Executive Summary

Miller & Associates contracted Kavinetor, Inc. to develop and optimize a separations process simulation for a new acetone production plant. The reactor system for this acetone plant was modeled by Group C. Williams. Using their findings, Kavinetor was charged with the task of using the effluent from their reactor design and isolating acetone at a 99.9% purity. When achieving this purity, this model had to be technically sound and cost-effective.

Kavinetor utilized rigorous techniques, which involved a balance between design and economic considerations. Through the analysis of different column configurations and the resulting economic costs, an optimal design was found. This model met design specifications and economic feasibility. The optimized design was found to have a capital cost of $972,500, as compared to the preliminary design cost of $1,078,000. The utility cost of the optimized design was calculated to be $2,450,000, as compared to the preliminary design cost of $4,021,000. This resulted in an overall savings of 33% over a 20-year plant life.

Based on these results, Kavinetor recommends that this design should be implemented or undergo further analysis. Kavinetor has confidence in the design and economic analysis of this optimized simulation.

Table of Contents

• Introduction
• Thermodynamic Considerations
• Effectiveness of Required Separations
• The Base Case
• Modeling Considerations
• Optimization Methodology
• The Optimized Design
• Design Details
  • Flash Tank
  • Packed Absorber
  • Waste Water Separation Column
  • Acetone Purifying Column
• Economic Analysis of the Optimized Design
  • Capital Cost
  • Utility Cost
• Conclusions
• References

Introduction

 Kavinetor, Inc was contracted by Miller & Associates to perform project work on their acetone production plant, in conjunction with Group C. Williams. Group C. Williams had developed an optimized design for the reactor system of the acetone plant which produces acetone from the dehydrogenation of isopropyl alcohol. The purpose of Kavinetor’s project was to design a separations system to purify the product of this acetone production plant. The goal was to develop a cost-effective model that will provide an acetone product of 99.9 mol% purity and maintain the azeotropic recycle stream. The original model used by Miller & Associates was based on the model discussed in Turton. This report describes the methods that Kavinetor, Inc. used to optimize this preliminary process design and provides the corresponding results.

The Turton model produces acetone through the dehydrogenation of isopropyl alcohol (IPA). The process begins with an azeotropic mixture of IPA and water. This is mixed with a recycle stream of unreacted IPA and water, which is fed into the reactor system. The reactor effluent becomes the feed to the separation system, the area of the plant under consideration. The purification system consists of four separations units. The reactor effluent enters the system through a two-stage cooling system of two heat exchangers in series. The base case begins at the entrance of the effluent into the second heat exchanger. After cooling to an appropriate temperature, the stream is sent to a flash drum, where the vapor is separated from the liquid. The overhead, consisting mainly of hydrogen, is sent to a scrubber. The purpose of the scrubber is to remove hydrogen from the process stream. It uses process water from another part of the unit to scrub acetone and IPA from the hydrogen. The hydrogen exits the scrubber through the overhead. The bottoms of the scrubber combine with the bottoms from the flash drum and enter the first distillation column. The first column separates the acetone from IPA and water. The overhead contains the 99.9% pure product. The bottoms of this column is then sent to the second column, where water is separated from the remaining materials. The overhead contains the azeotropic mixture of IPA and water, along with a small amount of acetone. This stream is recycled to the beginning of the reactor system. The waste water is sent off as the bottoms, which is then sent to the waste water treatment area of the plant. (10)

In simulation design, there is always a certain set of design specifications that must be met. Miller & Associates had agreed with Kavinetor, Inc that the following design specifications had to be used to develop the model:

1) The acetone product stream had to be 99.9% pure on a molar basis.

2) No more than 2.5 kmol/hr of acetone could leave as waste.

3) An azeotropic IPA-water mixture had to be maintained within the recycle stream with 88 wt% IPA.

Thermodynamic Considerations

ASPEN PLUS was chosen as the simulation package due to its thermodynamic capabilities. The feed coming into the separations process consists of acetone, water, isopropyl alcohol, and hydrogen. The first three of these components listed are polar, resulting in a non-ideal chemical system. The thermodynamics property package chosen to simulate this process had to be able to account for this non-ideality. NRTL was chosen for the following simulations due to its ability to handle combinations of polar and non-polar compounds in highly non-ideal chemical systems. However, the default NRTL package in ASPEN did not accurately model our system. One problem with this package is that it calculates the binary coefficients describing the interactions between the various compounds as an average over a wide range of temperatures and pressures. These binary coefficients yielded separations in the columns that did not agree with those specified in the base case. To solve this problem, the data regression method was employed. Experimental binary data for a range of temperatures and pressures similar to the ones of the acetone process was collected and regressed. The regressed parameters for acetone-water (1), acetone-IPA (2), and IPA-water (4) are compared to the experimental data and default parameters in the following graphs. The correct binary coefficients were obtained from this regressed data and were then used for simulation. These coefficients resulted in separations, which were in close agreement with the ones presented in the book.

