Application of Multiobjective Optimization in the Design and Operation of Reactive SMB and Its Experimental Verification Weifang Yu, K. Hidajat, and Ajay K. Ray* Department of Chemical and Environmental Engineering, National University of Singapore, 10 Kent Ridge Crescent, Singapore 119260 The performance of reactive simulated moving bed (SMBR) process was optimized for an experimentally verified mathematical model for the synthesis of methyl acetate ester. Multi- objective optimization was performed for an existing SMBR experimental setup, and optimum results obtained were subsequently verified experimentally. Thereafter, few other multiobjective optimization studies were performed for both existing setup and at the design stage. The effect of variable (distributed) feed flow rate on the optimum performance of SMBR was also investigated. The optimization was performed using AI-based nondominated sorting genetic algorithm (NSGA), which resulted in Pareto optimal solutions. The paper demonstrates usefulness of multiobjective optimization in the design of reactive SMB processes. Introduction The simulated moving bed (SMB) 1 is a practical way to implement continuous countercurrent operation of chromatographic separation processes. In SMB technol- ogy, the countercurrent movement of the fluid phase toward the solid phase is mimicked by switching the introduction and withdrawal ports periodically and simultaneously along a series of fixed columns in the direction of the fluid flow. For ease of operation, the columns are actually divided into sections (or zones). The number of columns within each section and the total number of columns are adjustable depending on the design of the system for any particular applications. SMB also provides opportunities for coupling reactions, 2-4 which allow higher conversion for equilibrium-limited reversible reactions by on-site separation of the prod- ucts, which leads to better yield and selectivity com- pared to typical fixed-bed processes. Additionally, the combination of two unit operations in one single ap- paratus reduces capital and operating cost. However, such integration of chemical reaction and separation complicates the process design and plant operation. The optimal design and determination of optimal operating parameters, such as switching time, flow rates in each section, and length of each column and its distribution, are therefore essential to evaluate the economic poten- tial of such processes and to successfully implement reactive SMB processes on industrial scale. The modeling, simulation, and experimental study of SMBR for the synthesis of methyl acetate (MeOAc) have been carried out details of which are reported else- where. 5,6 A rigorous mathematical model was developed to describe the dynamic behavior of SMBR for the synthesis of MeOAc. The numerical simulation results were then compared with the experimental results to validate the model. 6 Thereafter, a parametric sensitivity study was performed to investigate the effects of several design and operating parameters, such as switching time, number of columns in each section, and inlet and outlet flow rates, on the performance of SMBR. It was found that there is a complex interplay of these param- eters and some of the operating variables not only influence the yield, selectivity, and purity of MeOAc significantly but also act in conflicting ways. 6 This means any desirable change in one objective function (e.g., yield) results in an unfavorable change in another objective function (e.g., purity). In other words, it is not possible to improve yield and purity of MeOAc simul- taneously as when one is improved, the other is wors- ened. Although several studies 7-10 have been reported on the design and optimization of SMBR, they only involved single objective optimization, which is usually not sufficient for the real-life design of complex reactive SMB systems. Therefore, a more realistic approach, multiobjective optimization, 11 is necessary for the design of reactive SMB process. In this study, multiobjective optimization is used for the design of reactive SMB for the methyl acetate synthesis. Subsequently, some of the optimal solutions obtained from the optimization studies were verified experimentally to ascertain out whether optimum results can be achieved. The principle of multicriterion optimization with conflicting objectives is different from that of single objective optimization. 11 Instead of trying to find the best design solution, which is usually the global opti- mum, the goal of multiobjective optimization is to obtain a set of equally good solutions, which are known as Pareto optimal solutions. In a set of Pareto optimal solutions, no solution can be considered to be better than any other solutions with respect to all objective func- tions. When one moves from one Pareto solution to other, at least one objective function improve while at least one other worsened. So the selection of any optimal solution from a Pareto set will depend on auxiliary information. However, by narrowing down the choices, the Pareto sets does provide decision makers with useful guidance in selecting the desired operating conditions (the preferred solution) from among the (restricted) set of Pareto optimal solutions, rather than from a much larger number of possibilities. In earlier years, multiobjective optimization problems were usually solved using a single scalar objective function, which was a weighted-average of the several * To whom correspondence should be addressed. Fax: +65 6779 1936. E-mail: cheakr@nus.edu.sg. 6823 Ind. Eng. Chem. Res. 2003, 42, 6823-6831 10.1021/ie030387p CCC: $25.00 © 2003 American Chemical Society Published on Web 11/19/2003