DETECTING GLOBAL OPTIMAL STRUCTURES OF HEAT

LORENA BERGAMINI,; AGUIRRE, PIO; NICOLAS SCENNA,

Rio de Janeiro

Congreso; 2nd. Mercosur Congress on Chemical Engineering. 4th. Mercosur Congress on Process System Engineering; 2005

A new methodology for the global optimization of heat exchanger networks is presented, based on an outer approximation methodology, aided by physical insights. The problem is formulated as a MINLP problem and it considers a staged superstructure. Two lower bounding MILP master problems are constructed, such that they are outer approximations of the original problem and the NLP problem representing a fixed structure, respectively. The approximating problems include piecewise underestimators of the nonconvex terms, constructed over a grid point set. A solution of the bounding problems gives an approximated optimal solution that is used as initial point for solving the reduced NLP problem, or proves that there is no better solution. Meanwhile, the outer master problem selects feasible structures, which are fixed when NLP is solved. The algorithm explores iteratively each potential optimal structure, and finds the optimal level of operation of them. Each iteration tries to improve the best know solution. An integer cut is added to the bounding problem to prevent any explored structure to be selected in subsequent iterations. In order to reduce the number of feasible structures to be explored, rigorous constraints obtained from physical insights are provided to the algorithm. In this way, suboptimal structures are ignored in the search. Several examples show the performance of the proposed methodology.