Grid-search-and-box-search-assisted coordinate descent methodology for practical retrofit of the existing distillation columns to dividing wall columns

Abstract

A dividing-wall column (DWC) leading to savings in terms of substantial investment and operating costs, and a significant reduction in CO2 emission constitutes an optimal solution for distillation process intensification. Recently, industrial applications of the DWC gradually increased although the need for an efficient, practical, and simple to implement optimization methodology persists. This study proposes a practical method that employs coordinate descent methodology, which is assisted by a grid search to avoid local optimal points and box searches to seek a more promising solution to optimize the retrofit of existing distillation columns to DWCs. The optimal DWC structure and operating conditions are determined in a practical and effective manner that is simple to implement and requires minimal simulation runs, which is suitable for industrial uses. The proposed method is examined in the optimal retrofit of a side stream column and a conventional two-column sequence that are widely used in the chemical process industry. The results show that the grid-search-and-box-search-assisted coordinate descent methodology combining mathematical programing techniques and statistical methods effectively avoids local optimal points but also corresponds to a more promising solution, and thus, it is competitive to coordinate descent methodology and response surface methodology that are popularly used in optimization processes in the chemical process industry. The retrofitted DWC systems lead to a substantial decrease in operating costs while effectively removing the bottleneck problem and mitigating CO2 emission as well. Notably, operating costs are reduced to a maximum of 33.8% and 43.3% in retrofit of side stream columns and conventional two-column sequences, respectively. Both structural and operating variables are effectively and simultaneously optimized. It allows for the identification of interactions between design variables. Furthermore, the carbon dioxide emission is calculated and evaluated when retrofitting to the DWC.