- Email: [email protected]

METHYL-TERT-BUTYL-ETHER CATALYTIC DISTILLATION COLUMN Part II: Optimization H. S. ELDARSI and P. L. DOUGLAS Department of Chemical Engineering, University of Waterloo, Waterloo, Ontario, Canada

T

his is the second of two papers on the catalytic distillation process to produce methyltert-butyl-ether (MTBE) from methanol and isobutylene contained in a mixed butylenes stream. In this paper, an MTBE catalytic distillation column was optimized using AspenPlus. The optimization studies indicate that the market price of the MTBE product and the cost of the butylenes feed play a key role in determining the MTBE column pro® t. The pro® t is also affected by isobutylene composition changes. In addition, there are signi® cant economic penalties if the column decision variables are not updated for changes in utility costs. Keywords: MTBE catalytic distillation; steady state simulation; optimization

INTRODUCTION

PROCESS DESCRIPTION

Conducting a catalysed chemical reaction in a distillation process offers many potential advantages over a typical reactor and subsequent distillation such as, reduced capital and operating costs, utilization of the heat of reaction to improve the distillation, and drive the reversible chemical reaction to completion, thereby increasing reactant conversion and improving product selectivity. In addition, the synthesis of methyl-tert-butyl-ether (MTBE) exhibits additional economical incentives that are mainly driven from the conversion of isobutylene to MTBE, which has a lower vapour pressure, higher octane number, is environmentally friendly, and has a higher selling price. As a result of high isobutylene conversion, there is a reduction in catalytic reforming severity, thereby decreasing the re® nery fuel consumption and losses. This increases the overall re® nery throughput as a consequence of a greater number of components such as butane and lighter hydrocarbons that are added to the gasoline pool, Pecci 9 . By converting isobutylene to MTBE, a higher purity of the inerts (e.g., 1-butene) is obtained at a much lower cost compared to other chemical processes. Isobutylene and 1-butene contained in the butylenes stream are dif® cult to separate because their relative volatility is close to 1. Therefore, it is very expensive to obtain a stream with high purity of the 1-butene which has important chemical enduses, Clementi3 . However, the MTBE process offers an attractive method for separating isobutylene from the 1-butene since isobutylene reacts with methanol to produce MTBE which is separated as a bottoms product, and the 1-butene stream is recovered as an overhead product. The previous simulation work in catalytic distillation (CD) process of MTBE synthesis were limited to modelling experimental or industrial processes, Abufares1 , Eldarsi5 , Jacobs6 . This paper reports on a steady state optimization study on an existing CD column producing MTBE.

The optimization study was conducted on a CD column with con® guration and feed speci® cations illustrated in Figure 1; this column is identical to the one simulated by Jacobs5 . The RADFRAC module with user supplied reaction kinetics subroutine from AspenPlus simulator, Venkatarman12 , was used to model the CD column process. The gas and liquid phases were modelled using the Redlich-Kwong/UNIQUAC property routines in AspenPlus. The reaction kinetics and UNIQUAC binary interaction parameters were taken from Reh® nger11 . The ¯ ow rate and composition of the butylenes stream were assumed to be a function of upstream production unit operation (i.e., the ethylene unit in this case). Three different severities of this unit (low, moderate, and high) that result in three different ¯ ow rates and compositions of the butylenes stream were assumed. The ethylene unit severities correspond to isobutylene mole fractions 0.35, 0.42, and 0.53 respectively in the butylenes stream. The feed location of the butylenes feed was ® xed at stage 11. The methanol feed stage was ® xed on stage 10. The ¯ ow rate of the methanol was supplied at slightly above stoichiometricrate to isobutylene (198/197 mol s- 1 ) because:

· by continuously removing the product (MTBE) in the CD

column via distillation, a higher isobutylene conversion is achieved; thus, a large excess of methanol to increase isobutylene conversion is not required. · multiple steady states do not exist at this methanol ¯ ow rate and high isobutylene results, consequently, there is only a unique high isobutylene conversion, Eldarsi5 . Considerations such as safety, equipment capacities, and product purity determine the constraints imposed on the process. Two constraints were implemented, the maximum vapor ¯ ow in the CD column was set at 6.4 m3 (3000 mol s- 1 ) to limit the column diameter, and the 517

518

ELDARSI and DOUGLAS

Figure 1. A schematic representation of the column and feed speci® cations.

concentration of MTBE in the bottom product was maintained at # 98 mol% by varying the re¯ ux ratio. THE SELECTION OF DECISION VARIABLES The possible decision variables were restricted to those variables used to specify the CD system. These variables included:

· · · · · · · ·

number of stages (reactive and non-reactive) weight of dry catalyst per stage feed stages location ¯ ow rates and compositions of the butylenes feed ¯ ow rate of the methanol feed re¯ ux ratio bottom ¯ ow rate (MTBE solution) column operating pressure

The decision variables were considered further and several were eliminated as candidates for the optimization study.

