Azeotropic Distillation Problem & solution principle Ethanol cannot be obtained through simple distillation from an ethanol-water mixture because of an azeotrope. Different methods are available for this purpose. Here we will introduce azeotropic distillation using n-pentane as entrainer (see fig. 1). In a pre-treatment step, an ethanol-water mixture is concentrated to approx. 90% ethanol content (see fig. 1, column 1). Afterwards, the mixture is fed to a column together with the n-pentane, which results in a lowboiling ternary in the head (see fig. 1, column 2) that disintegrates into two liquid phases after condensation. The wet phase is then discharged into a flash and the arid phase returned to the column (see fig. 1, flash 5). Through formation of the ternary azeotrope and expulsion of the water, the ethanol turns into a high boiler and can be removed from column 2 in almost pure form as bottom component. Entrainer residues are removed from the water in column 4 and discharged afterwards. Figure 1 Flow sheet of azeotropic distillation focused on process simulation Implementation of azeotropic distillation in CHEMCAD Components Thermodynamics Feed streams k Unit Operations Ethanol K: NRTL Total flow = 1250 3 SCDS columns Water H: LATE xE = 0.61, xW = 0.39 Multipurpose Flash n-pentane VLLE (global phase) t = 30°C p = 1 bar Azeotropic distillation is a stationary process and can be simulated with CHEMCAD Steady State. To do so, the feed streams and further design parameters are defined in the flow sheet and the simulation is performed afterwards. Once the chemical properties of water, ethanol and n-pentane have been selected from the properties database, the three SCDS columns and a multipurpose flash are placed on the flow sheet and connected. For suitable selection of the feed quantities and the process structure, the thermodynamic properties of the three-substance mixture are evaluated using a residue curve map and a binodal plot. For this purpose, CHEMCAD contains the option [Plot][Binodal/Residue Curves...]. Once the three components have been selected, whereby the first and the third component form the miscibility gap, the previously entered atmospheric pressure is applied and the distillate temperature of the second column set at approximately 32°C as binodal temperature. Now the finished triangular diagram (see fig. 2) is issued. The boiling temperatures of the examined substances are stated in addition to the diagram, together with the azeotropic points with the corresponding composition and the respective boiling temperatures. CHEMCAD requires a suitable thermodynamic model for calculation of the miscibility gap as well as the residue curves, which is why the gE model as well as the VLLE phase characteristics must be set before plotting (both settings are located at [Thermophysical][Thermodynamic Settings]). Figure 2 illustrates the concentration gradient during the distillation process. Separation within the first column takes place along the binary water-ethanol line, whereby the red arrow (arrow 1) represents the rectifying section and the orange arrow (arrow 2) the stripping section. According to this, the two end-productions after the first distillation step are pure water in the bottom and the composition of the binary azeotropic point in the column head1. 1 The first distillation process represents a pre-treatment of the feed mixture and is omitted in case of a sufficiently high ethanol concentration in the feed stream (>75%). focused on process simulation Page 2 of 12 Figure 2 Residue curves/binodal plot In the next step, the bottom component of column 1 is fed into column 2 together with the entrainer n-pentane. In this process, a mixture compound forms in proximity to the ternary azeotropic point (green arrow tip, arrow 3). In the downstream flash, the condensed distillate stream then disintegrates into two liquid phases along the conode, due to the miscibility gap (blue arrow tips, arrow 4). The organic phase (arrow 4a) is added again to the second column, which results in a concentration shift of the ternary mixture across the distillation boundary (see [II] De Filliers, French, Koplos; 2002). The distillation boundary progresses along the residue curves, starting at the binary azeotropic points towards the ternary azeotrope. This way, after reflux of the recycle stream of the organic phase, the material system is located in the distillation section limited by the points binary azeotrope ethanol-n-pentane, pure ethanol and ternary azeotrope (see [I] Ulrich; 14 et seq.). Almost dry ethanol can now be extracted from column 2 as the bottom component (violet arrow tip, arrow 5). The wet part of the second phase (arrow 4b) in the flash is cleaned in a third distillation column (see fig. 1, column 4), so that pure water can be extracted as bottom component and a waterethanol-n-pentane mixture as top component. This is then added to column 2 again. focused on process simulation Page 3 of 12 The feed stream is defined before the columns are specified (see fig. 3). The pressure is set to 1 bar and the temperature to 30°C. The feed concentration is selected arbitrarily in the first distillation section (azeotropic point water/ethanol - ternary azeotrope - pure water). The mass flow is set to 250 water and 1000 ethanol, which corresponds to an ethanol concentration of 61 %. The feed composition can be selected with any ethanol content below the azeotropic point waterethanol, due to the pre-treatment. Figure 3 Feed definition The layout of the first column is defined after the feed definition. The feed stage is set to the 15th stage 1 at medium column height. Altogether, 30 stages 2 are simulated. With rigorous column design, numerous additional specification options are available, such as pressure losses, pressures or temperatures within the column (see fig. 4). As the water-ethanol mixture is supposed to be advanced as close as possible to the azeotropic point, the mole fraction in the distillate stream (distillate component mole fraction) is set to 0.9, while the mole fraction in the bottom (bottom component mole fraction) is set to 0.001 (see fig. 5). 1 The optimum feed stage has been calculated using a shortcut column as described by Fenske-Underwood-Gilliland. As this is based on ideal mixtures and constant volatilities, deviations from the real optimum feed stage may occur. 