The Effectiveness of Required Separations

 There are three major separations taking place in this process. One separation is between hydrogen and acetone. This separation can be performed rather easily considering the fact that hydrogen is a non-polar, supercritical gas. These properties of hydrogen make it insoluble in water and therefore easy to separate from the highly soluble acetone. The other separations are more complicated and occur between acetone and IPA and between IPA and water. These components must be separated through distillation. Distillation is the separation of a mixture into two or more products having different boiling points.

When there is a mixture of two components in the liquid and vapor phase, the components will diffuse until the two-phase compositions reach equilibrium. Therefore, several binary diagrams were used to analyze the desired separations (see Figures 1-3). These binary diagrams represent the mole fractions of the liquid and solid compositions of the volatile compound. X is the liquid mole fraction and y is the vapor mole fraction. Each graph has a reference diagonal; the diagonal represents the points where the liquid mole fraction equals the vapor mole fraction. All of the figures show the binary plot for the heavy and light key components for each of the splits found in the process. All the diagrams exhibit favorable separating conditions for the components involved with the exception of IPA and water.

As one can see from Figure 3, the vapor- liquid equilibrium curve crosses the reference diagonal indicating the presence of an azeotrope. The point of intersection represents a critical composition for which the vapor and liquid composition are identical. Once this critical composition is reached, the components cannot be separated at the given pressure. The IPA-water mixture is a minimum boiling azeotrope and boils at a temperature lower than either of the pure components. When distilling a mixture containing an azeotrope, the top product is the azeotrope and the bottom product is the high boiling component. While azeotropes generally make separations more complicated, it is really not a problem in this design since an azeotropic mixture of 88 wt % IPA to water is desired for the recycle stream.

Figure 1 exhibits the VLE interactions between Acetone and Water. This figure demonstrates favorable separating conditions for the two components.

Figure 2 also exhibits a clean separation between acetone and IPA. However, the binary curve is closer to the reference diagonal indicating a lower relative volatility and therefore a separation requiring more trays.

Figure 3 indicates the presence of an azeotrope. For the smaller liquid mole fractions of IPA the curve is further from the reference diagonal indicating a high relative volatility and an easier separation.

The Base Case

The base case was built from the model described in Turton. This model begins with a heat exchanger where the incoming stream is composed of acetone, water, hydrogen and isopropanol. The compositions of the feed stream were determined by Group C. Williams with 38.11% acetone, 3.75% IPA, 20.42% hydrogen, and 37.72% water. This stream is cooled from 40 C to 20 C, via the use of refrigerated water. It is then fed into a flash tank where most of the hydrogen leaves through the top. The top stream is then sent into a packed absorber, which scrubs the acetone and IPA from the hydrogen with process water and uses 1 inch ceramic Rashig rings. The bottom product of this absorber is mixed with the bottom product of the flash tank. This resulting acetone rich stream is then sent into the first separation column, which consists of 66 theoretical stages. For simulation purposes, sieve trays were modeled with the assumption of 85% tray efficiency. This assumption was provided by Miller & Associates. Due to this assumption, the process was modeled with 78 actual stages, with a reflux ratio of 1.76. The feed to the column enters stage 39, which is the middle tray of the column. Acetone leaves the top of the column with a purity of 99.9%. The bottoms of the column are then sent into a second separation column, where it feeds into the column at actual tray 18. This column has 19 theoretical stages, or 22 actual stages, using the same tray efficiencies seen in the first column. This column is operated with a reflux ratio of 30. From this column, waste water leaves as the bottoms. The overhead is composed of an azeotropic mixture of acetone and IPA, with 88% wt IPA. This azeotropic mixture is the recycle stream, which will return to the beginning of the process to be fed into the reactor system.