· the number of reactive and non-reactive stages and

weight of dry catalyst per stage were eliminated as possible decision variables since the CD column was assumed to be existing and in place · the ¯ ow rate and composition of the butylenes was assumed to be set by the upstream ethylene unit, and therefore not available to manipulate The location of the butylenes feed stage was ® xed at stage 11 (last reactive stage) because of the relatively high volatility the ¯ ow rate of the methanol was maintained at 198 mol s- 1 for the reasons described earlier. An additional check was carried out by varying the methanol ¯ ow rate

between 197 to 200 mol s- 1 : the objective function value was found not to change signi® cantly as will be illustrated later. It is important to note that the upper bound of methanol ¯ ow rate was set to 200 mol s- 1 to avoid the existence of multiple steady states. From the simulations there are two methanol feed locations (stages 10 and 11) that exhibit high isobutylene conversions at the above mentioned ¯ ow rate of methanol, Eldarsi5 . The feed tray location of the methanol was ® xed on stage 10 since stage 11 was used as feed stage of the butylene stream. The possible remaining decision variables therefore are:

· re¯ ux ratio · MTBE bottom ¯ ow rate · operating pressure of the CD column In order to determine how a variation of the above decision variables affects the objective function, a sensitivity study was performed. Based on this study, the range of maximum pro® t for each decision variables was determined, and good starting points were used in the simulations, allowing the optimizer to converge to the solution representing the maximum pro® t. FORMULATION OF THE OBJECTIVE FUNCTION For an existing CD column, the objective function (pro® t) was de® ned as profit = revenue - costs. The costs term included the raw material costs (i.e., butylenes and methanol) multiplied by their ¯ ow rates, and utility costs include steam and cooling water in the partial reboiler and the total condenser multiplied by their ¯ ow rates. The revenues term was the selling price of bottom and overhead products multiplied by their ¯ ow rates. Trans IChemE, Vol 76, Part A, May 1998

MTBE CATALYTIC DISTILLATION COLUMNS: PART II The bottom stream is the MTBE solution. The overhead stream is a high purity 1-butene stream (concentration $ 98 mol%) with negligible amounts of methanol, isobutylene, and MTBE. These trace components do not change the physical properties of the 1-butene such as molecular weight, vapour pressure, and octane number which are approximately similar to those of butylenes feed stream: the magnitudes of both over head product and the butylenes feed streams were assumed in order that their values might be assessed as a replacement for normal butane as gasoline pressuring stock, Mobil6 . For this reason, both streams were assigned the same marketing value. The value of both streams was determined from: (1.023 ´ price of normal butane - 0.023 ´ price of regular unleaded gasoline). The MTBE plant was assumed to be located on the US gulf coast with nominal production of 492,000 metric tons of MTBE/year. The plant was operated with its full capacity production rate 24 hours/day during 330 days/year. The raw material methanol cost was obtained from Chemical Marketing Reporter2 . The prices of normal butane, regular unleaded gasoline, and MTBE were obtained from Platt’ s Oilgram Price Report10 . The utility costs were obtained from Novocor, Sarnia8 .

OPTIMIZATION PROCEDURE The optimal decision variables of the CD were determined by implementing the following optimization procedure. The AspenPlus’ s Complex optimization algorithm together with the model and an economic objective function are performed to maximize the pro® t or minimize the operating cost, namely: Choose

P, RR, bottom ¯ ow rate (MTBE)

to Maximize

Pro® t = Revenues - costs

st.