2 With rigorous design, the number of stages can be selected by performing a sensitivity analysis or determining the theoretical number of stages with the McCabe-Thiele diagram (see [III] Stephan, Schaber, Stephan, Mayinger; 460 et seq.). focused on process simulation Page 4 of 12 Figure 4 Design parameters of the first column Figure 5 Design parameters of the first column focused on process simulation Page 5 of 12 Figure 6 Design parameters of the second column The number of stages of column 2 is set to 151 (see fig. 6). The product stream of the first column (stream 2) is fed at stage 82, while the recycle streams 4, 5 and 7 enter the column on the three next higher stages. An ethanol mole fraction of 0.999 is set as vaporizer specification to achieve the product specification. The condenser is specified with a reflux rate of 1, due to the minimum reflux rate of 0.9152. The minimum value has been determined beforehand using a shortcut column. Alternatively, graphic definition of the minimum reflux ratio using the McCabe-Thiele diagram is also possible. 1 With rigorous design, the number of stages can be selected by performing a sensitivity analysis or determining the theoretical number of stages with the McCabe-Thiele diagram (see [III] Stephan, Schaber, Stephan, Mayinger; 460 et seq.). 2 The optimum feed stage as well as the minimum reflux rate have been calculated using a shortcut column as described by Fenske-Underwood-Gilliland. As this is based on ideal mixtures and constant volatilities, deviations from the real optimum feed stage or from the minimum reflux rate may occur. focused on process simulation Page 6 of 12 Figure 7 Design parameters of the second column Figure 8 Design parameters of the Multipurpose Flash Once column 2 has been initialized, the flash mode is set to 1 in the "Multipurpose Flash" at [Flash Mode] (see fig. 8). A low mole vapour fraction is selected in order to completely condense a distillate while preventing convergence problems at the same time. The pressure may remain unspecified as the inlet pressure of the flash is applied. focused on process simulation Page 7 of 12 In column 4, the number of stages is set to 30 and the feed stage centrally to 15 (see fig. 9). Both parameters have been calculated using a shortcut column. Figure 9 Design parameters of the third column Figure 10 Design parameters of the third column focused on process simulation Page 8 of 12 Due to the low n-pentane fraction of the wet phase, a water-mole concentration of 0.999 is selected as vaporizer specification (see fig. 10). At the same time, the distillate is concentrated until an ethanol concentration of 85% is achieved so that the fraction of water recycled in the second column is not too large. Once all "Unit Operations" have been entered, linked and specified, the start values of the cut stream (see fig. 1, stream 3) have to be initialized. The simulation can be started with "run all", and an immediate convergence of the "Unit Operations" as well as of the material streams should be achieved. In case of convergence problems with the recycle stream, it may be necessary to restart the simulation. Cut Stream (stream 3) p = 1 bar T = 33°C ̇ Water = 100 ̇ Ethanol = 1000 ̇ N-Pentane = 20000 focused on process simulation Page 9 of 12 Assessment of the simulation results In order to obtain a simple overview of the material streams during the process, a list of characteristics of selected streams can be displayed in real time at [Format][Add Stream Box] (see fig. 11). Figure 11 Stream box with selected streams The approx. 21.7 ethanol fraction in the global feed can be removed almost dry at product stream 9. The process runs in a bar (absolute) here. The product specification has thus been achieved. The entrainer n-pentane is completely recycled without loss in the simulation model. In real operation, slight entrainer losses must be compensated for through constant addition. In CHEMCAD, this compensation is realized with the unit operation "Controller". focused on process simulation Page 10 of 12 Optimization of azeotropic distillation The energy input and the type of entrainer can be optimized. In order to reduce the energy input, economisers can be used and the purity of the top and bottom components reduced. CHEMCAD provides the option to perform the sensitivity study to realize optimization. In addition, selection of a suitable entrainer can influence the energy expenditure and the entrainer quantity. When using cyclohexane, for example, less entrainer is required due to the higher ethanol fraction in the tertiary azeotrope. However, as the boiling temperatures of the tertiary azeotropes as well as the impacts of the different organic substances on the environment vary, a suitable compromise must be found in practice. Benzol, for example, is no longer used due to its high degree of toxicity despite its good entrainer characteristics. Usually, the activity coefficients at infinite dilution are determined first when selecting a suitable entrainer, and the effects on the relative volatility of the substances to be separated are investigated (see [IV] Gmehling, Kolbe, Kleiber, Rarey; 2012; 512 pp.). The simulation discussed in this document was generated in CHEMCAD 6.5.3 and can be used with all versions as of CHEMCAD 5. Are you interested in further tutorials, seminars or other solutions with CHEMCAD? Then please contact us: Mail: support@chemstations.eu Phone: +49 (0)30 20 200 600 www.chemstations.eu Authors: Daniel Seidl Meik Wusterhausen Armin Fricke Sources focused on process simulation Page 11 of 12 I. Ulrich, Jan "Operation and Control of Azeotropic Distillation Column Sequences" Diss. ETH No. 14890, Swiss Federal Institute of Technology, Zurich, 2002 II. De Villiers, French, Koplos "Navigate Phase Equilibria via Residue Curve Maps" 2002, http://people.clarkson.edu/~wwilcox/Design/rescurve.pdf (Accessed on 10.09.2013) III. Stephan, Schaber, Stephan, Mayinger "Thermodynamik" Band 2 Mehrstoffsysteme und chemische Reaktionen, 15th edition, Springer IV. Gmehling, Kolbe, Kleiber, Rarey "Chemical Thermodynamics for Process Simulation" 2012, Wiley-VCH focused on process simulation Page 12 of 12