Design and Control of the Styrene Process

发布于:2021-06-11 03:02:54

Ind. Eng. Chem. Res. 2011, 50, 1231–1246


Design and Control of the Styrene Process
William L. Luyben?
Department of Chemical Engineering, Lehigh UniVersity, Bethlehem, PennsylVania 18015, United States

A recent paper by Vasudevan et al. (Ind. Eng. Chem. Res. 2009, 48, 10941-10961) presented a ?owsheet of the styrene process, which has several interesting design and control features. Their study concentrated on a comparison of three plantwide control methodologies. The economic optimum design of the process was given, but no quantitative details were given of how the optimum was obtained. The chemistry of styrene process involves the dehydrogenation of ethylbenzene. The reaction is endothermic, nonequimolar and reversible, so high temperatures and low pressures are conducive to high conversion in the adiabatic vaporphase reactors. Steam is mixed with the ethylbenzene (EB) fed to the reactors to lower the partial pressure of ethylbenzene and increase conversion. There are also several other side reactions that produce undesirable byproducts (benzene, toluene, ethylene and carbon dioxide), whose reaction rates increase with temperature and partial pressures. The main design optimization variables in this process are the steam-to-EB ratio, reactor inlet temperature, EB recycle ?ow rate and reactor size. Low reactor temperatures suppress side reactions but require higher EB recycle to achieve the same styrene production rate, which increases separation costs. Higher steam-to-EB ratios also suppress side reactions but increase furnace fuel costs and steam supply costs. The purpose of this paper is to develop a reasonable conceptural design considering capital costs, energy costs and raw material costs and then to develop a plantwide control structure capable of effectively handling large disturbances in production rate. The proposed design is signi?cantly different than the design of Vasudevan et al., featuring higher steam-to-EB ratios, lower reactor temperatures, larger EB recycle ?ow rates and larger reactors. Styrene yield is improved from 76 to 87%, which results in a 10% reduction in operating costs.
1. Introduction Jim Douglas was one of the pioneers in conceptual process design.1 His early book and papers led the way for much of the future practical, nonmathematical design work. One of his contributions was to point out the dominating economic effects of raw material costs in the design of a process. The fresh feed streams are much more costly than the energy used in the process or the cost of the capital invested to build the equipment. Reactant costs are typically an order of magnitude larger than energy or capital costs. This effect is completely generic (“Douglas Doctrine”) and is of vital importance in process design. The styrene process provides an excellent example of the application of the Douglas Doctrine. The consumption of ethylbenzene fresh feed can be reduced by two methods: 1. Increasing the process steam that is added to the EB feed to the reactor lowers partial pressures and helps to increase conversion of styrene and decrease the production of undesirable byproducts. However, it increases the cost of supplying the process steam, and it increases the cost of fuel needed in the furnace to heat the steam and EB to the desired reactor inlet temperature. 2. Lowering reactor temperatures lowers the production of undesirable side products that consume EB without producing styrene. However, lower reactor temperatures require higher EB recycle ?ow rates to raise EB concentrations so the desired production rate of styrene can be achieved. This increases the capital and energy costs of the separation section of the process. Thus there are signi?cant trade-offs that must be made to determine the optimum economic design.
E-mail: Phone: 610-758-4256. Fax: 610-7585057.

Several authors3,4 have considered the styrene process for evaluating sophisticated and systematic multiobjective rigorous optimization. The emphasis in these studies is on the optimization algorithm and not on providing process insight about the effects of the important design optimization variables. A more practical engineering approach is taken in this paper. In this study the production rate of styrene is ?xed for all cases, and the effects of other design optimization variables are quantitatively explored. The expenses associated with producing this ?xed amount of styrene include raw material (EB) cost, process steam cost, furnace fuel cost, column reboiler energy costs and capital costs (reactors, furnaces, heat exchangers, decanter, and distillation columns). As we will demonstrate, the raw material cost dominates, so investing capital and using energy to reduce the consumption of EB is usually justi?ed, up to the point of diminishing returns. Typically, more capital must be invested to reduce expenses of energy, steam, and raw material. The economic objective is to obtain a reasonable incremental return on incremental investment as changes are made in the design optimization variables. 2. Reaction Kinetics and Phase Equilibrium Figure 1 shows the styrene process described by Vasudevan et al.2 The details of the ?owsheet are discussed in the next section. Peng-Robinson physical properties are used in the Aspen simulations. 2.1. Reaction Kinetics. The production of styrene involves the dehydrogenation of ethylbenzene in a high-temperature, lowpressure gas-phase adiabatic reactor. The reaction is reversible and endothermic. Styrene Reaction: C6H5CH2CH3 S C6H5CHCH2 + H2


10.1021/ie100023s ? 2011 American Chemical Society Published on Web 10/14/2010


Ind. Eng. Chem. Res., Vol. 50, No. 3, 2011

Figure 1. Vasudevan et al. design.

There are several other side reactions given by Vasudevan et al.2 that consume ethylbenzene and produce undesirable byproducts. Benzene/Ethylene Reaction: C6H5CH2CH3 f C6H6 + C2H4 Toluene/Methane Reaction: C6H5CH2CH3 + H2 f C6H5CH3 + CH4 Carbon Monoxide Reactions: 2H2O + C2H4 f 2CO + 4H2 H2O + CH4 f CO + 3H2 Carbon Dioxide Reaction: H2O + CO f CO2 + H2

Table 1. Reaction Kineticsa k reaction reaction reaction reaction reaction reaction reaction

E (kJ/kmol) concentration (Pascals) 90 981 61 127 207 989 91 515 103 997 6723 73 638 PEB PSPH PEB PEBPH (PW)2PE PWPM PWPCO



1 forward 0.044 1 reverse 6 × 10-8 2 27,100 3 6.484 × 10-7 4 4.487 × 10-7 5 2.564 × 10-6 6 1779

Overall reaction rates have units of kmol s-1 m-3.

