Global optimization of nonconvex MINLP problems in process synthesis
journal contributionposted on 1987-01-01, 00:00 authored by Gary R. Kocis, Ignacio E. Grossmann, Carnegie Mellon University.Engineering Design Research Center.
This paper addresses the solution of nonconvex MINLP problems for process synthesis through the Outer-Approximation/Equality-Relaxation algorithm. A two-phase strategy is proposed where in phase I nonconvexities that may cut off the global optimum are systematically identified with local and global tests. In phase II, a new master problem is proposed that attempts to locate the global optimum which may have been overlooked in phase I. The proposed procedure, which has been implemented in DICOPT, is illustrated with several examples that include the design of multiproduct batch plants and a structural flowsheet optimization problem.