±the concentration of MTBE in the bottom ¯ ow rate is $ 98 mol% ±the maximum vapour ¯ ow rate in the CD column is 6.4 M3 (3000 mol s- 1 )

519

Table 1. High, low and nominal raw materials and product selling price. Raw materials costs and products selling prices

High

Low

Nominal

Methanol (US $/Gal)

0.55

0.35

0.44

Butylenes feed and overhead product (US $/Gal)

0.53

0.30

0.39

MTBE (US $/Gal)

0.93

0.67

0.76

In order to re¯ ect the actual costs and products selling price variations on the objective function values, a survey was conducted using Platt’ s Oilgram Price Report10 from October 1995 to June 1996. Based on this survey, low and high raw material costs and product selling prices during this period were found, and they are shown in Table 1. The utility costs were set to free and high utilities costs. The free utility costs represents free heated steam and cooling water, the high utility costs represents four times heated steam nominal costs. High, free, and nominal utility costs are shown in Table 2. The objective function values were calculated on the basis that one cost or selling price changed, while the rest of costs and selling prices remained at their nominal values. For example, by assuming that the methanol cost jumped from its nominal level to its higher level while the other costs and selling prices were ® xed at their nominal values, both the objective function value which represents the effect of higher methanol cost, and new set of decision variables were determined. Fourth, investigation of all possible combinations of the costs and selling prices (high and low) on the objective function, i.e., factorial design experiments were carried out. Each cost or selling price was to be estimated at several levels of other costs and selling prices. Therefore, the most signi® cant costs and selling prices that have a large impact on the objective function were found. Since there are four factors each at two levels there are 24 = 16 treatment combinations; the four factors are:

The optimization procedure was classi® ed into ® ve categories detailed below. First, calculation of the objective function value at an initial column speci® cation with nominal raw materials and utilities costs and the products selling prices. This is referred to as the `unoptimized objective function’ . Second, calculation of the objective function value and determination the optimal set points of the decision variables at nominal costs and the products selling prices. This is referred to as the `base case objective function’ . Third, calculation of the change in the objective function value with respect to change in raw materials and utility costs, and products selling prices. This is known as `postoptimality’ or `sensitivity analysis’ which is very important, Edger4 because:

1. the cost of the butylenes feed and selling price of the overhead product, factor (BS) 2. the cost of methanol, factor (M) 3. the selling price of MTBE, factor (MT); the high and low levels values of the above three factors are shown in Table 1. 4. the cost of utility, factor (U), with high level denoting the high stream and cooling water costs, and low level denoting their free costs as represented in Table 2.

· the raw materials and utility costs or products selling

Utility costs

High

Free

Nominal

Saturated steam (US $/1000 lb) Cooling water (US $/MMBTU)

0.017

0

0.004

0.16

0

0.16

markets may be poorly known and the objective function may be in¯ uenced by changes in these costs and prices. · the operating parameter set points (i.e., decision variables in this case) of the CD column may change in response to change in the objective function values. Trans IChemE, Vol 76, Part A, May 1998

Fifth, calculation of the penalty when the raw materials, utility costs, and product selling marketing change and the Table 2. High, free and nominal utility cost.

520

ELDARSI and DOUGLAS Table 3. The objective function values and decision variables (IB conc. = 35 mol%). Decision Variables

Scenario 1-UNOPT 2-BASE 3-FUTIL 4-HUTIL 3-HMTBE 3-LMTBE 3-HBUT 3-LBUT 3.HMeOH 3-LMeOH

Description

P (ATM)

RR

MTBE rate (mol s- 1 )

Nominal costs and selling prices without optimization Nominal costs and selling prices with optimization Free utility costs High utility costs High MTBE selling price Low MTBE selling price High butylenes cost and 1-butene selling price Low butylenes cost and 1-butene selling price High methanol cost Low methanol cost

11 8.87 9.20 8.62 8.91 8.75 8.78 8.85 8.92 8.86

7 3.65 6.96 3.15 4.17 3.46 3.38 3.86 3.69 3.65

197 196.40 197.36 195.71 196.73 196.20 196.10 196.55 196.42 196.40

optimization is not updated (i.e., on-line optimization is not in place). The calculation was carried out as follows: 1. apply the decision variables that result from `base case objective function’ into different costs and selling prices simulations. 2. run the simulations without optimization and calculate the objective function values at the new costs and prices. 3. the penalty for not updating the decision variables is the difference between steps 1 and 2 above. The above procedure was repeated for three ¯ ow rates and compositions of the butylenes stream using standard volumes (i.e., liquid at 60 F and 148 psi) for all feeds and products ¯ ow rates. RESULTS AND DISCUSSION Different Costs and Selling Prices Scenarios Tables 3, 4 and 5 show the results of categories 1, 2, and 3 in the optimization procedure. Each table represents speci® c isobutylene concentration in butylenes feed, the objective function values, and decision variables under different costs and prices scenarios. The bene® ts and drawbacks of the various scenarios (BASE, FUTIL, HUTIL, HMTBE, LMTBE, HBUT, LBUT, HMeOH, LMeOH) at three isobutylene compositions are discussed in the following paragraphs. A detailed analysis for each scenario is made. The BASE scenario, (nominal costs and selling prices

