A Study on Optimal Design of Single Periodic, Multipurpose Batch Plants

The purpose of this paper is to describe the design of a general multipurpose batch process or plant in terms of a series of mathematical programing models, and to develop approach solution methodologies. The proposed model for a single period is based on the formulation (MINLP; Mixed Integer Nonlinear Programming) of Papageorgaki and Reklaitis [1], but was linearized (MILP; Mixed Integer Linear Programming) so as to obtain an exact and practical solution, and to allow treatment of uncertainties to be considered in expanding the plant. As a solution strategy a modified Benders' Decomposition was introduced and was tested on three example problems. The optimizing solver, OSL code provided by the IBM Corporation, was used for solving the problems. The solution method was successful in that it showed remarkable reduction in the computing times as compared with the direct solution method.

본고는 다종, 소규모 회분식 공정 또는 공장을 수학적 프로그래밍 기법으로 최적 설계하는 방법에 대한 것이다. 제안된 일반 회분식 공장문제는 Papageorgaki와 Reklaitis에 의해 혼합정수비선형식(MINLP)으로 수립된 것인데, 최적해답을 보장하며 공장의 확장 등 불확실성을 고려하여 선형화 한 후(MILP) 푸는 방법론이 제시되었다. Bende식 문제분할 방식을 개조하여 몇가지 예제에 대한 풀이를 제시하였다. IBM의 OSL 최적화 패키지를 이용하였고 MILP를 직접 푸는 경우보다 계산시간을 크게 단축할 수 있었다.