The process flow diagram of Kavinetor’s preliminary design is as follows :

Modeling Considerations

In optimizing the preliminary acetone separations process, several modeling considerations had to be considered. There were two main types of considerations: design and economic considerations. Both must complement each other in order to produce the most cost-effective and realistic model. In the design considerations, the thermodynamic package and heuristics were important in simulating a reliable and accurate model. In the economic considerations, capital costs and utility costs were the important factors. Economic considerations are directly related to the design. For example, a reduction in the number of stages in a column will result in a smaller column, which reduces capital cost. However, a smaller column will require greater heat duties for the condenser and reboiler, which causes an increase in the utility cost. Therefore, a balance must be made between the design and economics of the model.

Optimization Methodology

One of the first considerations made for the optimization was the ordering of the distillation columns. The flash tank and absorber were not considered for this ordering since they are properly placed at the beginning of the separations train. They are used to separate hydrogen from the process stream. Hydrogen is the first component that must be removed since it does not liquefy easily, therefore the flash tank and absorber should remain at the beginning of the system. Distillation heuristics were used in determining the optimum order of the distillation columns. The following are the main heuristics that were considered: (9)

1) The two components with the lower relative volatility should be separated last.

2) The product requiring the highest purity should be separated last.

Based on these heuristics, it was determined that the ordering of the two distillation columns should be changed. As can be seen in Figure 2, Acetone and IPA have a lower relative volatility indicating that the acetone purifying separation is the hardest. Therefore, the waste water-producing column would be the first column, and the purified acetone becomes a product of the second column. After this ordering was determined, the columns had to be individually optimized.

Optimizing the two columns in the system involved the variation of different parameters to observe their effect on the final cost. For each column, the following variables were changed:

1) number of stages

2) reflux ratio

3) feed stage to the column

Sensitivity tests were run for the three variables in all possible combinations of configuration. Around 4,000 possible variations were calculated in the ASPEN simulation model for these runs. Samples of the converged results of the sensitivity test can be seen in Appendix D. From these variations, the ones providing the desired product specifications were selected. The best combinations were determined for cost comparison. Each configuration was costed according to capital and utility costs (see Appendix C for example spreadsheet). The net present values for each variation were calculated. Then, each variation was plotted with the net present value versus the number of stages. Each stage number represented the ideal feed stage and reflux rate for that specific number of stages. The minimum of this plot was determined to be the optimal design. This procedure was followed for both columns.

The Optimized Design

The main optimization performed was the change made in the order of the separation columns. In this design the waste water comes out of the bottom of the first column and the top is sent to the second column. The second column separates the acetone to the purity of 99.9% as the top stream and creates the azeotropic mixture as the bottom stream. Other more detailed aspects such as number of stages, heat duties and reflux ratios were directly affected and optimized due to the sequence change of the columns.

The new sequence of order can be seen in the optimized process flow diagram as shown in Figure 5.

**Detailed PFD can be seen in appendix along with process streams.

Design Details

The following sections discuss Kavinetor’s optimized design in detail.

Flash Tank

The flash tank was operated at the same conditions as the preliminary model, with a vertical orientation, height of 2.25 m and diameter of 0.75 m. Although the feed entering the flash tank is at a higher pressure (2.35 bar) compared to the model given by Turton (1.63 bar), the operability of the flash tank still remained the same and did not need to be altered.

The flows around the flash tank are seen in Table 1.

 Packed Absorber

The optimized case was chosen to follow a similar design as the base case. This vertical tower is 0.33 m in diameter, with a height of 3.2 m. Inside the column is 2.5 m of packing using 1inch ceramic Rashig rings. Different types of packing were researched to see if the Rashig rings could be replaced by more efficient methods. Rashig rings have very low efficiency, but are very cheap and cost-effective. A representative of Albemarle Corporation recommended Kavinetor to keep the use of Rashig rings (5). Rashig rings are sufficient since this column is so small and is used to remove hydrogen, which is an easy separation and does not require a great deal of efficiency. Therefore, Kavinetor decided to use the Rashig rings for the optimized design.

Table 2 shows the flows around the packed absorber. 

Waste Water Separation Column

The first column in the optimized separations train has waste water coming out of the bottoms and the rest of the components coming out of the overhead. As discussed in the Optimization Methodology section, different parameters had to be considered in order to find the most cost effective model. These parameters included number of stages, reflux ratio, and feed stage to the column. Sensitivity runs were performed by altering these parameters in order to find all combinations that would give the appropriate design specifications. The following design specifications were used: a 97.3 mol % water in the bottoms and a 3.2 mol % water in the overhead. These specifications were estimated by taking into account the necessary composition for the azeotropic mixture. This column’s overhead must contain the 88 wt % IPA to water mixture required by the azeotrope. The various combinations that were found to converge can be seen in Appendix D. Once the appropriate combinations were found, the capital and utility costs were calculated for each and the most cost-effective was chosen. This costing can be seen in the economic analysis portion of the report. The following chart shows the parameters of the final optimized design and those of the base case:

The reflux ratio is greatly reduced from that of the waste water column of the preliminary design and will have a significant effect on utility costs.

