Computational Experience Solving the Design of Multiproduct Batch Plants

J.M. MONTAGNA; A. R. VECCHIETTI

AICHE Symposium Series

CACHE Publications

Año: 2000; p. 294 - 297

This notes present the computational experience in the solution of a set of selected optimization problems for the design of multiproduct batch plants. Two representations for the model are assessed: MINLP and  disjunctive. The model is formulated with different degrees of difficulties: from weakly  non-linear convex to non-convex form of the constraints. For the solution of the MINLP problems two algorithms are applied: Extended Cutting Plane (ECP) and Outer Approximation/Equality Relaxation/Augmented Penalty (OA/ER/AP). DICOPT++ is the solver selected for the OA/ER/AP algorithm. A DICOPT++ modification linked to GAMS has been developed for the ECP algorithm. LOGMIP is used for the solution of the problems with the disjunctive formulation. The objective is to analyze the performance of the different approaches solving several problems. The goal is to increase the knowledge about the formulation behavior and the algorithm to apply to the design of batch plants.