(686b) Simulation-Based Derivative-Free Optimization for Hybrid Separation Design

As the chemical process industry continues in an energy-saving, high efficiency, and sustainable direction, researchers concentrate on redesigning/retrofitting existing facilities to achieve these goals. Distillation, as the most widely applied separation technique, is energy-intensive and has the lowest thermal efficiency [1]. New innovative separation technologies such as hybrid separation and intensified equipment provide new pathways to retrofitting the original distillation column(s) to lower energy costs and increase the process profitability. Implementation of these innovative separation methods requires the identification of optimized process configurations and associated operational parameters to balance any additional capital cost and expected utility savings. Caballero et al. [2] have presented an algorithm to identify the optimum hybrid separation configuration, which is formulated as a mixed integer nonlinear programming (MINLP) problem and solved it mathematically. Tula et al. [3] pointed out that without significant additional capital investment, a hybrid distillation membrane process can achieve reductions of 15-20% with respect to the energy consumption of the entire retrofitted process. O’Connell et al. [4] evaluated the influence of the switching composition in hybrid distillation membrane separation and have shown that it dramatically influences the retrofitted processes' economic feasibility. Therefore, a systematic means of identifying the optimal switching composition for hybrid distillation is needed to assist in the implementation of hybrid distillation separation.

This work focuses on a mathematical optimization-based approach for designing hybrid distillation columns. Here, the problem is formulated to select the optimal switching composition in hybrid distillation separation systems so that the annualized investment cost is minimized. The objective function, minimization of annualized investment cost, considers both the capital cost incurred by new equipment and also the operating cost associated with the entire separation task. Even though a trial and error based simulation run can help to identify the optimal point, it can be time-consuming or even impossible when dealing with complicated processes. An alternative approach is to apply derivative-free optimization algorithms, such as Bayesian optimization, to identify the optimal point. This approach has two levels (inner and outer loop). In the inner-loop, process simulations are performed, and the results are used to calculate the objective function, corresponding to the design decisions. The results from the inner-loop (objective function value and the associated decision variable values) are sent to the outer-loop where a suitable solver based on a gradient-free optimization algorithm is employed to obtain new/improved values of the decision variables such as design parameters. This iterative calculation continues until convergence of the outer-loop is achieved.

This optimization-based approach was used to solve a range of case studies, including alkane isomers separation, water-methanol. In all case studies, the developed approach was able to find the best switching composition for designing hybrid distillation columns. Furthermore, this approach can also be used to identify different hybrid schemes such as different membranes, adsorbents.

[1] Pellegrino, J.L., Margolis, N., Justiniano, M., Miller, M., Thedki, A. (2004), Loss and Opportunities Analysis. US Manufacturing & Mining.

[2] Caballero, J. A., Grossmann, I. E., Keyvani, M., & Lenz, E. S. (2009). Design of hybrid distillation− vapor membrane separation systems. Industrial & engineering chemistry research, 48(20), 9151-9162.

[3] Tula, A. K., Befort, B., Garg, N., Camarda, K. V., & Gani, R. (2017). Sustainable process design & analysis of hybrid separations. Computers & Chemical Engineering, 105, 96-104.

[4] O'Connell, J. P., Eden, M. R., Tula, A. K., & Gani, R. (2019). Retrofitting Distillation Columns with Membranes. Chemical Engineering Progress, 115(12), 41-49.