This column has a height of 18.9 m and a diameter of 1.3 m. The height of the column can be calculated by multiplying the number of stages by the tray spacing and adding an additional height which can be back-calculated from the base case. The following equation results:

H = (# of stages)*(tray spacing)*(actual height in base case) / (height due to trays)

The diameter of the column was specified instead of being found. Aspen and other rough correlations ( Reboiler Duty = 0.4*[diameter]² ) (6) approximated many of the column diameters to be approximately 0.8 meters. This seemed unreasonable in addition to the fact that CAPCOST would not allow this diameter for the heights involved. Since the size of the columns used in the optimized case fell between the size of those used in the base case, the base case diameters were used as an estimate. In the base case the column diameters ranged from 1.25 meters to 1.36 meters. For the optimized case, a diameter of 1.3 meters was used. It was found through trial and error that varying the diameter of the column did not significantly change our economic analysis over the 20 year plant lifetime. Therefore the 1.3 meters was felt to be an appropriate input for the simulations.

For a column of this size, a pressure drop of .134 atm will occur. The pressure drop across the column was calculated using the correlation of approximately 0.1 psi per tray (7). For our simulation purposes, the column was modeled using sieve trays with an 85% efficiency as in the base case. This was done so that parallel cost analyses could be performed between the base case and optimized case. Since there is such a wide range of efficiencies for each tray type, there is no way of constructing a model to accurately depict these effects. However, when implementing this design two types of trays in particular should be considered: sieve trays and valve trays. Bubble-cap trays were originally considered; however they have become antiquated and are only used when a large liquid holdup is required on each tray. Large liquid holdups are often necessary when a chemical reaction is happening at the same time as the distillation and for other specialized cases. Sieve trays are the cheapest of the two mentioned because they are simple and relatively easy to make. They also have the lowest pressure drop per tray. However, sieve trays have the smallest turndown ratio. Valve trays, while being more expensive and having a larger pressure drop per tray, have a larger turndown ratio. This results in a good operation at lower flow rates (7). Since valve trays are only a little more expensive then sieve trays while providing more flexibility in operations, both trays should be tested before a decision is made.

Condenser—Waste Water Separation Column

Due to the reduction in reflux ratio from 30 to 1.05, less heat duty is required to condense the overhead of the column. This results in a smaller heat exchanger and smaller amounts of utilities necessary for cooling.

This heat exchanger had to be properly sized once the optimum conditions were determined. As stated earlier, the optimum number of stages was found to be 20. At 20 stages, it was found from ASPEN simulation that the required condenser duty was 2465 MJ/hr. The heat transfer area required for this heat exchange was found using the following equation from Bird, Stewart, and Lightfoot (3):

= ( Equation 1

where Q_c is the heat duty of the exchanger, U _o is the heat transfer coefficient, A _o is the total heat transfer area of the exchanger, and (T_h-T_c) _ln is the logarithmic mean temperature. The logarithmic mean temperature is calculated as shown below:

Equation 2

where T_h2 is the outlet temperature of the hot side of the exchanger (process side), T_c2 is the outlet temperature of the cold side (cooling water), T_h1 is the inlet temperature of the hot side, and T_c1 is the inlet temperature of the cold side.

The heat transfer coefficient was found by using equation 1 with the Turton model’s values for duty, area and temperatures. The heat transfer coefficient was found to be 3.08 MJ/m²hr. After the U-value was found, equation 1 was used with Kavinetor’s values of the optimum heat duty and temperatures to solve for the area. The area of this condenser was found to be 33.4 m².

Cooling water is used as the heat transfer medium. To calculate the amount of cooling water needed for this exchange, the following equation was used:

Equation 3

where Qc is the heat duty of the exchanger, m is flow rate of cooling water required and Cp is the average heat capacity of water within the operating temperature range of cooling water (30-40 deg C) and dT is the change in temperature of the cooling water. The flow rate of cooling water was found to be 58.7 tonne/hr.

The following table shows a comparison between the condenser’s heat duties, transfer areas and utilities for the base case and the optimized design:

As can be seen in the table, a smaller heat exchanger and less utilities are required for the optimized design. The benefits of this can be seen in the economics, which is discussed later in the report.

