(376bv) Multi-Objective Optimization of a Batch Distillation Column

Over the years, intuition and experience based heuristic approaches are used for the development and operation of batch processing in a distillation column. An optimal operating condition of a batch distillation may lead to run the column most efficiently in terms of both utility consumption and cost. Robust optimization like genetic algorithm has a significant advantage over gradient-based optimization to find the global optimal point due to non-convexity of design space. This work aims at formulating a multi-objective optimization (MOO) strategy to optimize the design and operating parameters of a conventional batch distillation (CBD), probably for the first time in the literature. Elitist non-dominated sorting genetic algorithm (Deb et al., 2002) is employed for the development of the optimization strategy along with the selection of the optimal point by using the technique for order of preference by similarity to ideal solution method (TOPSIS) along with entropy information for weighting method. Two objective functions in the aspect of performance criteria, i.e. total annual cost (TAC), and total annual production (TAP) are considered, which are often conflicting in nature. The number of trays, reboiler heat duty and reflux ratio are considered as decision variables to be optimized. The Pareto optimal front is constructed from the stated MOO studies, and then the optimal point is chosen by the TOPSIS-entropy weighting method. A CBD column separating a mixture of acetone and water (Miladi and Mujtaba, 2004) is considered for the demonstration of the proposed optimization strategy. The optimum value for the TAC and TAP are found to be 94689 $/year and 65610 kmol/year, respectively; and the corresponding decision variables are 22 trays (excluding the reboiler and total condenser), constant reboiler heat duty of 8204 kJ/min and a constant reflux ratio of 0.72.

References

Deb K., Pratap A., Agarwal S., Meyarivan T., A fast and elitist multi-objective genetic algorithm: NSGA-II, IEEE Tran. Evol. Comput. 6 (2002) 182-197.

Miladi M.M., Mujtaba I.M.. Optimization of design and operation policies of binary batch distillation with fixed product demand. Comput. Chem. Eng. 28 (2004) 2377–90.