Heavily constrained, mixed integer optimization problems present a difficult test for optimization algorithms. This paper examines optimization of the ethylene process (from ethane) using a flowsheet simulation model and the adaptive random search. Optimization resulted in an improvement in annual earnings of $1.15 MM over the usual ethylene process design. This process has both continuous and discrete variables and has sensitive implicit constraints in the pyrolysis furnace. The search efficiency was found to be improved by relaxing these constraints early in the search and then restoring the constraints as the optimum is approached. The search efficiency was also improved by testing each independent variable for its optimal explicit bound and using these bounds as the starting point for the optimization. Significant improvements in the economics ($0.7 MM per year) were found by examining and slightly adjusting the variable boundaries at the optimum. These results suggest that variable bounds should be established in conjunction with the optimization study. © 1979, American Chemical Society. All rights reserved.


Chemical and Biochemical Engineering

International Standard Serial Number (ISSN)


Document Type

Article - Journal

Document Version


File Type





© 2023 American Chemical Society, All rights reserved.

Publication Date

01 Oct 1979