"PROPS, (Process Optimization System), has been used to simulate and optimize the Universal Oil Products solid phosphoric acid catalyst process for producing gasoline. PROPS, a modified version of CHESS (Chemical Engineering Simulation System), performs a complete design and economic evaluation of the process for each set of independent variables. The computational time during optimization was reduced by the use of a regression equation to determine K-values as a function of temperature and pressure. K-value data used for the regression was generated from the Chao-Seader correlation using mixtures of the gasoline components to be encountered during optimization. The regression equation produces K-values within 5% for the liquid-vapor equilibria over the range of 60-300°F and 50-350 psia.
The nine independent variables used in this study were the feed/ effluent exchanger outlet temperature, the feed heater outlet temperature, the recycle split, the quench splits to the first three reactor beds, the depropanizer pressure, and the depropanizer and debutanizer refluxes. The percent return on investment was used to measure the desirability of the process at each design. Six dependent variables were constrained during the design optimization. Four of these implicit constraints, the reactor bed temperatures (which had to be below 485°F to prevent tar formation in the reactor) increased calculation time seriously because they could be maintained only by iterative calculation throughout the entire plant.
The objective function of the polymerization process seems to be unimodal in the valid region. The gradient of the objective function is large in the direction of higher reactor temperatures or higher conversions. The gradient is relatively small along the reactor temperature constraints.
Three optimization techniques, the Complex Method, the Pattern Search, and the Adaptive Random Search, were each started from two points in an attempt to find the optimum. The Complex Method, in both cases, contracted and failed to progress once the upper temperature constraints in the reactor were reached. Penalty functions were applied to keep the Pattern Search in the valid region. The penalty functions created a ridge along the temperature constraints and the first Pattern run failed to move along this ridge. The search from the second starting point moved a considerable distance along the ridge before progress stopped near the optimum.
The Adaptive Random Search required fewer steps to obtain better return than the other methods used. With 198 Return-on-Investment (ROI) evaluations in 12 minutes (CPU-370/165) the Adaptive Random Search found an ROI of 14.93% compared to the value of 14.24% ROI for the best of the other methods, which had terminated itself after 277 evaluations in 8 minutes (CPU-370/165)"--Abstract, pages ii-iii.
Gaddy, J. L.
Crosser, Orrin K.
Rigler, A. K.
Kieffer, John C.
Chemical and Biochemical Engineering
Ph. D. in Chemical Engineering
University of Missouri--Rolla. Department of Chemical Engineering
National Defense and Education Act Title IV Fellowship
University of Missouri--Rolla
viii, 76, A-v, A-87, B-7 pages
© 1974 Larry Dale Gaines, All rights reserved.
Dissertation - Open Access
Chemical engineering -- Computer simulation
Chemical engineering -- Simulation methods
Chemical engineering -- Mathematical models
Print OCLC #
Electronic OCLC #
Link to Catalog Record
Gaines, Larry Dale, "Process optimization by flow sheet simulation" (1974). Doctoral Dissertations. 304.