The separation portion of the design suggested by Turton involves a system of six separation towers: one quench tower, one absorption tower, one liquid-liquid extractor, and three distillation columns. As discussed above, a significant portion of the separation process takes place at low pressure (under high vacuum) to allow relatively low temperature separation.

The feed to the separation section is the product stream from the reactor. This stream moves from the reactor directly to a quench tower where a stream of dilute aqueous acrylic acid quickly cools, or "quenches", it from 310^oC to 40^oC to prevent further oxidation reactions.

Most of the acetic acid, acrylic acid, and water condense during cooling. Most of this bottoms stream (98.5% by volume) is recycled back through the quench tower, while the remainder is sent to the liquid-liquid extraction unit. The O₂, CO₂, N₂, propylene, and some of the water remain in vapor phase and are sent to an off-gas absorber to recover the acrylic acid and acetic acid that have been entrained with this vapor stream. In the off-gas absorber, deionized water is used to recover the acrylic acid and acetic acid and the purge gases are sent to an off-gas incinerator.

A fraction of the quench tower dilute acid bottoms is sent to the liquid-liquid extractor. There, it is contacted with a countercurrent flow of organic solvent in order to preferentially dissolve the acrylic acid and acetic acid from the water. This process is possible because acrylic acid and acetic acid have a higher affinity for the solvent than they do for water, and because the solvent is not very soluble in water. Regardless, the stream leaving the top of the reactor will contain a relatively small amount of water along with solvent, acrylic acid and acetic acid. The bottoms stream will consist mostly of water with a few impurities. In the suggested design, diisopropyl ether is employed as the organic solvent. Though diisopropyl ether works tolerably well in extracting the acids, its normal boiling point of 68 ^oC is so low that expensive refrigerated water must be used in separation units further downstream. If a higher boiling point solvent can be found, then simple cooling tower water, which costs about one-tenth as much as refrigerated water, may be used in solvent recovery.

The bottoms of the acid extractor is sent to a waste water tower to remove enough of the impurities from the water so that it may be sent to a waste water treatment plant. The top stream of the tower consists mostly of diisopropyl ether which is recycled back to the liquid-liquid extractor. The bottoms stream, which is mostly water, is then sent to a waste water treatment plant.

The top stream from the acid extractor is fed to the solvent recovery tower. The water and solvent is taken off the top of the column and the acetic acid and acrylic acid come out the bottom. This stream is fed to the acid purification tower where acetic acid is taken off the top at 95% purity, and acrylic acid comes out the bottom at 99.9% purity.

Also, to increase the accuracy of our model, experimental data was used to calculate the binary parameters to be used in the simulation. Several different property packages were tested, and it was found that using UNIQUAC yielded results most consistent with the base case.

The results yielded by our model of the base case were in close agreement with those provided by Turton. It was then concluded that our modeling techniques would yield accurate results, and, therefore, any results produced by our simulation could justifiably be compared with Those provided in the base case.

Several solvents were considered including: n-heptane, di-n-propyl ether, ethyl acetate, and isopropyl acetate. The ideal solvent would have a boiling point higher than diisopropyl ether but lower than acetic acid. This condition is so that the use of refrigerated water could be avoided. The solvent should also be relatively insoluble in water but have a high affinity for acrylic acid and acetic acid so that it will be successful in extracting the acrylic and acetic acid out of the water. The ideal solvent would also have a relatively low heat of vaporization so as to minimize the duty of the distillation tower reboilers. It is also important that the solvent not form any azeotropes that would hinder separation.

Solvents such as n-heptane, which is almost completely insoluble in water but has an incredibly high affinity for acrylic acid, were also considered. The problem with using n-heptane, however, is that it does not successfully remove the acetic acid from the water. Because we want to sell the acetic acid, it would need to be separated from the water. Because there is only about 6 kmol of acetic acid in 1200 kmol of water, this separation would prove to be very difficult and would require high reflux.

