点击次数：

论文类型：期刊论文

发表时间：2019-07-20

发表刊物：CHEMICAL ENGINEERING SCIENCE

收录刊物：SCIE、EI

卷号：202

页面范围：300-317

ISSN号：0009-2509

关键字：Computer-aided molecular design; Reaction solvent; COSMO-SAC; Solvation effect; Decomposition-based algorithm

摘要：Solvents have been widely used in chemical manufacturing processes. When involved in liquid homogeneous-phase kinetic reactions, they can have significant impacts on the reaction product yield. In this paper, an optimization-based framework is developed for reaction solvent design. The framework first identifies a reaction kinetic model using a hybrid method consisting of three steps. In step one, a rigorous thermodynamic derivation based on CTST (Conventional Transition State Theory) is performed to formulate a primary reaction kinetic model. In step two, a knowledge-based method is used to select additional solvent properties as supplementary descriptors to account for quantitative correction to the model and thereby improving the prediction accuracy. In step three, model identification is performed to obtain the best regressed reaction kinetic model. This hybrid modelling method is tested through two case studies, namely Diels-Alder and Menschutkin reactions, and an impressive consistency of the results is observed when the infinite dilution activity coefficients (calculated by COSMO-SAC model), hydrogen-bond donor, hydrogen-bond acceptor and solvent surface tension are selected as descriptors in the final reaction kinetic model. The GC-COSMO and GC (Group Contribution) methods are combined for the prediction of these descriptors. Finally, the Computer-Aided Molecular Design (CAMD) technique is integrated with the derived kinetic model for reaction solvent design by formulating and solving a Mixed-Integer Non-Linear Programming (MINLP) model. A decomposition-based solution algorithm is employed to manage the complexity involved with the nonlinear COSMO-SAC equations. Promising reaction solvents are identified and compared with those reported by others, indicating wide applicability and high accuracy of the developed optimization-based framework. (C) 2019 Elsevier Ltd. All rights reserved.