(4) (5)

R2 ) pEBk2e-E2/RT R3 ) pEBpHk3e-E3/RT R4 ) pW(pE)0.5k4e-E4/RT R5 ) pWpMk5e-E5/RT R6 ) pWpCOk6e-E6/RT

(8) (9) (10) (11) (12)


Table 1 gives the kinetic parameters used in this study, which are somewhat modi?ed from those of Vasudevan et al.2 to match the reactor exit stream leaving the second reactor. Simple power law kinetics are used for all reactions. All overall reaction rates have units of kmol s-1 m-3. Concentration units are partial pressures in Pascals. R1F ) pEBk1Fe-E1F/RT R1R ) pSpWk1Re-E1R/RT


The chemistry immediately tells us that high temperatures will favor the ?rst reaction because the activation energy of the forward reaction is greater than the activation energy of the reverse reaction since the reaction is endothermic. The kinetics also tell us low pressure and high ethylbenzene concentrations will favor the production of styrene.

Ind. Eng. Chem. Res., Vol. 50, No. 3, 2011


Figure 2. (A) Txy diagram for EB/styrene at 10 kPa. (B) Txy diagram for EB/styrene at 50 kPa.

However, the other reactions have speci?c reaction rates that increase with temperature. So the production of undesirable byproduct that waste the reactant ethylbenzene can be reduced by keeping the temperature low. Thus there is a fundamental con?ict between conversion (favored by high temperatures) and selectivity (favored by low temperatures). Low pressure also slows these undesirable reactions. 2.2. Phase Equilibrium. The styrene process has a decanter, which involves liquid-liquid-vapor equilibrium, and two distillation columns, which involve vapor-liquid equilibrium. Parts A and B of Figure 2 give the Txy diagrams for the EB/ styrene system at 10 and 50 kPa, which are the pressures at the top and bottom of the product column C1. The separation is run under vacuum to suppress styrene polymerization. It is a dif?cult separation, so the column has many trays and a fairly high re?ux ratio. Figure 3 gives the Txy diagram for the toluene/ EB separation at 120 kPa. The separation is only modestly dif?cult so fewer trays and a lower re?ux ratio are required.

In the decanter, a multicomponent mixture is separated into two liquid phases (aqueous and organic) and a vapor phase. The major components are water, EB, styrene, and hydrogen with small amounts of benzene and toluene. Figure 4 gives a ternary diagram for the EB/styrene/water system at 10 kPa. The total pressure in the decanter is 120 kPa, but the presence of hydrogen and other light components lowers the effective pressure of the other heavier components. A pressure of 10 kPa gives a temperature close to the 40 °C temperature in the decanter and illustrates the LLE properties. The aqueous phase is essentially pure water. The organic phase contains a small amount of water. 3. Vasudevan et al. Flowsheet Figure 1 shows the ?owsheet of the process with the equipment sizes and conditions derived from the paper by Vasudevan et al.2 Flow rates and compositions are similar to those given in their paper, but there are some small differences.


Ind. Eng. Chem. Res., Vol. 50, No. 3, 2011

Figure 3. Txy diagram for toluene/EB at 120 kPa.

Figure 4. Ternary diagram for EB/styrene/water at 10 kPa.

The fresh feed is 152.6 kmol/h of ethylbenzene. It is combined with 74.6 kmol/h of a recycle stream of mostly ethylbenzene and 612 kmol/h of low-pressure process steam. The stream is heated in a feed-ef?uent heat exchanger (E2), which uses the hot reactor ef?uent to heat the feed stream to 487 °C. Additional low-pressure process steam (2788 kmol/h) is heated in a furnace (E1) to 777 °C and mixed with the stream from E2 to achieve a reactor inlet temperature of 650 °C. The heat duty in the furnace is 16.9 MW. The total process steam is 3400 kmol/h. 3.1. Reactors. There are two gas-phase adiabatic reactors in series. Each has a diameter of 3.3 m, a length of 3.5 m, and a catalyst loading of 35 700 kg. The exit temperature of the ?rst reactor is 588 °C because of the endothermic reaction. A furnace (E3) heats this stream back to 650 °C before it enters the second reactor. The furnace heat duty is 3.5 MW.

The total EB entering the ?rst reactor R1 is 225.1 kmol/h. The ethylbenzene leaving the second reactor has a molar ?ow rate of 80.9 kmol/h, so the per-pass conversion of EB in the two reactors is 64%. The molar ?ow rate of styrene leaving the second reactor is 119 kmol/h. If all of the ethylbenzene in the fresh feed were converted into styrene, there would be 152.6 kmol/h of styrene produced. Therefore, a signi?cant amount of EB is wasted in producing byproducts. In addition, there are signi?cant losses of EB and styrene in the two gas streams leaving the process (to be discussed in the next section). The production rate of stryrene leaving the process from the bottom of the ?rst distillation column is only115.6 kmol/h, giving a styrene yield of only 76%. It is obvious that a lot of ethylbenzene is not ending up in the styrene product stream. This means that raw material cost is

Ind. Eng. Chem. Res., Vol. 50, No. 3, 2011


higher than it would be if smaller amounts of byproducts were produced and losses of EB and styrene were lower. There appears to be plenty of opportunity for process improvement. 3.2. Condenser and Decanter. After being cooled in E2 and partially condensed in heat exchanger E4 using cooling water, the process stream enters a decanter operating at 40 °C and 120 kPa. The dimensions to give 20 min of liquid holdup when half full is a diameter of 2.8 m and a length of 5.6 m. Water is withdrawn from the bottom of the vessel at a rate of 3355 kmol/ h. The gas phase (“Lights”) leaving the top of the decanter at a ?ow rate of 194.1 kmol/h contains most of the hydrogen formed in the basic reaction, but it also contains signi?cant amounts of other components. The styrene lost in this stream is 1.71 kmol/h, and the EB loss is also 1.71 kmol/h. 3.3. Product Column C1. The organic phase is fed to the ?rst distillation column C1 at a ?ow rate of 233.6 kmol/h and a composition of 33.83 mol % EB, 49.78 mol % styrene, 6.08 mol % benzene, 4.71 mol % toluene, 5.29 mol % water, and smaller amount of some of the light components. The 82-stage column operates under vacuum with a re?ux-drum pressure of 10 kPa. The pressure drop per tray is assumed to be 0.5 kPa. A theoretical-tray model is used in the Aspen RadFrac simulation, but in determining the capital cost of the column, structured packing at a cost of $1000 m-3 is assumed to ?ll the vessel. The diameter of the column is 4.86 m, and the reboiler duty (supplied by low-pressure steam) is 8.31 MW. The re?ux ratio is 4.76. The feed is fed on Stage 37, which minimizes reboiler duty. The two design speci?cations for the column are a bottoms purity of 99.75 mol % styrene and a distillate impurity of 1 mol % styrene. 3.4. Recycle Column C2. The bottoms from column C1 are fed to Stage 17 of a 38-stage distillation column whose function is to recover ethylbenzene for recycle. The two design speci?cations are a bottoms impurity of 1 mol % toluene and a distillate impurity of 1 mol % EB. The distillate is mostly benzene and toluene with some water. The bottoms is recycled back to the reaction section. The column operates at 120 kPa, has a diameter of 1.09 m, and requires a re?ux ratio of 3.38. The reboiler duty is 1.3 MW. As we will show in a later section of this paper, the total capital investment of the Vasudevan et al. plant is $7 181 000. The total expenses for purchasing the fresh ethylbenzene feed, the process steam, and the energy required in the furnaces and distillation columns is $174 400 000 per year. A large percentage of this expense is in raw material cost. To produce 115.6 kmol/h of styrene, the fresh feed of EB is 152.6 kmol/h. In the proposed design discussed in a later section, the same amount of styrene is produced from less EB fresh feed (132.8 kmol/h). This results in a much lower expense of $155 800 000 per year. The capital investment is only slightly higher ($8 764 000). 4. Effects of Design Optimization Variables The design discussed in the previous section appears to offer bountiful opportunities for cost reduction, primarily in terms of improving the yield of styrene by reducing the production of unwanted byproducts and reducing losses of raw material and product. In this section we explore the effects of the major design optimization variables on the economics of the process. The major design optimization variables are the process steam, the reactor inlet temperature, the ethylbenzene recycle ?ow rate and the size of the reactors. These will each be explored separately.

