Optimal synthesis of flexible heat exchanger networks

In-Beum Lee, Purdue University

Abstract

Systematic procedures for the synthesis and design of heat exchanger networks (HEN) have been under study for some 20 years with over 100 papers published on the subject. Progress has been made towards synthesizing networks with maximum energy recovery (MER) and minimum number of units (MNU) which exhibit local minima with respect to the investment cost. However, satisfactory methods do not exist for determining the global minimum cost network; rather, most of the work has focused on improving the initially synthesized local MER and MNU solutions through merging, loop-breaking and split/bypass optimization. In this work, a systematic, but ultimately heuristic, procedure is described which, beginning with an initial feasible MNU/MER network, generates a sequence of such networks with successively improving annual cost. We first examine unpinched problems and then present extensions which handle pinched problems with multiple subnetworks. The key steps in the procedure consist of path tracing/list processing constructions that allow development of networks which are in some sense adjacent to the initial network while retaining the MER and MNU features. The approach is evolutionary in nature, involving a set of rules. However, the properties of the constructions and procedures are rigorously explored and the effectiveness of the composite algorithm is demonstrated by many test problems. These tests show that the proposed approach can find the optimum network for the known standard problems, and for several problems, such as 4SP2 and TC-2, new optima are identified which to date have not been reported in the literature. For heat exchanger networks which are required to accommodate multiple-periods of operation, feasible networks are synthesized at each period and then these networks are combined to form a feasible super network. The super network guarantees MER at each period and features MNU. For the general problem with uncertain parameters, all the flexible networks which can cope with the uncertainties are enumerated to generate the flexible minimum cost network after a flexible network is obtained from an MNU/MER network synthesized at the nominal parameters. The procedures are made up of must-match search and path processing constructions that allow not only a combination of networks but also development of flexible networks adjacent to the initial one. For all the synthesized networks, the flexibility is analyzed over the given ranges of uncertain parameters.

Degree

Ph.D.

Advisors

Reklaitis, Purdue University.

Subject Area

Chemical engineering

COinS