(84c) Dynamic Modeling and Optimization of Thermally Coupled Dividing Wall Column Batch Distillation Processes | AIChE

(84c) Dynamic Modeling and Optimization of Thermally Coupled Dividing Wall Column Batch Distillation Processes

Authors 

Joglekar, G. S. - Presenter, Batch Process Technologies, Inc.
Agrawal, R. - Presenter, Purdue University
Reklaitis, G. V. - Presenter, Purdue University


A novel, dividing wall column batch distillation process has been developed to distill multicomponent mixtures into multiple pure product streams. Unlike in continuous columns, the dividing wall in the proposed batch column extends upto the bottom of the column, creating two chambers in the lower section of the column. Each lower chamber has its own reboiler and heat input. Like a conventional batch column, there is a condenser at the top. An important feature of this process is that it can operate as a conventional batch distillation column if necessary. The dividing wall batch distillation processes are expected to significantly reduce the capital cost as it requires only one column shell. The potential applications are in pharmaceutical, specialty chemical and agricultural industrial facilities that utilize a wide range of solvents.

In this study, a detailed process dynamics model of the thermally coupled columns was constructed. Distefano's quasi steady-state formulation(2) was used for mass balance on all plates, while the rest of the variables were updated after the specified elapsed time. The important variables that affect the column performance are the reflux ratio, cut durations and fraction of liquid stream within the column that is split between the two chambers created by the dividing wall. A mixture of Benzene, Toluene and Ortho-Xylene with mass fractions 0.33, 0.33 and 0.34, respectively, was charged to the column.

The following key operating variables associated with each cut affect the overall column performance: reflux ratio, duration and fraction of liquid stream within the main column that is diverted to the side column. To find the optimum values of the operating variables, a two-level approach was used. The Genetic Algorithm function in Matlab was used to create the members in the population. The fitness function value for each member was computed by simulating the column dynamics. The objective function value of each member in turn was used by the GA to create new population.

For a three component mixture of Benzene, Toluene and Ortho-Xylene, the optimum cycle time for the dividing wall column process is reduced by 34% compared to the optimum cycle time for a conventional column; also slop cut is eliminated leading to environmentally friendly solution.

1. AIChE presentation, 2007 Annual Meeting, paper 211a.

2. AIChE Journal, 14(1), 190-199 (1968).