular flowsheeting converge monotonically. On the basis of this findings, the eigenvalues are assumed to be real and positive. As part of this project, the impact of a nonideal mixture on the convergence behavior produced by the mass balance iterative calculations was systematically examined. Geometrical interpretations of the findings were presented.;The use of the successive underrelaxation approach to control the convergence behavior of the iterative mass balance calculations was also investigated and reported.