Objective function value (US $/yr) 52,605,500 53,403,200 54,580,100 51,556,700 83,413,700 37,528,600 34,269,500 65,725,400 46,788,600 58,817,500

after optimization), show higher objective function values compared to the UNOPT scenario, (nominal costs and selling prices without optimization). In other words, the bene® ts for implementing the optimization at nominal costs and selling prices; the economic bene® ts of optimization were estimated to be greater than US$500,000/yr. However, the value of the objective function will vary with various costs and selling prices scenarios, as will be illustrated later. The FUTIL scenario, (free utility costs), will result in higher re¯ ux ratios. As a consequence of increasing the re¯ ux ratio, the MTBE ¯ ow rate increases resulting in a higher objective function value compared to that of the BASE scenario, (nominal costs and selling prices after optimization). However, the increase in re¯ ux ratio led to higher operating pressures which in turn led to high temperatures. An opposite effect was observed by applying the HUTIL scenario, (high utility costs), which results in a reduction in re¯ ex ratio and hence less MTBE was formed. As a result, the objective function value was less than that of the BASE scenario, (nominal costs and selling prices after optimization). The decrease in the re¯ ux ratio also causes a decrease in the operating pressure. The HMTBE scenario, (high MTBE selling price), offers the highest objective function value, though higher demands on the re¯ ux ratio were observed (the second highest re¯ ux ratio from the FUTIL scenario, free utility costs). The signi® cant increase in the objective function value is attributed to the higher MTBE selling price, since the

Table 4. The objective function values and decision variables (IB conc. = 42 mol%). Decision Variables Scenario 1-UNOPT 2-BASE 3-FUTIL 3-HUTIL 3-HMTBE 3-LMTBE 3-HBUT 3-LBUT 3-HMeOH 3-LMeOH

Description

P (ATM)

RR

MTBE rate (mol s- 1 )

Nominal costs and selling prices without optimization Nominal costs and selling prices with optimization Free utility costs High utility costs High MTBE selling price Low MTBE selling price High butylenes costs and 1-butene selling price Low butylenes cost and 1-butene selling price High methanol cost Low methanol cost

11 8.88 9.25 8.50 8.89 8.78 8.79 8.86 8.87 8.86

7 4.54 9.54 3.83 4.81 4.44 4.30 4.62 4.52 4.53

197 197.67 198.36 196.90 197.80 197.63 197.51 197.72 197.66 197.67

Objective function value (US $/yr) 53,175,200 53,865,600 54,873,000 52,230,500 84,064,900 37,891,500 34,550,400 66,297,800 47,248,300 59,280,500

Trans IChemE, Vol 76, Part A, May 1998

MTBE CATALYTIC DISTILLATION COLUMNS: PART II

521

Table 5. The objective function values and decision variables (IB conc. = 53 mol%). Decision Variables Scenario 1-UNOPT 2-BASE 3-FUTIL 3-HUTIL 3-HMTBE 3-LMTBE 3-HBUT 3-LBUT 3-HMeOH 3-LMeOH

Description

P (ATM)

RR

MTBE rate (mol s- 1 )

Nominal costs and selling prices without optimization Nominal costs and selling prices with optimization Free utility costs High utility costs High MTBE selling price Low MTBE selling price High butylenes cost and 1-butene selling price Low butylenes cost and 1-butene selling price High methanol cost Low methanol cost

