(311f) A Node-Edge Representation for Reactor Network Synthesis

Reactor network synthesis problem is one class of chemical process synthesis problem that determines the reactor network to transform the given raw materials into desired products [1]. It has wide applications in identifying reaction pathways for waste reduction, methane conversion, and polymerization [1-3]. Optimization-based reactor network synthesis problem consists of three steps: (1) graph (superstructure) representation, (2) mathematical modeling of superstructure, and (3) solution of mathematical model [4]. The richness of graph representation that embeds potential reaction pathways influences the tractability of mathematical model and finally the quality of optimal solution [5,6]. Despite many reported graph representations for reactor network synthesis problem [1-6], open questions remain regarding: (1) How does superstructure representation influence the tractability of mathematical model for reactor network synthesis problem? (2) How can a generic and flexible graph representation be developed to enable efficient solution of reactor network synthesis problem?

In this work, we propose a generic graph representation for reactor network synthesis problem. In the proposed graph representation, nodes allow feed assignment, product withdrawal, assignment of reaction operation, and assignment of hot and cold utility. In each node with reaction operations, we use continuous stirred tank reactor (CSTR) as building block and assembly of these nodes to represent plug flow reactors (PFRs), maximum mixed reactors (MMR), segregated flow reactors (SFRs), and cross-flow reactors (CFRs). Edges among nodes allow material and energy flow. We model this graph representation as a mixed-integer nonlinear programming (MINLP) problem to design reaction networks that maximize production profit. We propose a preprocessing method to remove redundant edges representing material and energy flow in the graph representation to improve the computational performance of the corresponding MINLP model. We apply the proposed graph representation and the corresponding MINLP model to design reactor networks. Moreover, we benchmark the computational performance of the proposed model with other state-of-the art models based on graph representations from literature.

[1] Schweiger, C.A. and Floudas, C.A. Optimization framework for the synthesis of chemical reactor networks. Industrial & Engineering Chemistry Research, 1992, 38(3), pp.744-766.

[2] Lakshmanan, A., Rooney, W.C. and Biegler, L.T. A case study for reactor network synthesis: the vinyl chloride process. Computers & Chemical Engineering, 1999, 23(4-5), 479-495.

[3] Diaconescu, R., Tudose, R.Z. and Curteanu, S. A case study for optimal reactor networks synthesis: styrene polymerization. Polymer-Plastics Technology and Engineering, 41(2), 297-326, 2002.

[4] Li, J., Demirel, S.E. and Hasan, M.M.F. Process synthesis using block superstructure with automated flowsheet generation and optimization. AIChE Journal, 64(8), 3082-3100, 2018.

[5] de Lima, A.E.P. and Maravelias, C.T. A generalized framework for reactor network synthesis: a graph theoretic approach. Computers & Chemical Engineering, 160, 107722, 2002.

[6] Mencarelli, L., Chen, Q., Pagot, A. and Grossmann, I.E. A review on superstructure optimization approaches in process system engineering. Computers & Chemical Engineering, 136, 106808, 2020.