Reboiler—Waste Water Separation Column

The reboiler calculations follow the same procedure that was followed with the condenser. At 20 stages, it was found that the required reboiler duty was 3,160 MJ/hr. A proper size for this amount of duty was determined using equation 1 stated above. The heat transfer coefficient was found by using equation 1 with the Turton model’s values for the duty, area and temperatures for this reboiler. One change made is that the logarithmic mean temperature becomes (Th-Tc). The modified equation 1 becomes:

where T_h is temperature of the process side outlet, and T_c is the temperature of the heating side outlet.

After the heat transfer coefficient was found, equation 4 was used with Kavinetor’s values of the optimum heat duty and temperatures for the 20-staged column to solve for the area. The heat transfer coefficient was found to be 2.23 MJ/m²hr and the area was found to be 23.7 m².

Low pressure steam is used as the heat transfer medium. To calculate amount of low pressure steam needed, the following equation was used:

Equation 5

where Q_c is the heat duty required, m is the steam flow rate and H_v is the heat of vaporization for the operating conditions. H_v was calculated using the heat duty and flow rate values seen in Turton’s model. This heat of vaporization was then used with the optimized model’s heat duty to solve for the flow rate. This mass flow rate was found to be 1.52 tonne/hr.

The following table shows a comparison between the reboiler’s heat duty, transfer areas and utilities for the base case and optimized design:

As can be seen, a smaller exchanger and utilities are needed for the reboiler as well. This reduces the cost needed for capital and utilities. This is shown in the economic analysis.

Acetone Purifying Column

The second column in the optimized separations train has an azeotropic mixture of IPA and water coming out of the bottoms, and the acetone product is coming out of the overhead. The same procedure used to determine the most cost effective column configuration for the waste water separation column is used for the acetone purifying column. Once again, the costing of the various parameters can be seen in the economic analysis portion of the report.

The number of stages is greatly reduced from the acetone purifying column of the preliminary design and will cause a significant decrease in the capital costs. The column will have a height of 39.7 m and a diameter of 1.3 m. For this size column, a pressure drop of 0.282 atm will occur. All of these parameters were determined in the same way as they were for the waste water separating column. As with the first column, both sieve trays and valve trays should be tested before implementation.

Condenser—Acetone Purifying Column

This heat exchanger had to be properly sized once the optimum conditions were determined. All methods of calculation follow that of the waste water separation column. As stated earlier, the optimum number of stages for this column was found to be 42. At 42 stages, it was found from ASPEN simulation that the required condenser duty was 2,902 MJ/hr.

The heat transfer coefficient was found by using equation 1 with the Turton model’s values for duty, area and temperatures. The heat transfer coefficient was found to be 3.08 MJ/m²hr. After the U-value was found, equation 1 was used with Kavinetor’s values for the optimum heat duty and temperatures to solve for the area. The area of this condenser was found to be 43.3 m².

Cooling water is used as the heat transfer medium. The mass flow rate of cooling water was found to be 69.2 tonne/hr.

The following table shows a comparison between the condenser’s heat duties, transfer areas and utilities for the base case and the optimized design:

 The operating temperature for this condenser in the base case is at 83 ^oC. The operating temperature in the optimized design is near 60 ^oC. Since there is a difference in the operating temperatures of this column between the base case and the optimized design, the heat transfer coefficients used in the calculations are different. Therefore, no accurate comparison can be made.

Reboiler—Acetone Purifying Column

The reboiler calculations follow the same procedure that was followed in the waste water separation column. At 42 stages, it was found that the required reboiler duty was 2,925 MJ/hr. A proper size for this amount of duty was determined using equation 4 stated in the waste water separation column section. The heat transfer coefficient was found by using equation 4 with the Turton model’s values for the duty, area and temperatures for this reboiler. Once this was found, equation 4 was used with Kavinetor’s values for the optimum heat duty and temperatures for the 42-staged column to solve for the area. The heat transfer coefficient was found to be 1.62 MJ/m²hr and the area was found to be 24.0 m².

Low pressure steam is used as the heat transfer medium. This mass flow rate was found to be 1.40 tonne/hr.

The following table shows a comparison between the reboiler’s heat duty, transfer areas and utilities for the base case and optimized design:

As can be seen, a smaller exchanger and less utilities are needed for the reboiler. This reduces the cost needed for capital and utilities. This is shown in the economic analysis.