4.1. Effect of Process Steam. Using more process steam lowers partial pressures of the reactant and products in the styrene reversible reaction and helps to drive the reaction to the right. However, steam costs increase, and furnace duties increase. A process steam cost of $16.27 per 1000 kg is assumed on the basis of information from Turton et al.5 Remember that boiler feedwater must be generated and converted into steam before it is injected as live steam into the process. The water from the decanter cannot be used as boiler feedwater. The cost of the fuel used in the two furnaces is assumed to be $9.83 per gigajoule on the basis of energy cost data from Turton et al.4 The ?ow rate of process steam was varied to see its effect on process conditions and economic results. In all the results shown below, the production of styrene is kept constant at 115.6 kmol/h. This is achieved by varying the ethylbenzene recycle ?ow rate in each case. Reactor inlet temperatures are kept at 650 °C. As we will see, the fresh feed of ethylbenzene required to produce the 115.6 kmol/h of styrene changes as the process steam ?ow rate is changed. The cost of ethylbenzene is assumed to be $0.50 per pound. It should be noted that the values of the byproducts have not been considered in the economic analysis. The hydrogen in the Lights stream and the benzene and toluene in the D1 distillate could potentially be recovered and sold. However, signi?cant additional capital investment and energy would be required. Developing realistic values for these streams would strongly depend on the prices and costs assumed, which have a large uncertainty. The main purpose of this paper is to develop a reasonable conceptual design. The upgrading of the byproducts has not been considered since it is a minor issue and would involve a great deal of guesswork. The installed costs of the furnaces are calculated from the correlation given on page 915 in Turton et al.5 log10 Cn ) 7.3488 - 1.168 log10 QEn + 0.2028(log10 QEn)2 (13) where Cn is the installed cost of furnace n with an energy requirement of QEn in kW. Figure 5 shows the results of changing process steam from the original 3400 kmol/h up to 4500 kmol/h. As the process steam ?ow rate increases, less EB fresh feed is needed, which translates into signi?cant reductions in raw material cost. This occurs because there is less production of undesirable byproducts, as demonstrated by the reduction in “Lights” from the decanter and “Gas” from the re?ux drum of C1. However, the energy consumption in the furnace E1 increases, which raises energy cost and furnace capital cost. There is little change in the EB recycle needed to keep the styrene production constant at 115.6 kmol/h. Notice the nonmonotonic changes in EB recycle shown in Figure 5 as process steam increases. The initial effect of increasing steam is to decrease partial pressures, which improves selectivity. However, as even more process steam is used, partial pressures become so small that conversion is affected, which eventually requires more EB recycle to maintain the ?xed styrene production. These changes in EB recycle are quite small. Table 2 gives details of the costs for each case. We use the incremental savings and the incremental capital investment in going from one case to another to select the optimum process steam ?ow rate. The savings are the differences between the expenses (EB fresh feed, fuel, and process steam) in moving from one case to the next. Expenses are reduced as the process steam ?ow rate is increased because the reduction in EB fresh feed is much larger than the increase in the cost of process steam


Ind. Eng. Chem. Res., Vol. 50, No. 3, 2011

Figure 5. Effect of process steam. Table 2. Effect of Process Steam Flow Ratea process steam ?ow rate (kmo/h) 3400 QE1 (MW) QE3 (MW) fresh EB (kmol/h) lights (kmol/h) gas (kmol/h) EB recycle (kmol/h) furnace energy cost ($/h) cost process steam ($/h) capital investment furnace (106 $) incremental furnace capital ($) fresh EB savings ($/h) energy and process steam cost increase ($/h) net savings (106 $/y) incremental ROI

3800 18.80 3.40 149.70 183.9 22.63 73.35 785.6136 1112.9 1.7506 61171 343.8634 178.7191 1.4467 23.64

4000 19.79 3.36 148.6 180.4 22.34 73.46 819.2322 1171.4 1.7838 33200 128.6564 92.1906 0.31944 9.6217

4200 20.67 3.32 147.77 177.0 22.07 74.05 848.9581 1230.0 1.8130 29135 97.0771 88.2979 0.076906 2.6396

4500 22.10 3.28 146.72 173.3 21.72 75.47 898.0006 13179 1.861 47977 122.8084 136.9058 -0.12349 -2.574

16.95 3.50 152.64 194.1 23.38 74.61 724.0 995.7 1.690

Styrene production ) 115.6 kmol/h; 650 °C reactor inlet temperatures.

