(374a) Tailor-Made Solvent and Process Design for the Separation of Azeotropic Mixtures
AIChE Annual Meeting
2016
2016 AIChE Annual Meeting
Computing and Systems Technology Division
Process Design I
Tuesday, November 15, 2016 - 12:30pm to 12:49pm
Marta Gonzalez Garcia1, Ioulietta Paraskeva1, Anjan Kumar Tula1, Emmanouil Papadakis1, Deenesh Kavi Babi2, Rafiqul Gani1
1SPEED, Department of Chemical and Biochemical Engineering, Technical University of Denmark, Soltofts Plads 1, Building 229, DK-2800 Lyngby, Denmark
2Utility and Solvents, Department of Insulin Manufacturing 1, Novo Nordisk A/S, Hallas Allé DK-4400 Kalundborg, Denmark
A typical example of a process that does not require the addition of a solvent for achieving the desired separation of an azeotropic mixture is pressure swing distillation. An important condition for its operational feasibility is that the azeotrope must be pressure sensitive over a moderate change in pressure1. However, if the azeotrope does not satisfy this condition or the process is energy intensive and/or environmentally unfriendly then a promising alternative is a solvent â?? based separation process2. The solvent â?? based process configuration for a two phase system (vapor-liquid) consists of an extraction column followed by a solvent recovery column and the process design together with solvent selection is usually obtained through a trial and error approach (heuristics, mathematical programming, and/or hybrid). However, can the design be obtained without any trial and error? It is possible by first determining the process design for a desired (target) solvent and then finding a solvent that matches the desired solvent properties. This decomposition of the design problem is feasible because the process design is based on the properties of the system and does not need the identities of the chemicals involved. On the other hand, the selection problem needs the identities of the chemicals involved but not details of the process. Also, a unique solvent-azeotropic mixture cannot be obtained for all homogeneous azeotropic mixtures. This means that the azeotropic mixtures need to be classified first and for each class, the optimal solvent behavior and the corresponding apriori process design determined. Also, can the solvent-based separation problem further improved in terms of energy consumption for the same separation specification?
Assuming a desired solvent-azeotropic mixture behavior and a set of solvent properties, the design of the two column process can be determined, for example, in terms of product purities, number of stages, feed location, and reboiler-condenser duties without knowing the identity of the solvent. This means that having apriori designed the separation process, if a solvent that matches the targeted behavior and properties could be found, a simultaneous tailor-made solvent and process design could be very easily achieved. The conceptual basis for this method is that solvent properties have a direct and unique influence on the separation process behavior, for example, separation feasibility is defined by the distillation boundary (corresponding to the ternary system of solvent and the azeotropic mixture); energy consumption is influenced by solvent capacity and solvent pure component properties; waste generation and environmental impact is influenced by solvent selectivity and solvent environmental properties. By using the diving-force concept3, it is ensured that for the apriori design, the extraction and recovery columns operate at the maximum driving force, which translates into the design of a more sustainable energy efficient process.
The following reverse design problem consisting of two sub-problems is formulated and solved: Sub-problem 1: Given ternary phase equilibrium data, driving force diagrams and properties of the chemical system, determine the optimal energy efficient design of the extraction and the solvent recovery columns. Sub-problem 2: Given, a homogenous azeotropic mixture and an apriori process design; find the best (optimal) solvent that satisfies the process design in the form of targeted solvent-azeotropic mixture behavior plus solvent properties such as normal boiling point, heat of vaporization, solubility of solute and many more. Note that matching these targets implicitly mean satisfying all the process constraints such as mass and energy balance, separation feasibility and energy demands. Sub-problem 3: Replace the solvent recovery column (SRC) with a liquid-liquid extraction column plus an evaporator (LL+E) by swapping the original solvent with a swap solvent. The swap solvent takes the solute from the original solvent in the liquid-liquid extraction column and the solute is removed by simply heating the solute-swap solvent mixture. The original solvent is recycled back to the original extraction column and the swap solvent is recycled to the liquid-liquid extraction column. The energy needed for the removal of the solute by evaporation is significantly lower than that required by the SRC using the original solvent.
