You are on page 1of 5

Journal of Food Engineering 51 (2002) 2731

www.elsevier.com/locate/jfoodeng

Computer simulation of a continuous whisky still


M. Gaiser b, G.M. Bell b, A.W. Lim b, N.A. Roberts b, D.B.F. Faraday R.A. Schulz b, R. Grob a
b

b,*

a Aspentech Europe SA/NV, Rue Colonel Bourg 127-129, B-1140 Brussels, Belgium Department of Chemical and Process Engineering, School of Engineering in the Environment, University of Surrey, Guildford, Surrey GU2 7XH, UK

Received 29 September 2000; accepted 15 January 2001

Abstract Grain whisky is typically manufactured in a continuous distillation unit comprising of two columns, the beer stripper and the rectier. For the functional design of a respective sample still, Aspen Plus , an engineering software package developed by AspenTeche, was used. Based on the literature data, the process parameters were manipulated to meet known characteristics of the production process and the spirit product. The results of the simulation were in good agreement with the data found in the literature. 2001 Elsevier Science Ltd. All rights reserved.
Keywords: Whisky; Distillation; Computer simulation; Ethanol

1. Introduction Whisky manufacture, irrespective of the place of origin, results in the production of two distinctive products. The heavy and full-bodied single malt whisky distilled in traditional pot stills, and the lighter grain whisky, typically produced in continuous distillation units according to the patent or ``Coey'' still design (Lyons & Rose, 1977; Varnam & Sutherland, 1994). The raw material for both processes is a grain-based fermentation wash that could be characterised as unhopped beer. Traditionally, barley is the grain used for most Scotch whiskies and Irish whiskeys (spelled with an ``e'' as opposed to Scotch whisky). Single malt whisk(e)y is produced exclusively (in Scotland) or primarily (in Ireland) from malted barley. Grain whisky is usually produced from grains supplemented with a certain malt fraction in order to provide the enzymes necessary for starch saccharication. While the traditional pot still simply consists of two or three consecutive single-step distillation units, the patent still is built in the form of two multi-stage distillation columns. The patent still was patented in Dublin by Aeneas Coey in 1830. Its principal design remains virtually unchanged since then, only being upCorresponding author. Tel.: +44-1483-879474; fax: +44-1483876581. E-mail address: d.faraday@surrey.ac.uk (D.B.F. Faraday).
*

dated by improvements in process control (Varnam & Sutherland, 1994). One column, the beer stripper or analyser, is used to strip the feed from most of its ethanol. The consecutive rectier concentrates the ethanol and puries it to some extent. A special characteristic of the Coey still is the beer ow in a separate pipe cascading through the rectier (see Fig. 1). This serves two purposes, preheating of the beer before feeding into the top of the beer stripper, and increasing the condensate and thus the separation eect in the rectier. Typical number of trays is 1527 in the beer stripper and 3045 in the rectier. However, this is subject to individual construction. Generally, steam is injected directly into the bottom of the analyser. The rectier is heated by feeding the vapour from the top of the analyser into the bottom of that column. Liquid from the rectier's bottom has to be pumped into the top of the beer stripper. The ethanol stripped medium, the so-called spent wash, is withdrawn as bottom fraction of the beer stripper. Since tax is calculated from the potential amount of ethanol that could be gained in the distillery rather than the amount of alcohol actually collected as spirit, the distiller has to aim to keep the alcohol concentration in that part of the still as low as possible. From the rectier three separate fractions can be drawn. Low boiling components such as aldehydes, acids and esters can be drawn as head fraction at the top condenser of the column. The product is typically drawn at 510 stages below the top condenser at a concentration of

0260-8774/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved. PII: S 0 2 6 0 - 8 7 7 4 ( 0 1 ) 0 0 0 3 3 - 4

28

M. Gaiser et al. / Journal of Food Engineering 51 (2002) 2731

Fig. 1. Process ow diagram of a typical coey still the number of heating coils displayed does not necessarily represent the actual number.

approximately 93.5% v/v (Panek & Boucher, 1989). Between the centre and the bottom of the rectier the higher boiling substances, the fusel-oils are removed (Panek & Boucher, 1989). Fusel-oils and light draws can be collected separately or can be recycled into the beer stripper.