and furnace energy, up to some point of diminishing returns. However, capital investment increases. For example, with a process steam ?ow rate of 3800 kmol/h (second column in Table 2), the capital investment in the furnaces is $1 750 600. With a process steam ?ow rate of 4000 kmol/h, the capital investment is $1 783 800. The incremental capital investment is $33 200. At 3800 kmol/h, the cost of steam is $1112.9 per hour and the cost of furnace fuel is $785.61 per hour. At 4000 kmol/h, these increase to $1171.4 and $819.23 per hour, respectively. This is an increase in total cost of $92.19 per hour. However, the reduction in EB fresh feed (149.7 - 148.6 ) 1.1 kmol/h) represents a raw material savings of $128.65 per hour. So the net savings is $128.65 per hour minus $92.19 per hour, which amounts to $319 440 per year. The increase in furnace capital investment in designing for 4000 kmol/h of process steam instead of 3800 kmol/h is $33 200. Investing this amount to save almost 10 times this amount is an attractive investment. Going from 4000 to 4200 kmol/h of process steam requires an incremental investment of $29 135 and yields a net savings of $76 906 per year, which is still attractive. However, in going to 4500 kmol/h, steam and fuel costs increase by $136.90 per

hour while EB fresh feed savings only increase by $122.81 per hour and capital investment increases by $47 977. To remain on the conservative side, a process steam ?ow rate of 4000 kmol/h is selected as the design point. 4.2. Effect of Reactor Inlet Temperature. With process steam ?xed at 4000 kmol/h, the effect of reactor inlet temperature is explored. Styrene production is held at 115.6 kmol/h for all cases. The inlet temperatures of both reactors are assumed to be the same. Figure 6 shows how several important variables change as reactor inlet temperature is changed from the original 650 °C down to 550 °C. As the top graph on the left shows, the consumption of raw material ethylbenzene decreases as the reactor inlet temperatures are decreased. This is re?ected in the decreases in the ?ow rates of vapor from the decanter (“Lights” shown in the second graph from the top on the left) and gas from the re?ux drum of the ?rst distillation column (“Gas” shown in the second graph from the top on the right). However, the rate of reduction of the required EB fresh feed slows up as temperatures are reduced, so there is a point of diminishing returns. The third graph from the top on the right shows that net expenses (raw material savings minus the costs of process

Ind. Eng. Chem. Res., Vol. 50, No. 3, 2011


Figure 6. Effect of reactor inlet temperature. Table 3. Effect of Reactor Inlet Temperaturea reactor inlet temperature (°C) 550 process steam (kmol/h) fresh EB (kmo/h) EB recycle (kmol/h) lights (kmol/h) gas (kmol/h) QE1 (MW) QR1 (MW) ID1 (m) total capital (106 $) total expenses (106 $/yr) total energy (106 $/yr)

560 4000 131.268 327.8 151.1 12.55 15.23 19.75 7.783 11.485 157.52 12.764

580 4000 132.996 232.2 154.8 12.96 16.25 15.47 6.823 9.9659 158.05 11.529

600 4000 135.618 165.9 159.5 14.01 17.22 12.42 6.06 8.8467 159.95 10.733

650 4000 148.63 73.46 180.4 22.34 19.73 8.236 4.839 7.2156 172.63 10.08

650 3400 152.6426 74.61 194.1 23.38 16.95 8.313 4.86 7.1808 174.44 9.321

4000 130.6469 391.7 149.7 12.67 14.68 22.53 8.355 12.458 157.72 13.603

Styrene production ) 115.6 kmol/h.

steam, furnace energy, and reboiler energy) reach a minimum at about 560 °C. Figure 6 also shows (top graph on the right) that the required recycle of ethylbenzene increases as reactor inlet temperatures are decreased. Higher EB concentrations are required in the reactors to achieve the desired styrene production rate. The higher recycle raises separation costs in the distillation columns. Energy costs in the large E1 furnace and in the reboiler of the ?rst column increase. Capital investments in the furnace and in the ?rst distillation column increase as reactor inlet temperatures are reduced. Table 3 gives details of the economics for each case. The incremental savings in going from 580 to 560 °C is the difference between the total expenses: $158 050 000 per year minus $157 520 000 per year ) $530 000 per year. The incremental increase in capital investment is $11 485 000 minus $9 965 900 ) $1 519 000, giving a return on investment of 35%. Going from 560 to 550 °C increases total expenses instead of decreasing them and requires an increase in capital investment. Therefore, a reactor inlet temperature of 560 °C is selected for the design. 4.3. Effect of Reactor Size. The next design optimization variable considered is reactor size. The diameters of the two reactors are held constant at 3.3 m, and their lengths are varied from the base case 3.5 m up to 10 m. Process steam is 4000 kmol/h and the reactor inlet temperature is 560 °C in all these cases.

Figure 7 shows that the consumption of fresh ethylbenzene increases slowly as the reactor length is increased. This occurs because less EB recycle is required to maintain styrene production, which increases the production of undesirable byproducts. The vapor from the decanter increases as the reactor length increases. However, the very signi?cant reduction in EB recycle reduces energy and capital costs in the separation section, so total capital investment decreases despite the increase in reactor capital investment. The net result in the balance between raw material cost and energy cost is a reduction in total expenses as we move from 3.5 to 7 m in reactor length where a minimum is encountered. As shown in Table 4, going from 10 to 8 m in reactor length increases expenses slightly (raw material cost increases while energy consumption in the separation section decreases) by $580 000 per year. The required increase in capital investment in going from 10 to 8 m is $225 000, which is a 250% return on incremental investment. So the 8 m reactor is better than the 10 m reactor. Going from 8 to 7 m in reactor length again results in a slight increase in expenses but reduces capital investment still further. The savings in expenses in going from 7 m to 8 m is $70 000 per year. The required increase in capital investment is $279 400, which is only a 35% return on incremental investment. Going from 7 m to 6 m results in increases in both expenses and capital investment, which obviously is unattractive. The 8 m reactor is selected for the proposed design.


Ind. Eng. Chem. Res., Vol. 50, No. 3, 2011

Figure 7. Effect of reactor length. Table 4. Effect of Reactor Sizea reactor length (m) 3.5 Fresh EB (kmo/h) EB recycle (kmol/h) lights (kmol/h) gas (kmol/h) QE1 (MW) QR1 (MW) ID1 (m) total capital (106 $) total expenses (106 $/yr) total energy (106 $/yr)

5 131.6287 328.3 153.9 12.31 14.76 15.28 6.778 9.9912 156.12 10.764

6 131.9723 191.0 155.7 12.34 14.59 13.56 6.356 9.4365 155.80 10.326

7 132.356 165.0 157.5 12.42 14.47 12.35 6.043 9.0432 155.73 9.858

8 132.7634 146.0 159.2 12.54 14.38 11.46 5.804 8.7638 155.80 9.517

10 133.7437 1 21.2 163.6 12.85 14.27 10.31 5.479 8.5388 156.38 9.090

131.268 327.8 151.1 12.55 15.23 19.75 7.783 11.485 157.52 12.764

Styrene production ) 115.6 kmol/h; process steam ?ow rate ) 4000 kmol/h; reactor inlet temperatures ) 560 °C.

