Optimal design of superstructures for placing units and streams with multiple and ordered available locations. Part II: Rigorous design of catalytic distillation columns. (4th August 2020)

• Record Type: Journal Article
• Authors: Liñán, David A.; Bernal, David E.; Ricardez-Sandoval, Luis A.; Gómez, Jorge M.
• Is Part Of: Computers & chemical engineering. Volume 139(2020)
• Publication Date: 2020-08-04
• Subjects: Catalytic distillation -- Optimal design -- Decomposition strategy -- Discrete-steepest descent algorithm; CD catalytic distillation -- CPU central processing unit -- ECP extended cutting plane -- ETBE Ethyl tert-butyl ether -- FNLP fixed-nonlinear program/problem -- GBD generalized benders decomposition -- GDP generalized disjunctive programming/program -- MINLP mixed-integer nonlinear programming/program/problem -- IPLC integer program/problem with linear constraints -- NLP nonlinear programming/program/problem -- OA outer-approximation -- BB branch and bound -- TAC total annualized cost; Chemical engineering -- Data processing -- Periodicals; 660.0285
• DOI: 10.1016/j.compchemeng.2020.106845
• Languages: English
• ISSNs: 0098-1354

Abstract

Highlights: Large application of the Discrete-Steepest Descent Algorithm (D-SDA) is presented. A case study involving a catalytic distillation column is presented. MINLP of a catalytic distillation superstructure is considered. The D-SDA found better solutions than the available deterministic MINLP solvers. Abstract: This work addresses the optimal design of catalytic distillation (CD) columns taking into account discrete design variables, i.e., the number of stages and location of feed stages and reactive stages. This optimization problem is challenging due to the combinatorial complexity introduced by these discrete decisions. In this work, the binary variables of the superstructure are expressed as a function of a reduced variable set: the external variables, which allows the use of a local algorithm that solves a series of master and primal sub-problems introduced in our previous study (Part I ). The master problem is constructed from the reduced variable set and solved with a Discrete-Steepest Descent Algorithm (D-SDA). The primal sub-problem is a nonlinear programming problem obtained by fixing the binary terms in accordance with the reduced variable set. This strategy avoids finding multiple local minimizers, thus making this methodology more practical and efficient than existing Mixed-Integer Nonlinear Programming (MINLP) local and global optimization solvers. The proposed approach was evaluated on a catalytic column used for the production of ethyl tert-butyl ether (ETBE). It was found that an alternating distribution pattern of the reactive trays, the optimization of the location of feed stages, and the optimization of the total number of trays results in a more cost-effective CD column. Graphical abstract: Image, graphical abstract