Validation of a Fast Solvent Evaluation and Selection Model for Extractive Distillation

Extractive distillation separates a two-component mixture of close-boiling or azeotropic components by adding a third component (entrainer/solvent) to the process. The selection of the third component is critical because it affects the column design and cost. Here, we validate a Fast Solvent Evaluation and Selection (FSES) model [1] by combining process simulation with derivative-free optimization (DFO). The model [1] employs physical properties, such as boiling point, selectivity, and molecular weight, and process properties, such as minimum reflux ratio, minimum reboiler duty, and column stages to predict solvent performance and rank the solvents.

The framework for validating the FSES model integrates a rigorous extractive distillation process simulation in the inner loop with a DFO algorithm in the outer loop iteratively. Given a separation task, the goal is to find the optimum process design and operating conditions, i.e., inlet stages, solvent ratio, column stages, distillate rates, and reflux ratios, for each solvent that the FSES model ranked. The optimum is defined as the design and operating conditions that minimize the total annualized cost (TAC). In each framework iteration, the DFO algorithm generates a set of design and operating conditions and passes them to the process simulation. Once the simulation is converged for the conditions, the total annualized cost is calculated and passed back to the DFO algorithm, starting a new iteration. This iterative process continues until the DFO algorithm no longer generates candidate design and operating conditions with lower TAC values or the maximum number of simulation runs is reached. At termination, the framework yields the minimum TAC and the optimum design and operating conditions for each solvent. Then, the candidate solvents are ranked from lowest to highest based on their minimum TACs, and these rankings are compared to those obtained from the FSES model.

We applied the validation framework to two case studies, acetone/methanol and benzene/cyclohexane separations. For the acetone/methanol separation, the FSES ranked the solvents from best to worst as dimethylsulfoxide (DMSO), water, ethylene glycol, ethanol, and 2-propanol. The framework yielded the same order. For the benzene/cyclohexane separation, the FSES model ranked the solvents from best to worst as sulfolane, furfural, dimethyl phthalate, N-methyl-2-pyrrolidine (NMP), and aniline. The framework yielded the same order, except it ranked NMP as a better solvent than dimethyl phthalate.

References

[1] Xu, S., Crump, T., Cremaschi, S., Eden, M. R., & Tula, A. K. A Short-Cut Method for Synthesis of Solvent-based Separation Processes. In preparation.