Table 5. Optimum Distillation Column Design T1 NT NFopt ID1 (m) QR1 (MW) QC1 (MW) TACa (106 $/yr) 72 32 5.979 12.05 11.47 5.468 T2 NT NFopt ID2 (m) QR2 (MW) QC2 (MW) TACa (106 $/yr)

82 37 5.806 11.45 10.87 5.372

102 45 5.674 11.056 10.41 5.491

32 16 1.408 2.077 1.174 0.8206

38 17 1.381 1.987 1.099 0.8127

42 18 1.371 1.965 1.072 0.8153

TAC ) energy cost + capital investment/3 year payback period.

4.4. Optimum Distillation Column Design. The number of stages in each of the distillation columns was optimized by ?nding the number that minimized the total annual cost. The results are shown in Table 5. The optimum values are the same as those given in the Vasudevan et al. design. 4.5. Number of Reactors. The ?owsheets considered above used two reactors in series with two furnaces for preheating. A ?owsheet with three reactors and three furnaces was studied to see if there is any economic incentive to add an additional reaction stage. The process steam ?ow rate was kept at 4000 kmol/h, and the reactor inlet temperature was kept at 560 °C.

A third reactor and a third furnace were added. Capital investment increased because of this change and its effect on other parameters. With three 8 m reactors, the consumption of fresh EB jumped from 132.76 to 136.28 kmol/h, indicating poor economics. With three 3.5 m reactors, the consumption of fresh EB was reduced from the base case of 132.76 to 132.19 kmol/h. However, the required EB recycle climbed from the base case of 146.0 to 194.8 kmol/h. The resulting increase in column energy and capital made capital investment climb from $8 763 800 to $9 921 000 and expenses increased from $155 800 000 per year to $156 290 000 per year. Therefore, a two-stage reaction section is better than three-stage. 4.6. Reoptimization. In the previous sections the design optimization variables have been changed one at a time to ?nd the best values. Several additional cases were examined in which parameters were changed from the optimum values to see if there were better conditions. The independent variation of parameters appears to have resulted in the optimum economic design. For example, with a 8 m reactor and 560 °C reactor inlet temperatures, the process steam ?ow rate was varied from its optimum value of 4000 kmol/h. Results showed that 4000 kmol/h is still the optimum. Increasing steam ?ow rate to 4200 kmol/h reduced the fresh EB feed from 132.7634 to 132.1605 kmol/h. But the total energy increased from $9.517 × 106 per year to $9.765 × 106 per year and capital investment increased from $8 763 800 to $8 817 300.

Ind. Eng. Chem. Res., Vol. 50, No. 3, 2011


Figure 8. Effect of benzene value on net cost.

The net effect on expenses was an increase from $155.80 × 106 to $155.95 × 106 per year. So increasing the process steam ?ow rate from the proposed value is unattractive. Decreasing the steam ?ow rate to 3800 kmol/h increased the fresh EB feed from 132.7634 to 133.4902 kmol/h. The total energy decreased from $9.517 × 106 to $9.2776 × 106 per year and capital investment also decreased from $8 763 800 to $8 719 800. But the increase in raw material cost produced a net increase in expenses from $155.80 × 106 to $156.82 × 106 per year. So decreasing the process steam ?ow rate from the proposed value is also unattractive. The “sweet spot” appears to be 4000 kmol/h.

4.7. Other Improvements. There are several other features that could be explored to improve the ?owsheet. First, the feed to C1 could be preheated by the bottoms from the column to reduce reboiler energy consumption. Second, the hot stream from the feed-ef?uent heat exchanger E2 is at a high enough temperature (392 °C) that it could be used to generate steam or be heat-integrated with the reboilers in the distillation columns instead of throwing the energy away to cooling water. After generating steam, a water-cooled condenser would be required to get down to 40 °C. Neither of these features is included in the proposed ?owsheet because they could also be applied in the Vasudevan et al. ?owsheet and would not affect the comparison. 4.8. Value of Byproducts. One of the reviewers of this paper pointed out that no value has been assigned to the byproduct of the process. The Lights stream contains a lot of hydrogen, and the distillate from the second column contains benzene and toluene. These components could be recovered and sold. However, assigning a realistic value for the components in their impure stage presents a dif?cult problem. The reviewer suggested one approach to this problem. The idea is to concentrate on the benzene and toluene produced and look at a range of possible values. A reasonable price for highpurity benzene and toluene is $94 per kilomole, so this would be the absolute upper limit. The starting value is zero. The “net cost of materials” for the two processes is calculated by subtracting the value of the benzene and toluene produced from the cost of the fresh ethylbenzene.

Figure 9. Proposed design.