11 8.85 9.11 8.38 8.87 8.81 8.64 8.86 8.70 8.73

7 6.77 14.22 5.80 7.05 6.31 6.08 6.86 6.63 6.67

197 199.03 199.46 198.61 199.10 198.87 198.80 199.05 199.03 199.03

MTBE production rate was less than that of the FUTIL scenario, (free utility costs). On the contrary, the low MTBE selling price scenario (LMTBE), shows the second lowest objective function value; the lowest results from the HBUT scenario, (high butylenes feed cost and 1-butene selling price), though it has a higher re¯ ux ratio. This occurs because a low selling MTBE price in the LMTBE scenario is still higher than the high butylenes cost and 1-butene selling price in the HBUT scenario. Consequently, more MTBE was formed to maximize the pro® t. HBUT scenario, (high butylenes feed cost and 1-butene selling price), shows the lowest objective function value even with lower re¯ ux ratio (the second lowest from the HUTIL scenario, a high utility costs). The lowest objective value occurred because of higher butylenes feed cost. Furthermore, the butylenes feed ¯ ow rate was ® xed in the simulations (the butylenes feed rate cannot be reduced to maximize the objective function value). The decrease in re¯ ex ratio took place, since a little more overhead product was formed in response to its higher selling price; therefore, less MTBE was produced. Although the LBUT scenario, (a low butylenes feed cost and 1-butene selling price), represents in part a low overhead product (1-butene) selling price, this scenario demonstrates a signi® cant increase in the objective function (the second highest objective function value from the HMTBE scenario, a high MTBE selling price). Higher objective function is the result of lower raw material butylene cost, since the ¯ ow rates of the butylenes feed to overhead product (1-butene) is varying from 1.5±2.1:1 as observed in the simulations. This indicated that the less the raw material butylenes cost the higher the objective value. The HMeOH, (high methanol cost), and LMeOH, (low methanol cost), scenarios did not signi® cantly change the optimal decision variables compared to the BASE scenario, (nominal costs and selling prices after optimization), or to each other. However, the objective function values of both scenarios are different as a direct in¯ uence of high and low methanol costs. As was stated earlier, the methanol ¯ ow rate was selected as an additional manipulated variable and varied from 197 to 200 mol s- 1 . The results showed that manipulating the ¯ ow rate of the methanol has only a minor impact on the objective function values, as well as the set point of the ¯ ow rate of the methanol. Three speci® ed values Trans IChemE, Vol 76, Part A, May 1998

Objective function value (US $/yr) 53,717,900 54,313,100 55,193,200 52,834,600 84,706,700 38,237,400 34,819,200 66,862,800 46,697,500 59,727,00

of methanol ¯ ow rate and objective function in Table 6 are compared with their counterparts in Table 3. These results are not surprising for the following reasons: 1. ® rstly, the ¯ ow rate of the methanol is still smaller than the ¯ ow rate of the butylenes in all three butylene stream conditions 2. secondly, the difference in costs between both feed streams is not very high. An important point to note is that as isobutylene concentration increases in the butylenes feed stream (i.e., increasing the ethylene unit severity), both the objective function and the re¯ ux ratio increase. This can be explained by recalling the rate of reaction equation which is expressed in terms of component activities, Reh® nger11 . The increase in isobutylene concentration causes an increase in the isobutylene activity, since the activity is the product of the isobutylene concentration (mole-fraction) and its activity coef® cient. Therefore, as the isobutylene concentration increases, the reaction rate increases accordingly. This causes more MTBE to be formed; consequently, higher objective function values result since the MTBE selling price is assigned the highest value. However, as more MTBE is produced, the re¯ ex ratio increases to maintain the desired concentration of MTBE. The increase in the re¯ ux ratio does not result in lower objective function values for the following reasons:

· as the severity of the ethylene unit increases, the ¯ ow rate

of the butylenes to the CD column decreases. Consequently, there is a smaller costs term in the objective function. · a reduction in the reboiler duty occurred as a result of decreased ¯ ow rate of the butylenes feed. The reboiler duties of isobutylene mole fraction, 0.35, 0.42, and 0.53 in the butylenes feed were calculated to be 16, 14, and 13 MW respectively. The decrease in the reboiler duties results in a Table 6. The effect of manipulating methanol ¯ ow rate on objective function. Scenario

MEOH rate set points (mol s- 1 )

Objective function value (US $/yr)

BASE HMeOH LMeOH

198.1 197.9 198.3

54,403,600 46,780,500 58,823,100

522

ELDARSI and DOUGLAS Table 7. Design for 24 MTBE column experiment. Factor level settings