2. Material and methods The patent still is a comparably well-described process for manufacturing grain whisky. Because of its widespread application it is well investigated in the literature. However, depending on process parameters such as point of product draw, reux rate and number of stages, variations in the result of the application of this process are unavoidable. For dening the inputs and validating the outputs of the computer simulation, the presented literature data have been used (Panek & Boucher, 1989; Pyke, 1965; Whitby, 1992). Pyke (1965) described the operation of a patent still along with concentration proles in the rectier column and a wash composition. Other references give at least an idea of the ethanol concentration at the dierent side draw stages (Panek & Boucher, 1989; Whitby, 1992). However, the wash composition described by Pyke (1965) appears to be the result of fermentation suering from bacterial contamination thus resulting in an excess level of lactic acid and a comparably low ethanol concentration. After using the data given by Pyke (1965), a second simulation was calculated, based on the composition of typical beers as an approximation of the feedstock for a whisky still. Modern computer-based design tools provide the engineer with sophisticated means for designing and

optimising a process such as the one described. However, since whisky manufacture is a traditional process, traditional empirical design knowledge is still the preferred option. For this publication, Aspen Plus by AspenTeche (version 10.1) is used to simulate the steady-state operation of a patent still. The system platform of choice was Microsoft Windows 98. Former versions of Aspen Plus are also available for Windows NTe, OpenVMSe and Unix platforms such as Digital Unixe, Solaris2e, HP-UXe and AIXe. Based on the literature information, the crude design of the still was congured and the beer feed dened. For determining the required property data of the substances considered, standard AspenTeche databases on the thermo-physical data included in Aspen Plus were used. The model chosen to describe the vapourliquid equilibria concerned was NRTL-2, the second version of the built-in Non-Random Two Liquids equations, which gives a slightly better approximation of the ethanolwater azeotrope as known from the literature. Aspen Plus has built-in models for typical unit operations such as rectifying distillation (RADFRAC) and heat exchange (e.g. HEATX). Results can either be displayed as tables or in the form of orthogonal graphs. Through typical Microsoft Windows functions, such as copy and paste, results can be exported to other applications such as spreadsheets, word processors or graphic tools. For determination of the required reux rate and the steam rate the Aspen Plus tools for sensitivity analysis and design specications were used. Sensitivity analysis produces a table to show (graphically or by direct comparison) what impact input parameter changes in a given range will have on certain process results. Through design specications Aspen Plus can be used to optimise certain parameters in a given range in order to achieve pre-set targets. The tools were used to optimise the required steam input and to ensure ethanol concentration at the product draw to be at a given level. One characteristic of Aspen Plus is that it counts the stages in a distillation column from top to bottom, the top condenser being stage number 1 and the reboiler being the nal stage. In a conguration without reboiler or top condenser, these stage numbers are assigned to the nearest trays. RADFRAC oers the simulation of fractionating distillation with any number of side draws and pump-arounds. Heat losses, vaporisation or Murphree eciencies and pressure drops can be assigned for the entire column, or, for increased precision, for any smaller part of it. 3. Results and discussion In each simulation the distillation plant was represented by a single column, comprising both the analyser and the rectier, with direct steam injection into the

M. Gaiser et al. / Journal of Food Engineering 51 (2002) 2731

29

bottom. To reect the beer passing through an inner tube inside the rectier for pre-heating of the feed, a heat loss was congured distributed over the rectier stages. The pre-heating was realised by conguring an external heat exchanger (model HEATX) with the respective heat duty. The heat loss of the rectier is split into two sections. Based on the literature information (Pyke, 1965) heat stream for the rst sector was congured to obtain a beer temperature of 48C at the product draw stage. The beer stream was supposed to get heated up to an exit temperature of 90C passing through beer coils in the rectier. Based on this assumption the heat loss of the remaining rectier stages, below the product draw, was congured. The module RADFRAC was used for describing the distillation column. The column's reboiler was eliminated to make the steam injection the only energy source. The two liquid phases developing in the rectier column stages with high iso-amyl alcohol concentrations (Whitby, 1992) were taken into account by the models through the two liquid phases option in RADFRAC. The respective owsheet is shown in Fig. 2. For developing the process model the parameters steam input, product draw, fusel draw and the ratio lights draw/beer feed were varied. First, simulations for simple ethanol/water mixtures were developed. These simplied models were then altered to take into account characteristic congener substances. After including the considered congener substances in the model tray-sizing and tray-rating functions in AspenPlus were used to obtain a rst design of the actual distillation columns. The column was sized and rated in two segments, the analyser and the rectier. In tray rating, the distillation column trays were congured, according to literature information (Whitby, 1992), as sieve trays with a hole diameter of 1/2 in. in the beer stripper, and 4.4 mm in the rectier part. Using the resulting pressure drop prole the respective model was re-run in order to update the results. 3.1. First simulation data according to Pyke (1965) The input for the rst simulation was congured according to Pyke (1965), even though the concentrations given might not reect a typical grain whisky wash. They are detailed in Table 1.