Ind. Eng. Chem. Res., Vol. 50, No. 3, 2011

net cost ) (EB feed)($117 per kilomole) - (benzene flow + toluene flow)(value $ per kilomole) (14) An ethylbenzene cost of $117 per kilomole ($0.50 per pound) is used and the value of the benzene and toluene is varied over a wide range. Styrene production is exactly the same in both processes. In the Vasudevan process (see Figure 1), the consumption of EB is 152.6 kmol/h and the benzene and toluene produced are (20.1 kmol/h)(0.4273 mf benzene + 0.4193 mf toluene) ) 17.02 kmol/h. In the proposed process (see Figure 9), the EB consumption is 132.4 kmol/h and the benzene and toluene production is (16.07 kmol/h)(0.1506 + 0.3853) ) 8.61 kmol/h. Figure 8 compares the “net cost of materials” as a function of the value of the benzene/toluene for each process. The cost of the EB feed does not change, but the net cost decreases as the value of benzene/toluene is increased. Since the Vasudevan process makes more benzene/toluene, its net cost line drops more quickly than that of the proposed process. Note that the ordinate is in $1000 per hour. The Douglas Doctrine is clearly demonstrated in these results. The cost of the raw material EB dominates the economics. The value of the byproducts benzene/toluene would have to be very high to justify wasting EB. Notice that the plot goes up to values of $100 per kilomole of byproducts. The ICIS price for benzene is $4 per gallon, which converts to $94 per kilomole. So even if the impure benzene and toluene in the distillate from column T2 cost nothing to separate from the other components, the proposed design has a much lower net cost. The proposed design uses more process steam (4000 versus 3400 kmol/h), but the resulting cost differential is only (600 kmol/h)($16.27/1000 kg)(18 kg/kmol) ) $176 per hour, which is small compared to the $1500 net cost difference between the two processes, even with an unrealistic byproduct value of $100 per kilomole. 5. Proposed Design Figure 9 shows the ?owsheet with the revised parameters. The process steam ?ow rate is 4000 kmol/h. The reactor inlet temperature is 560 °C. Reactors are 8 m in length. Notice that the styrene production rate from the bottom of the Product Column C1 is exactly the same as that in the Vasudevan et al. design. However, the fresh EB feed has been reduced from 152.6 to 132.8 kmol/h. Recycle EB has almost doubled (74.6 to 146 kmol/h). Decanter vapor has dropped from 194.1 to 159.2 kmol/h. Gas from the re?ux drum of C1 has dropped from 23.4 to 12.54 kmol/h. Styrene yield has increased from 76% to 87%. The economic result of these changes in equipment and operating conditions is a reduction in total expenses of 10% (from $174 400 000 per year to $155 800 000 per year), which is a very signi?cant savings in operating costs. This improvement requires an 22% increase in capital investment from $7 181 000 to $8 764 000, which is well justi?ed to attain the cost reduction. Figure 10 gives the composition pro?les in the two reactors. Figure 11 gives the temperature and composition pro?les in the Product Column. Figure 12 gives these pro?les in the Recycle Column.

Figure 10. (A) Composition pro?les in reactor R1. (B) Composition pro?les in reactor R2.

6. Plantwide Control Before exporting the steady-state Aspen Plus simulation into Aspen Dynamics, column re?ux drums and bases are sized to provide 5 min of liquid holdup when at 50% level. The decanter is sized to provide 20 min holdup. Dead times of 5 min are inserted in the two furnace temperature control loops to account for the dynamic lags in furnace ?ring. It is interesting to note that the steady state predicted by the dynamic model in Aspen Dynamics is slightly different from that predicted by the steady-state model in Aspen Plus. Running the Aspen Dynamic simulation out in time until all variables stopped changing gave somewhat different conditions. The total ethylbenzene was ?xed at 278.6 kmol/h. The resulting fresh feed of ethylbenzene was 129.53 kmol/h instead of 132.8 kmol/h found in Aspen Plus. The styrene product changed from 115.6 to 112.5 kmol/h. The production of Lights changed from 159.2 to 155.9 kmol/h. The production of Gas changed from 12.54 to 12.39 kmol/h. The explanation for this small difference in the two steady states probably lies in the difference in the plug?ow reactor models. In Aspen Plus, a distributed ordinary differential equation model is used. In Aspen Dynamics, a multicell “lumped” model is used. 6.1. Control Structure. Figure 13 shows the plantwide control structure developed for this process. Conventional PI controllers are used in all loops. All level loops are proportional with KC ) 2. The C2 column temperature loop has a 1 min

Ind. Eng. Chem. Res., Vol. 50, No. 3, 2011


Figure 11. (A) Temperature pro?le in column C1. (B) Composition pro?les in column C1.

Figure 12. (A) Temperature pro?le in column C2. (B) Composition pro?les in column C2.

dead time. The composition loops have 3 min dead times. All are tuned by using relay-feedback tests (Yu6) to obtain ultimate gains and periods and then applying Tyreus-Luyben tuning rules. The various loops are listed below with their controlled and manipulated variables. 1. The total ethylbenzene ?ow rate (fresh plus recycle from the Recycle Column) is controlled by manipulating the fresh feed of ethylbenzene. 2. The ?ow rate of the steam that is mixed with the total EB stream before feeding into the feed-ef?uent heat exchanger E1 is ratioed to the ?ow rate of the total ethylbenzene. 3. The ?ow rate of the steam that is heated in furnace E2 is ratioed to the ?ow rate of the total ethylbenzene with the ratio adjusted by the output signal from a temperature controller holding the inlet temperature to the ?rst reactor (TCR1). 4. The inlet temperature to the second reactor is controlled by manipulating the fuel to furnace E3 (TCR2). 5. The decanter temperature is controlled by manipulating the heat removal in the condenser E4. 6. The decanter pressure is controlled by manipulating the vapor stream (“Lights”). 7. The water/organic level in the decanter is controlled by the aqueous-phase drawoff rate. 8. The organic level is controlled by the organic drawoff rate, which is the feed to the Product Column. 9. Re?ux in the Product Column is ratioed to the feed to the column, but a 5 min lag is inserted between the feed

?ow signal and the re?ux ?ow to account for the dynamic difference between the effect of feed ?ow rate and re?ux ?ow rate. 10. Gas from the re?ux drum of the Product Column is removed by changing compressor speed to control re?uxdrum temperature. 11. The ethylbenzene impurity on Stage 65 is controlled by manipulating reboiler heat input (CC1). 12. The base level is controlled by manipulating the bottoms ?ow rate, which is the styrene product from the process. 13. Distillate controls the re?ux-drum level and is the feed to the Recycle Column. 14. Distillate in the Recycle Column controls the re?ux-drum level. 15. Re?ux in the Recycle Column is manipulated to keep a speci?ed re?ux ratio, with the re?ux ratio being adjusted by a Stage-7 temperature controller (TC2). 16. The base level is controlled by manipulating the bottoms ?ow rate and is the EB recycle stream back to the reaction section. 17. The toluene impurity on Stage 30 in the Recycle Column is controlled by manipulating the reboiler heat input (CC2). 18. Pressures in both columns are controlled by manipulating condenser heat removal rates. The rationale for the selection of the distillation control structures outlined above is discussed in the next section. Since the product styrene leaves from the bottom of the ?rst column,


Ind. Eng. Chem. Res., Vol. 50, No. 3, 2011

Figure 13. Control structure.

