Molecular Design Of Mixed Solvents And Dynamic Simulation For Batch Extractive Distillation

Batch extractive distillation(BED) provides the advantages of batch distillation and extractive distillation, e.g. more flexibility, multi-product with a single column and low cost etc... It is very suitable for the separation of close boiling-point or azeotropic mixtures in fine chemicals and pharmaceutical industries so that increasing attention is payed to BED's related studies. Single solvent is mostly applied for extractive distillation previously, with the more and more rigorous need of product purity and economical efficiency, study on BED with mixed solvents will be the emphasis of corresponding technical research for the future.The selection of solvent and process simulation is the key problem of extractive distillation. In this paper, the molecular design method of solvent, the selection of mixed solvents and the dynamic simulation of BED were studied by theoretical analysis and experiment.In all kinds of molecular design methods of solvent currently, genetic algorithm(GA) has unique advantage for it's strong capacity of global optimization, but it's poor ability of local optimization maybe lead to a large evolutional generation and a long computation time. On the basis of characteristic of GA, the standard GA(SGA) was improved from several aspects of established population, fitness function, design of genetic operator and determination of parameter values. Then, this improved GA was combined with SQP algorithm, which has strong capacity of local optimization, to establish a high efficient blending GA(BGA) for molecular design of solvent based on Lamarckian theory. Two typical functions were tested and the result showed that BGA was superior to SGA.As for the selection method of mixed solvents, corresponding research and reporting which mostly belong to experiential or semiempirical screening are very little by far. Based on the intermiscibility of mixed solvents, the miscible parameter was defined as criterion of secondary solvent performance. The cohesive energy of preselected UNIFAC groups were calculated by group contribution method and molecular design of secondary solvent was implemented through solubility parameter theory of cohesive energy-based.Mixed solvents of BED for ethanol-ethyl acetate and acetonitrile-toluene systems were designed applying BGA. It can be found that BGA can be convergent within 100 generations which is less than SGA. DMSO and DMF at 1:2(volume ratio) are preferable mixed solvents for ethanol-ethyl acetate system, isopropyl benzene and NMP at 1:1(volume ratio) are appropriate to acetonitrile-toluene system. The VLE of ethanol-ethyl acetate and acetonitrile-toluene systems was investigated experimently and the result indicates that experimental result agrees well with that of molecular design.The BED experiment for acetonitrile-toluene system with isopropyl benzene and mixed solvents was carried out respectively and the influence of reflux ratio and solvent ratio on BED process was analyzed in this paper. The result shows that the performence of mixed solvents is superior to single solvent and the molar fraction of acetonitrile in column top can reach 0.994; reflux ratio and solvent ratio have little influence on the yield of acetonitrile; reflux ratio can be 2~3 in favor of separation of acetonitrile and isopropyl benzene at stage of acetonitrile distillated; solvent ratio of mixed solvents at 1~1.5 is adequate; reflux ratio can be moderately decreased to reduce operational time at subsequent stage.BED belongs to an unsteady state where the feed is strongly nonideal. A solution which is effective, quick and easy to run is hard to find out. To this problem, the equilibrium dynamic models of BED with mixed solvents were respectively built at total reflux start stage, total reflux with solvent stage and product distillated stage in this paper. The dynamic models and resolving process were implemented through S-function, embedded in SIMULINK model. Hence a dynamic simulation platform was developed for BED with mixed solvents based on SIMULINK. Through comparison with different solutions, it can be concluded that ODE45 is fit for the resolving of total reflux start stage and total reflux with solvent stage; ODE15s and ODE23s are fit for the resolving of product distillated stage and ODE23s is correspondingly quicker. ASPEN PLUS and PROII, two commercial simulation softwares, were adopted for the dynamic simulation of BED with mixed solvents in this paper. Compared with the simulation results of ODE15s and ODE23s, integration algorithm supplied by simulation software is only correct for total reflux start stage and total reflux with solvent stage, not the same with product distillated stage.