(32c) A Synthesis Framework for Hybrid Separation Sequences Based on Reduced Directed Graph Superstructure

Chemical engineering process usually involves separation of multi-component mixtures, the energy consumption and capital cost of separation process takes a great part in the whole investment, thus separation process is of great importance in process system engineering. Distillation, which is still the major separation technique in chemical engineering, with the advantages of the low capital investment, the operational flexibility, and the low operational risk. However, when distillation boundaries or low relative volatilities complicate distillation, it is desirable to substitute distillation with unit operations like extraction, crystallization, adsorption, or membrane separation, if possible. The key problem of this hybrid separation synthesis is to choose appropriate separation techniques and determine optimal separation sequence.

Previous work mainly focused on synthesis between two separation techniques^[1-4], when extraction, crystallization, adsorption, and membrane separation are introduced into the synthesis framework simultaneously, as there are many candidates which can be found in separation techniques above, problem of combinatorial explosion becomes apparent. This problem was usually solved using heuristic sequential search strategy^[5], however, this strategy usually leads to a suboptimal solution. Besides, rules of thumb in the past were not applicable in this problem because of their weakness in handling with too many choices.

In this work, a synthesis framework is proposed to reduce the space of separation directed graph by introducing matrix analysis. First, maturity of separation techniques, requirement for feed composition and product purity was considered in this methodology, aiming to preliminary screen feasible techniques. Then for the candidates of techniques which can complete the same separation task, shortcut models were introduced to determine the cost of different candidates. In this way, disadvantaged candidates would be removed from set of technique, and the space of set decreased.

Considering the separation characters of technique, separation technique–component incidence matrix analysis was proposed in this methodology: directed graph of distillation sequence was generated first to build basic set of distillation task. Then the separation tasks of non-distillation techniques were derived from top to bottom, from more components to fewer components, using the incidence matrix mentioned above. It was worth noticing that new generated tasks were added into the basic set of task continually. After this, the entire separation directed graph was completed, all illegal separation tasks were removed, and solution space decreased significantly.

To improve the efficiency of searching solution space, based on ordered branch search strategy^[6], the separation sequence was searched in the order: the non-distillation separation which separate multiple components at a time would be ranked first; for components which were separated from mixture, if the relative volatility between them was large, this task would be ranked front; the non-distillation separation prior to distillation separation. After this, the advantaged set of separation sequence were obtained. Then rigorous model was introduced to verify the cost of these sequences.

Two illustrative examples were presented to demonstrate the validity and advantages of the proposed approach. Mixtures which contained twelve components from C₆-C₈ and mixtures which contained fifteen components from C₈-C₁₀ were going to be separated, where solvent extraction, pervaporation, adsorption and crystallization techniques were introduced. The number of sequence in the new directed graph was less than 1% of the original one.

Reference:

[1] Errico M, Sanchez-Ramirez E, Quiroz-Ramìrez J J, et al. Synthesis and design of new hybrid configurations for biobutanol purification[J]. Computers & Chemical Engineering, 2016, 84: 482-492.

[2] Beierling T, Micovic J, Lutze P, et al. Using complex layer melt crystallization models for the optimization of hybrid distillation/melt crystallization processes[J]. Chemical Engineering and Processing: Process Intensification, 2014, 85: 10-23.

[3] Ritter J A, Wu F, Ebner A D. New approach for modeling hybrid pressure swing adsorption–distillation processes[J]. Industrial & Engineering Chemistry Research, 2012, 51(27): 9343-9355.

[4] Harvianto G R, Ahmad F, Lee M. Vapor permeation–distillation hybrid processes for cost-effective isopropanol dehydration: modeling, simulation and optimization[J]. Journal of Membrane Science, 2016, 497: 108-119.

[5] Floquet P, Pibouleau L, Aly S, et al. Fuzzy heuristic approach for sharp separation sequence synthesis[J]. Computers & chemical engineering, 1994, 18(9): 785-795.

[6] Rodrigo B, Fernando R, Seader J D. Synthesis of separation sequences by ordered branch search[J]. AIChE Journal, 1975, 21(5): 885-894.