Run

BS

M

MT

U

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

+ + + + + + + +

+ + + + + + + +

+ + + + + + + +

+ + + + + + + +

Objective function US $/year 56,405,600 24,706,600 44,411,800 12,681,800 102,509,000 73,818,200 90,478,500 58,780,000 53,371,500 22,227,100 41,340,900 10,196,600 99,094,000 67,761,100 87,063,400 55,730,500

smaller utility costs term, thus, an increased objective function. Factorial Design Experiments Results Analysis A 24 factorial design experiment for an isobutylene concentration of 35 mol% in the butylenes feed was carried out using 16 treatment combinations. The objective function values were determined as represented in Table 7. The `+ ’ and `- ’ signs refer to the high and low level values for each factor respectively. The contrasts which demonstrate the total effect of each factor or combination (BS, M, MT, U and interaction terms such as BS-M, etc.) were calculated. The BS contrast is calculated by multiplying the objective function in Table 7 by either - 1 or + 1 (depending on the sign in the BS column for each run) and summing the results. Similarly the contrast for M can be calculated by multiplying the objective function by - 1 or + 1 (depending on the sign in the M column for each run) and summing the results. The results were then ordered and plotted on normal probability paper. From the graph, one can clearly see that the MT, M and BS factors are signi® cant while the others are near 0. In other words the MT, M and BS factors have a large value, either positive or negative, far from the other contrasts which form straight line and have negligible effects on the objective

Figure 2. Normal probability plot of all contrast shows the signi® cance of both the MTBE selling price and the butylenes feed cost on the objective function.

function. The factors that have large magnitudes usually corresponds to signi® cant effects. From Figure 2, the factors MT (selling price of MTBE) and BS (the cost of butylenes feed and selling price of overhead product [1-butene]) were found to be the most signi® cant factors; the factor M (methanol cost) is also potentially important. It is important to note that the signi® cant of the factor BS refers to the cost of butylenes feed since the ¯ ow rates of the butylenes feed is larger than its counterpart of the over head product by 1.5 Experimental design results indicted that the MTBE selling price and the cost of the butylenes feed are the major factor in deciding the values of the objective function (pro® t). Furthermore, these results are similar with those from different costs and selling prices scenarios (category no. 3 in the optimization procedure). For these reasons, there is no need to run factorial design experiments for isobutylene concentration of 42 and 53 mol% in the butylenes feed. Penalty for Not Updating On-Line Optimization When utility costs and selling prices change and the optimization is not `re-run’ or updated one may face signi® cant economic penalties or lost pro® ts. Tables 8, 9 and 10 show the penalties for each scenario if the CD column manipulated variables representing the base case scenario are not updated by re-running the optimization. The results in the above tables indicated that an MTBE column is moderately sensitive to changes in utility costs;

Table 8. The penalty for not updating optimization (IB conc. = 35 mol%).

Scenario

Objective function values with on-line optimization (US $/yr)

Objective function value without on-line optimization (US $/yr)

Penalty (US $/yr)

FUTIL HUTIL HMTBE LMTBE HBUT LBUT HMeOH LMeOH

54,580,100 51,556,700 83,413,700 37,528,600 34,269,500 65,725,400 46,788,600 58,817,500

54,255,700 51,250,200 83,401,100 37,525,800 34,245,800 65,722,800 46,786,200 58,816,800

- 324,400 - 306,200 - 12,600 - 2,800 - 23,700 - 2,600 - 2,400 - 700

Trans IChemE, Vol 76, Part A, May 1998

MTBE CATALYTIC DISTILLATION COLUMNS: PART II

523

Table 9. The penalty for not updating optimization (IB conc. = 42 mol%).

Scenario

Objective function values with on-line optimization (US $/yr)

Objective function value without on-line optimization (US $/yr)

Penalty (US $/yr)

FUTIL HUTIL HMTBE LMTBE HBUT LBUT HMeOH LMeOH

54,873,000 52,230,500 84,064,900 37,891,500 34,550,400 66,297,800 47,248,300 59,280,500

54,614,600 51,963,800 84,022,400 37,885,500 34,532,200 66,293,100 47,248,100 59,278,700

- 258,400 - 266,700 - 42,500 - 6,000 - 18,200 - 4,700 - 200 - 1,800

Table 10. The penalty for not updating optimization (IB. conc. = 53 mol%).