The optimization of the columns using the guidelines discussed above resulted, in most cases, in fewer stages and lower reflux ratios than specified in the base case.

As mentioned earlier, the conditions as specified by heuristics could not be met exactly since the minimum values were obtained by short-cut methods. However, comparison of table 5.3 to table 5.2 shows that the values in the final design did not stray too far from those specified by heuristics.

The reflux ratios seen in table 5.3 resulted in the following heat duties.

Table 5.4: Heat Duties
	Base Case (GJ/hr)		Optimized Case (GJ/hr)
	T-304	T-305	T-304	T-305
Condenser	108.300	2.280	43.915	2.416
Reboiler	101.000	2.230	43.280	2.374

The lower reflux ratios for the solvent recovery column, T-304, resulted in smaller heat duties in the condensers and reboilers. The smaller duties led to smaller required heat transfer areas and smaller steam and cooling water flow rates. This resulted in a lower capital cost and lower operating costs.

The duties for T-305, the acid purification tower, were pretty much the same as specified in the base case. This was not surprising since this tower was not really affected by the change in solvent and was, therefore, performing the same separation in both the base case and the optimized case. The column in the base case, however, had not been optimized so the size of the column could be decreased during optimization, cutting back on the capital cost.

Though the optimal reflux ratio and number of stages mentioned above were obtained from heuristics, it is important to reconfirm their validity. Therefore, the Winn-Underwood-Gilliand method was used to prepare plots relating the number of stages to the reflux ratio required to achieve the desired separation. The following graph is the plot for the solvent recovery tower, T-304.

Figure 5.1

The optimal conditions occur on the graph where the plot begins to curve upward (around 20-23 stages). Operating near the right side of the plot where the curve is very flat is undesirable because it requires a very large reflux ratio. It is more economical to operate with a lower reflux ratio even if it means the column will need more stages. Adding more stages affects the capital cost which is a one-time expenditure. In contrast, the reflux ratio affects the amount of cooling water and stream needed in the reboiler and condenser, so it affects the utilities costs, which must paid throughout the plant’s life. However, operating at a point where the slope of the curve becomes very steep should be avoided because, on this part of the curve, increasing the number of stages results only in a minimal decrease in the reflux ratio.

Referring back to Table 5.3, it can be seen that, using the distillation heuristics as a guide, the solvent recovery column was designed with 21 stages and was run with a reflux ratio of 0.2. This point falls in the range of optimal conditions identified in Figure 5.1.

The following plot is for T-305, the acid purification tower.

Figure 5.2

It can be seen from this plot that the optimal reflux ratio should be within the range of 14 to 16 and the optimal number of stages should be in the range of 22 to 26 stages. Looking back at Table 5.3 it can be seen that the column was designed with 28 stages and a reflux ratio of 14.3. The reflux ratio falls in the desired range, but the number of stages is a bit higher than the optimal values. The point at which the column had been designed does not even fall on the curve seen in Figure 5.2. This is because this plot was made using short-cut methods and is not very accurate. It is, however, accurate enough to determine approximate optimal values. The reflux ratio of 14.3, as determined by heuristics, falls in the optimal range specified by this curve. The number of stages is a little higher than the optimal values, but it is definitely more desirable to have a higher number of stages than a higher reflux ratio, as discussed earlier.

The optimal reflux ratios and number of stages identified in Figures 5.1 and 5.2 are in agreement with those specified by heuristics. Because the designs of T-304 and T-305, which were based on heuristics, have been confirmed by another method, it is concluded that design based on the heuristic specifications is reliable and that the columns have been designed for optimal performance.

After these two columns had been successfully sized and specified, the last column, the waste water tower, was optimized. The purpose of this last tower is to remove enough impurities from the water so that it can be sent to a wastewater treatment plant. It was found that the tower could be run with zero reflux and with only five trays and still remove enough of the impurities from the water.