The main features of the method are: classification of the homogeneous azeotropic mixtures, classification of the solvent-azeotropic mixture behavior; classification of the swap solvent mixture behavior, apriori process design based on driving force, solvent-azeotropic mixture behavior and swap solvent-original solvent-solute mixture behavior; finding solvents and swap solvents that match the design targets; and a database containing apriori process design from which designs of other problems can be extrapolated or interpolated. For each class of azeotropic mixture in the database, a list of pre-selected solvents is already available. Therefore, for a separation involving a known azeotrope, the optimal solvent (or swap solvent) and process design are obtained with minimal calculations. For an azeotropic mixture not available in the database, simply identifying the class of the azeotropic mixture, all other necessary details can be extrapolated or interpolated from the available data.
 The presentation will highlight the concepts and the method for tailor-made solvent and swap solvent selection and process design, together with tools developed to implement and test this method. The method comprises of 4 â??stages. In stage 1, the azeotropic mixture class is identified. For this stage, a database of 43 homogenous azeotropic mixtures that have an azeotropic composition between 30-70% mol is developed using available experimental data. In stage 2, the solvent target behavior is retrieved for the identified azeotropic mixture from a second database. For this stage, a classification of the solvent-azeotropic mixture behavior is developed using a graphical approach based on ternary plots and the position of the distillation boundaries. The distillation boundary of interest is drawn from the azeotropic point (located on the binary vertex of the two compounds that constitute the azeotrope with compound 1 not removed by the solvent) until it touches and overlaps the vertex of compound 2-solvent binary line. The classification is done between compound 2 and solvent, that is, high or low composition (more or less than 50%mol) of compound 2 in the azeotrope and high or low composition (more or less than 50%mol) of solvent in the distillation boundary overlap. For example, if a mixture has an azeotrope composition of 40%mol of compound 2 (low) and a distillation boundary overlap of 60% of solvent composition (high), it will be classified as Low-High. Therefore, the 43 azeotropic mixtures are classified into four groups: High-High, High-Low, Low-High and Low-Low. This behavior also requires that the solvent must not form azeotropes with either of the compounds and that the ternary system must be totally miscible. In stage 3, based on the selection from stages 1 & 2, the corresponding driving force based apriori design is retrieved from the database. For this stage, validated process designs by rigorous simulation for all classes of azeotropic mixtures and solvent-azeotropic behavior are determined and stored in a database. Also, additional information that helps to extrapolate from the stored data are provided. For example, plots of solvent flow rate vs composition of solvent in the compound 2-solvent mixture and solvent flow rate vs driving force. In stage 4, the solvent selection problem is solved by database search and/or standard solvent selection-design techniques (3). Final design is verified through rigorous simulation using as initial estimate, the retrieved apriori design data. For the swap-solvent selection, a target liquid-liquid phase boundary is defined together with properties such as melting points and boiling points, which must be much greater than the solute.
The method is generic and able to directly determine the best solvent, solvent flow rate, the driving force and the energy required for separation of the classes of homogeneous azeotropic mixtures available in the database. It allows safe extrapolations for azeotropic mixtures not covered in the database but that match the classified behavior. The method has been tested for a wide range of homogeneous azetropes and a collection of these results will be presented. Until the final rigorous simulation verification stage (for which also a good initial estimate is provided), the method hardly needs any calculations and is able to provide more sustainable and near optimal simultaneous selection of solvents and its corresponding optimal process design. The concept and method, can in principle, be applied for many other simultaneous product-process design problems. The savings is not only in energy and material but also time to operate and solve problems. References
1. Seader, J. D., Henley, Ernest J. y Roper, Keith. Separation Process Principles. s.l. : John Wiley, 2011.
2. Igor Mitrofanov, Sascha Sansonetti, Jens Abildskov, Gürkan Sin and Rafiqul Gani, 2012, â??The Solvent Selection framework: solvents for organic synthesis, separation processes and ionic-liquids solventsâ?, Computer-Aided Chemical Engineering , 30, 762-766
3. R. Gani, E. Bek-Pedersen, 2001, â??A Simple New Algorithm for Distillation Column Designâ?, AIChE Journal, 46 (6), 1271-1274