Table 1 Wash composition for rst simulation run Pyke (1965) Substance Ethanol n-Propanol iso-Butanol iso-Amyl alcohol Acetaldehyde Concentration 4.8% v/v 3.4 g/100l 5.0 g/100l 13.5 g/100l 0.78 g/100l

For simplicity, the other given substances were not considered. However, judging from the unusually low pH given for the wash (pH 3.7), it becomes evident that the concentration of acid-related substances might be excessive. The plant was congured to have 15 tray stages in the beer stripper and, according to the reference, 30 distillation stages in the rectier. The Murphree eciencies of the distillation stages were assumed to be at 0.9 in the rectier and 0.8 in the beer stripper, respectively. At rst, the product draw stage was set at stage 7 and the fusel-oil draw at stage 20. The fusels are separated in a decanter into two liquid phases, the fusel-oil and a return stream. Using these settings the reux rate and the steam input were examined for their impact on the alcohol concentration in the product draw. The design specications were set to achieve an ethanol concentration in the product draw of 93.5% v/v. The simulation considering only the ethanol and the water content in the beer feed revealed that at the given alcohol concentration in the beer, the set ethanol concentration in the product draw could not be achieved. All following calculations were thus performed based on an average ethanol concentration of 7.5% v/v and the respective concentrations of the other substances considered based on the composition of typical high gravity beer as found in the literature (Hough, Briggs, Stevens, & Young, 1997). From the concentration ranges given for the substances considered the concentrations were chosen as displayed in Table 2. Starting with a higher ethanol content, the desired product concentration could easily be obtained. After including the congener substances considered, the parameters of the simulation were adjusted in order to meet the typical concentration proles.
Table 2 Wash composition for further simulation Substance Ethanol n-Propanol iso-Butanol iso-Amyl alcohol Acetaldehyde Concentration/% v/v 7.5 0.005 0.0083 0.0142 0.001

Fig. 2. Flowsheet for coey still simulation according to Pyke.

30

M. Gaiser et al. / Journal of Food Engineering 51 (2002) 2731

Fig. 3. Concentration prole in liquid phase, rst simulation stage numbers from top to bottom for rectier and analyser.

As a result, the positions of the concentration peaks of n-Propanol, iso-Butanol and iso-Amyl alcohol in the rectier column were as described in the reference. The absolute concentrations, especially those of iso-Butanol, were slightly dierent, which is mainly due to potential dierences in the feed concentration. The concentration range, however, is similar. The calculated concentration prole is shown in Fig. 3. The major dierence between the prole shown and the one presented by Pyke (1965) is the concentration of aldehydes. In Pyke (1965) graph their concentration prole does not show a signicant peak throughout the rectier column, whereas this congener's concentration is calculated to be enriched in the head fraction. However, judging from proles published in similar papers dealing with dierent types of spirit (Packowski, 1963; Murtagh, 1995), aldehydes are usually expected to be enriched in the head fraction, as calculated in the simulation. In addition, the prole by Pyke shows the ethanol concentration in the rectier bottom at a level near zero, while it has to be assumed that it should have at least beer strength. Thus, the prole published (Pyke, 1965) should be used rather as a direction than as an absolute reference. A comparison between the composition of the freshly distilled raw whisky as given in the reference (Pyke, 1965) and the composition of the product stream as calculated by the computer simulation shows a very similar prole. The concentrations of n-propanol and iso-butanol were calculated to be 611 and 506 ppm as compared to the concentrations given at 500 and 760 ppm, respectively. 3.2. Second simulation data by Whitby (1992) Following Whitby (1992) publication, the owsheet was slightly changed. According to the continuous process outlined in that paper, the decanter was replaced