Economic Analysis of the Optimized Design

In this section, the waste water separation column is referred to as "Column 1" and the acetone purifying column is referred to as "Column 2." The two separation columns were optimized through sensitivity tests in ASPEN. The variables of the sensitivity tests were reflux ratio, number of stages, and feed stage. The output combinations that yielded the required acetone purity were selected. The capital costs of the selected results were then found using CAPCOST and the utility costs were calculated as specified in Appendix C. The net present value was calculated over twenty years and then plotted versus the number of stages in order to find the configuration that would result in minimum costs. This configuration was then used in the optimized design. Figures 6 and 8 were constructed for the waste water separation column and the acetone separating column respectively.

As can be seen from Figure 6, minimum costs are incurred with a 20-stage column.

The following figure shows that the optimized operating condition is located at the point on the curve where both the number of stages and the reflux ratio are near a minimum.

This figure supports the results seen in Figure 6, where the optimum stage is around 20 and the reflux ratio is around 1.

The following figure represents the cost-optimization curve for Column 2.

As can be seen from Figure 8, minimum costs are incurred with a 42-stage column.

The following figure shows that the optimized operating condition is located at the point on the curve where both the number of stages and the reflux ratio are near a minimum.

This figure supports the results seen in Figure 8, where the optimum stage is around 42 and the reflux ratio is around 1.9.

Capital Costs

The capital costs of the optimized design were lower then those of the base case. One of the reasons for savings is the reduction of the column sizes due to the use of less overall stages. The optimized columns consisted of 20 and 42 stages as opposed to the 22 and 78 stages found in the base case. The optimized columns also operated at lower reflux ratios than the base case columns. Due to the lower reflux ratios, less duty is needed from the heat exchangers. Thus, capital costs are also lowered by the use of smaller heat exchangers. Total capital costs were lowered from $ 1,078,000 to $927,000. In other words, 14% was saved in capital costs over the base case. For detailed costing of the individual parts of equipment see Appendix B.

Utility Costs

In addition to reducing capital costs indirectly, the smaller duties required for the reboilers and condensers directly translated into savings in utilities, specifically in low pressure steam for the reboiler and cooling water for the condenser.

The utility costs were reduced from $4,021,000 to $2,450,000 by comparing the net present value over 20 years at 10% discount rate. This represents a 39% savings over the base case. For individual calculations of utilities see Appendix B.

Overall, a savings of 33% was achieved. This is an indication that this design should be considered.

Conclusions

In conclusion, Kavinetor, Inc. has developed an optimized model for the separations train of the acetone production plant. This has been based on rigorous economic and design analysis. All design specifications have been met and have been found cost-effective. With an overall savings of 33%, this separations design is recommended for use or further investigation. In conjunction with Group C. Williams, a complete simulation of this new plant has been achieved.

References

1. Arlt, W., Gmehling, J.and Onken, U. Vapor-Liquid Equilibrium DataCollection:Aqueous-Organic Systems (Supplement 1), Vol 1 Part 1a. Dechema, Frankfurt, 1981.

2. Arlt, W., Gmehling, J.and Onken, U. Vapor-Liquid Equilibrium Data Collection Organic Hydroxy Compounds, Alcohols and Phenols, Vol 1 Part 2b. Dechema, Frankfurt, 1978.

3. Bird, R.B., Stewart, W.E., and Lightfoot, E.N. Transport Phenomena. John Wiley and Sons, New York, 1960.

4. Gmehling, J.and Onken, U. Vapor-Liquid Equilibrium Data Collection: Aqueous- Organic Systems, Vol 1 Part 1. Dechema, Frankfurt, 1977.

5. Huang, Y.F, Personal correspondence on November 1, 1999 by phone.

6. Hyprotech Corporation. Hysys Reference Guide 1. 1997.

7. Kister, Henry Z. Distillation Design. McGraw Hill, New York, 1992.

8. Luck, Patrick et. Al. "The Wonderful World of Acetone, Isopropyl Alcohol, Water, and Hydrogen".1998.

9. Seader, J.D. et. Al. Process Design Principles: Synthesis, Analysis, and Evaluation. John Wiley & Sons, New York, 1999.

10. Turton, Bailie, Whiting, and Shaeiwitz. Analysis, Synthesis, and Design of Chemical Processes. Prentice Hall PTR, 1998, p. 728–735.

11. Williams,C. et. al. Acetone Production Via Isopropanol Dehydrogenation: Reactor Design. 1999.