an effective column control structure for this column is vital in terms of product-quality control. 6.2. Column Control Structure Selection. Many industrial distillation columns use some type of single-end temperature control because of its simplicity and low maintenance cost. However, this simple structure may not provide effective control for some columns. Even if a single-end control structure is possible, we have to decide how to select the other control degree of freedom. The most common choices are holding a constant re?ux-to-feed ratio or holding a constant re?ux ratio. The methodology used for selecting column control structures has been discussed in a previous paper.7 A. Selecting Re?ux Ratio or Re?ux-to-Feed Ratio. To explore this question, a series of steady-state runs are made in which the effects of changes in feed composition on the required changes in re?ux-to-feed ratio and re?ux ratio were determined while holding both products at their speci?ed composition. Table 6 gives results of these calculations for the two distillation columns. (1) Product Column C1. In the ?rst distillation column, the feed compositions of the light-key component ethylbenzene and the heavy-key component styrene are varied around the design point (50.61 mol % EB and 40.31 mol % styrene). As more EB and less styrene are fed to the column, there is very little change in the re?ux-to-feed ratio (see Table 6). On the other hand, the re?ux ratio changes about 17% over the range. Therefore, a re?ux-to-feed ratio should work well.

Table 6. Column Control Structure Selection feed composition C1 design 0.55615 EB 0.35307 styrene 0.50615 EB 0.40307 styrene 0.45615 EB 0.45307 styrene 0.02722 toluene 0.9028 EB 0.04722 toluene 0.8828 EB 0.06722 toluene 0.8628 EB 2.587 2.590 2.595 0.4787 0.6202 0.6913 3.966 4.305 4.711 6.078 6.254 5.782 re?ux-to-feed ratio re?ux ratio

C2 design

There may be some dif?culties with this structure because of the high re?ux ratio in this column. Conventional distillation wisdom suggests that the re?ux-drum level should be controlled using re?ux when re?ux ratios are high. A recent paper8 suggests a method for handling the situation in which a re?ux-to-feed ratio structure can be used even in a high re?ux ratio column by using reboiler heat input to control the re?ux-drum level. Since the Product Column operates under vacuum conditions, rapid manipulation of the reboiler heat input is undesirable. Therefore, the re?ux-to-feed structure is selected. Dynamic performance results presented later show that effective control is achieved.

Ind. Eng. Chem. Res., Vol. 50, No. 3, 2011
Table 7. Controller Parameters TCR1 controlled variable manipulated variable SP transmitter range OP OP range dead time KC τI R1 inlet temperature furnace E2 heat duty 560 °C 500-700 °C 0.1859 QE2/TotEB 0-0.3 QE2/TotEB 3 min 0.225 14.5 min TCR2 R2 inlet Temperature furnace E3 heat duty 560 °C 500-700 °C 9.958 GJ/h 0-20.67 GJ/h 3 min 0.915 13.2 min CC1 stage 65 liquid composition reboiler heat input 0.0665 mf EB 0-0.133 mf EB 41.33 GJ/h 0-100 GJ/h 3 min 0.271 37 min CC2 stage 30 liquid composition reboiler heat input 0.0463 mf toluene 0-0.10 mf toluene 7.254 GJ/h 0-14.31 GJ/h 3 min 0.193 33 min TC2


stage 7 liquid composition re?ux ratio 125.9 °C 100-200 °C 7.254 RR2 0-10 RR2 1 min 1.60 15.8 min

(2) Recycle Column. In the second distillation column, the feed compositions of the light-key component toluene and the heavy-key component EB are varied around the design point (4.722 mol % toluene and 88.28 mol % EB). The changes in the re?ux-to-feed ratio over the range of feed compositions are about 34% (see Table 6), and the changes in the re?ux ratio are about 8%. These results suggest that a single-end structure may not provide effective control for feed composition disturbances. Therefore, a dual-end structure is selected for this column. A composition controller and a temperature controller are used, as discussed below. B. Selecting Temperature/Composition Control Tray Location. Another important issue in distillation control structure selection is the location of the tray used for temperature or composition control. A simple approach for temperature control that usually works quite well is to select a location where the temperatures are changing signi?cantly from tray to tray. (1) Product Column. The ethylbenzene/styrene separation is a dif?cult one (low relative volatility), which requires many trays and gives a ?at temperature pro?le. The vacuum operation results in signi?cant pressure changes through the column, and the resulting temperature pro?le is more affected by pressure than composition (see Figure 10A). Therefore, temperature cannot be used in this column. Direct composition measurement and composition control are required. The natural choice of what composition to control would be the EB impurity in the bottoms. However, trying to control the bottoms composition in this high-purity column (99.75 mol % styrene) would result in slow closed loop dynamic response and a high degree of nonlinearity. Distillation control wisdom suggests that it is better to move away from the high-purity end of the column to a tray having larger compositions. Process gains are larger and there is less nonlinearity to complicate closed loop performance. Figure 11B shows that the EB impurity on Stage 65 is 6.38 mol % EB. We select this stage to control. (2) Recycle Column. Figure 12A shows that there are signi?cant changes in temperature from tray to tray in the rectifying section of the column. So we should be able to control one temperature. Since this temperature is near the top of the column, it should do an effective job of maintaining the EB impurity in the distillate. Stage 7 is selected. The temperature controller adjusts the re?ux ratio. The second loop in the Recycle Column must be a composition loop since there is only one temperature break in the column. The bottoms recycle stream is fairly pure ethylbenzene with only 1 mol % toluene. We might consider controlling it directly. Control might be better if we control a composition higher up in the column. Both of these alternatives were tested. Controller tuning with a 3 min dead time produced an integral time of 65 min for control of the bottoms composition. Controller tuning with the same 3 min dead time for controlling the toluene impurity on