by a fusel-oil column. This unit was set up as a simple distillation column. The head condenser of the main distillation column was congured as a partial condenser with a vent condenser. Any liquid condensed in the vent condenser is pumped, together with the heads of the fusel-oil column, back into the beer stripping part of the main process. Again, for simplicity, the two columns are represented by one single column. According to the data given, the rectier part was set up having 35 stages and the beer stripper with 27. Spirit is drawn at stage number 8 (from the top), fusels are drawn from stage number 28. Using the assumed concentration for the beer feed as displayed in Table 2 the simulation was easily set up to reach an alcohol concentration of 94% v/v in the product. Again, after including the congener substances the model parameters had to be adjusted in order to meet the concentration proles as given in the publication. The concentration prole obtained is displayed in Fig. 4. As in the previous simulation, the peak position of the respective congener concentrations in the rectier part of the distillation column is comparable to the position as stated in the reference paper. All the concentrations of the congeners given, the iso-amyl alcohol, the n-propanol and the iso-butanol, are similar to the concentrations as stated by Whitby. Obviously, since no information was available on the concentration prole of the beer used in the distillation presented in the publication, there are a variety of parameters present that could be adjusted in the simulation in order to alter these concentrations slightly. The typical concentration range for n-propanol, isobutanol and iso-amyl alcohol in spirits is given by Whitby in the region 0.120.17% w/w, 0.040.07% w/w and 0.0002% w/w, respectively. The product's concentration prole as it results from the simulation with 744 ppm of n-propanol, 267 ppm of iso-butanol and 4 ppb

Fig. 4. Concentration prole for second computer simulation stage numbers from top to bottom for rectier and analyser parts together.

M. Gaiser et al. / Journal of Food Engineering 51 (2002) 2731

31

of iso-amyl alcohol shows slightly lower concentrations for n-propanol, but is overall in good agreement with the concentrations given. 4. Conclusions Aspen Plus is a suitable tool for the design of distillation processes utilised in the potable spirit industry. The built-in database on thermo-physical properties includes most of the spirit components of interest. The result of any simulation is obviously dependent on the availability and precision of input data such as raw material composition, geometric design targets and heat requirements. Considering the typical variation in concentration proles of both feedstock and nal product the results of the simulations are in good agreement with the data found in the literature. In the second simulation, a very good agreement between the simulation and the data given in the reference (Whitby, 1992) was obtained, despite the limited input data. Acknowledgements The project, which this research is a result of, is partfunded by the European Commission under the title ``Proof of Concept in the Development of a Clean Multi-purpose Distillation Technology for Potable Spirit Production''. The project comprises two research parties, the University of Surrey, Guildford, UK, and

the Dublin Institute of Technology, Ireland. The industrial consortium carrying the project comprises of the Welsh Whisky Company, Hirwaun, UK, Kiteen Pontikka, Pioj arvi, Finland, AspenTech Europe SA/ NV, Belgium, Brussels, Carbery Milkproducts, Ballineen, Ireland, Distilleria Andrea Da Ponte, Conegliano, Italy, Paar Scientic Ltd., Kingston, UK, and Graham Hart Engineering, Bradford, UK.

References
Hough, J. S., Briggs, D. E., Stevens, R., & Young, T. W. (1997). Constituents of beer: (b) Volatile. Malting and Brewing Science 2 (pp. 790797). London: Chapman & Hall (reprint). Lyons, T. P., & Rose, A. H. (1977). Whisky. In A. M. Rose (Ed.), Alcoholic Beverages (pp. 635692). London: Academic Press (Chap. 10). Murtagh, J. E. (1995). Neutral spirit production with notes on gin and vodka preparation. In T. P. Lyons, J. E. Murtagh, & D. R. Kelsall (Eds.), The alcohol textbook (pp. 193212). Nottingham University Press (Chap. 14). Packowski, G. W. (1963). Alcoholic beverages, distilled. Encyclopaedia of Chemical Technology (3rd ed.). (Vol. 1, pp. 501531). Panek, R. J., & Boucher, A. R. (1989). Continuous distillation. In J. R. Piggott, & R. E. Duncan (Eds.), The science and technology of whiskies (pp. 150173). Longman Scientic Technical. Pyke, M. (1965). The manufacture of scotch grain whisky. Journal of the Institute of Brewing, 71, 209218. Varnam, A. H., & Sutherland, J. P. (1994). Alcoholic beverages: distilled spirits. Beverages technology, chemistry and microbiology (pp. 400447). London: Chapman & Hall. Whitby, B. R. (1992). Traditional distillation in the whisky industry. Fermentation, 5(4), 261267.

You might also like