Stage 30 produced a 33 min integral time. These results demonstrate that much tighter control can be achieved by using a composition away from the high-purity base of the column. 6.3. Dynamic Performance Results. Several large disturbances were made to test the ability of the proposed plantwide control structure to provide stable regulatory control and hold product streams near their desired speci?cations. Table 7 gives controller parameters for the major loops. A. Throughput Disturbances. Figure 14 gives the responses of a number of important variables when 20% changes in the set point of the ?ow controller on the total ethylbenzene fed to the reactors are made at a time equal 0.2 h. The solid lines are for 20% increases, and the dashed lines are for 20% decreases. Stable regulatory control is achieved for these large disturbances. The change in the total EB immediately changes the two steam addition rate through the action of the ratio elements. Notice that there are large transient changes in the ?ow of the fresh EB feed. The upper right graph in Figure 14A shows that the fresh EB feed climbs to almost 300 kmol/h for an increase in total EB ?ow and drops to zero for a while for a decrease in total EB ?ow. The result of the 20% changes in total EB are 16% changes in the fresh feed of EB and 16% changes in the styrene product (from 112.5 up to 129.3 kmol/h or down to 94.7 kmol/h). The change in EB recycle is about 23%, and the changes in Lights are about 18%. The third graph on the left in Figure 14A shows that the control of reactor R1 inlet temperature is quite good with a deviation of only 10 °C. This is due to the QE2-to-TotEB ratio that makes immediate changes in the furnace fuel as inlet ?ows change. The re?ux changes (through the lag) as the feed to the column (“Organic”) changes. The top right graph in Figure 14B shows that the purity of the styrene product xB1(S) is maintained very close to its speci?cation. Remember that this composition is not directly controlled. The composition of Stage 65 is controlled by manipulating reboiler heat input. The second graph on the right in Figure 14C shows that the impurity of EB in the distillate of the Recycle Column undergoes some transient disturbances but is brought back to very close to its speci?ed value in about 3 h. The changes in the ?ow rate of the distillate D2 from the Recycle Column are very slow, taking almost 10 h to come to a new steady state. The changes in the ?ow rate of the EB recycle (B2) occur fairly quickly, taking only about 2 h to come to steady state. B. Reactor Inlet Temperature Disturbances. The solid lines in Figure 15 give the responses of the system when the set points of the two reactor inlet temperature controllers are increased from 560 to 570 °C at a time equal 0.2 h. The total ethylbenzene is held constant. The higher reactor temperatures lead to increases in Lights (155.9-171.1 kmol/h), Gas (12.29-13.62 kmol/h), and fresh EB (129.5- 139.4 kmol/h). There is only a small increase in


Ind. Eng. Chem. Res., Vol. 50, No. 3, 2011

Figure 14. 20% disturbances in total EB.

Ind. Eng. Chem. Res., Vol. 50, No. 3, 2011


Figure 15. Disturbances in reactor inlet temperature and steam/EB ratio.


Ind. Eng. Chem. Res., Vol. 50, No. 3, 2011 QRn ) reboiler heat input in column n Rn ) re?ux ?ow rate in column n Tot EB ) sum of fresh and recycle EB TRn ) inlet temperature to reactor n T7 ) temperature on Stage 7 in Recycle Column xB1(S) ) mole fraction styrene in bottoms from Product Column C1 xB2(T) ) mole fraction toluene in bottoms from Recycle Column C2 xD1(S) ) mole fraction styrene in distillate from Product Column C1 xD2(EB) ) mole fraction EB in distillate from Recycle Column C2 x30(T) ) mole fraction toluene on Stage 30 in Recycle Column C2 x65(EB) ) mole fraction EB on Stage 65 in Product Column

styrene product (112.5-118.3 kmol/h), so the styrene yield drops from 86.8 to 85.6% as a result of the higher temperature. C. Steam-to-EB Ratio Disturbances. The dashed lines in Figure 15 give the responses when the ratio between the steam fed to furnace E2 and the total EB is reduced from 12.155 to 9.724. There are some small transients, but the long-term effect on the process is small. There are small increases in the re?ux and reboiler heat input in the Recycle Column. 7. Conclusion The styrene process illustrates a number of interesting design trade-offs. Low reactor temperatures improve styrene yield but require more ethylbenzene recycle to maintain conversion, which increases energy and capital costs in the separation section of the process. Higher process steam ?ows improve yield and selectivity but increase furnace capital and fuel costs and increase the cost of providing the process steam. Economics are dominated by improving yield to reduce raw material costs. The proposed ?owsheet uses more process steam, larger reactors and higher EB recycle than the original ?owsheet, and increases the yield of styrene from 76 to 87%. A 10% reduction in operating costs is achieved ($18 000 000 per year) for a 22% increase in capital investment (from $7 181 000 to $8 764 000). The control of the process requires effective control structures for the distillation columns from which the styrene product and the recycle EB come. The Product Column uses a re?ux-tofeed single-end control structure, but temperature control cannot be used because of the ?at temperature pro?le. Composition control on an intermediate tray avoids slow dynamics and nonlinearity. A two-end control structure is required in the Recycle Column (one temperature controller and one temperature controller). Nomenclature
Bn ) bottoms ?ow rate from column n Cn ) distillation column n Dn ) distillate ?ow rate from column n FF EB ) fresh feed of EB Fn ) feed ?ow rate to column n

Literature Cited
(1) Douglas, J. M. Conceptual Design of Chemical Processes; McGrawHill: New York, NY, U.S., 1988. (2) Vasudevan, S.; Rangaiah, G. P.; Murthy Konda, N. V. S. N.; Tay, W. H. Application and evaluation of three methodolgies for plantwide control of the styrene monomer plant. Ind. Eng. Chem. Res. 2009, 48, 10941–10961. (3) Lee, A. K. Y.; Ray, A. K.; Rangaiah, G. Multiobjective optimization of an industrial styrene reactor. Comput. Chem. Eng. 2003, 27, 111–130. (4) Tarafder, A.; Rangaiah, G.; Ray, A. K. Multiobjective optimization of an industrial styrene monomer manufacturing process. Chem. Eng. Sci. 2005, 60, 347–363. (5) Turton, R.; Bailie, R. C.; Whiting, W. B.; Shaelwitz, J. A. Analysis, Synthesis and Design of Chemical Processes, 2nd ed.; Prentice Hall: Englewood Cliffs, NJ, U.S., 2003. (6) Yu, C. C. Autotuning of PID Controllers, 2nd ed.; Springer: Berlin, Germany, 2006. (7) Luyben, W. L. Method for evaluating single-end control of distillation columns. Ind. Eng. Chem. Res. 2009, 48, 10594–10603. (8) Luyben, W. L. Unusual control structure for high re?ux ratio distillation columns. Ind. Eng. Chem. Res. 2009, 48, 11048–11059.

ReceiVed for reView January 4, 2010 ReVised manuscript receiVed September 1, 2010 Accepted September 26, 2010 IE100023S