Scenario

Objective function values with on-line optimization (US $/yr)

Objective function value without on-line optimization (US $/yr)

Penalty (US $/yr)

FUTIL HUTIL HMTBE LMTBE HBUT LBUT HMeOH LMeOH

55,193,200 52,834,600 84,706,700 38,237,400 34,819,200 66,862,800 47,697,500 59,727,000

55,002,000 52,556,600 84,699,000 38,225,900 34,793,500 66,860,800 47,695,900 59,726,500

- 191,200 - 278,000 - 37,700 - 11,500 - 25,700 - 2,000 - 1,600 - 500

therefore, there is economic penalty if the model is not updated, whereas small penalties were observed under other different scenarios. Consequently, to justify on-line optimization applications one would need more information about the frequency of such changes in utility costs as well as the cost of implementing on-line optimization. CONCLUSIONS 1. An AspenPlus model representing an existing industrial scale CD column for the production of MTBE was developed and optimized. 2. A sensitivity and factorial design analysis indicated that the most signi® cant factors affecting the pro® t of MTBE column are the MTBE product selling price and the butylenes feed cost. Not surprisingly, the pro® t can be signi® cantly improved by increasing the MTBE product selling price or decreasing the butylenes feed costs. 3. Increasing an isobutylene concentration in the butylenes feed results in higher MTBE formation and lower reboiler and condenser duties; consequently, higher column pro® t. 4. Changes in the utility costs requires changes in the decision variables of the CD column to remain at the optimum. If the model is not updated to re¯ ect these changes, an economical penalty will be incurred. The penalty is the most severe for changes in the utility costs ranges from US$200,000/yr to US$300,000/yr. This is the incentive for real time optimization applications. NOMENCLATUR E BASE BS CD

nominal costs and selling prices after optimization scenario the cost of butylenes feed and selling price of the overhead product factor in factorial experimental design catalytic distillation

Trans IChemE, Vol 76, Part A, May 1998

conc. FUTIL HBUT HMeOH HMTBE HUTIL IB 1B LBUT LMeOH LMTBE M MeOH MT MTBE U UNOPT

concentration free utility costs scenario high butylenes cost and 1-butene selling price scenario high methanol cost scenario high MTBE selling price scenario high utility costs scenario isobutylene 1-butene low butylenes cost and 1-butene selling price scenario low methanol cost scenario low MTBE selling price scenario the cost of methanol factor in factorial experimental design methanol the selling price of MTBE factor in factorial experimental design methyl tert butyl ether the cost of utility factor in factorial experimental design nominal costs and selling prices without optimization scenario

REFERENCES 1. Abufares, A. A. and Douglas, P. L., 1995, Mathematical modelling and simulation of an MTBE catalytic distillation column using SpeedUp and AspenPlus, TransIChemE, 73(A1): 3±12. 2. Chemical Marketing Reporter, October 1995±June 1996. 3. Clementi, A., Oriani, G., Ancillotti, F. and Pecci, G., 1979, Up-grade C4 with MTBE process, Hydrocarbon Processing, December, 109± 113. 4. Edger, T. F. and Himmelbau, D. M., 1988, Optimization of Chemical Process (McGraw-Hill). 5. Eldarsi, H. S. and Douglas, P. L., 1998, Methyl-tert-butyl-ether catalytic distillation column. Part I: Multiple steady states, Trans IChemE, 76(A4): 509±516. 6. Jacobs, R. and Krishna, R., 1993, Multiple solutions in reactive distillation for methyl tert-butyl ether synthesis, Ind Eng Chem Res, 32: 1706±1709. 7. Mobil, 1995, Personal communication with Mr D. J. Rasmussen, customer relations, Mobil, December. 8. Novacor Personal communication with Mr Mark Kzyonsek, Engineer, Novacor-Sarnia. 9. Pecci, Giancarlo and Floris, Telemaco, 1977, Ether ups antiknock, Hydrocarbon Process, December, 98±102.

524

ELDARSI and DOUGLAS

10. Platt’ s Oilgram Price Report, October 1995±June 1996. 11. Reh® nger, A. and Hoffmann, U., 1990, Kinetics of methyl tertiary butyl ether liquid phase synthesis catalyzed by ion exchange resin-I. Intrinsic rate expression in liquid phase activities, Chem Eng Sci, 45: 1605±1617. 12. Venkataraman, S., Chan, W. K. and Boston, J. F., 1990, Reactive distillation using ASPEN PLUS, Chem Eng Prog, 96(8): 45±54.

ACKNOWLEDGEMENTS The authors would like to thank AGOCO Oil Co., Libya and Natural Science and Engineering Council (NSERC) for their ® nancial support of

this project, Mr Rasmussen from Mobil Co. for providing the formula to calculate the cost of the butylenes feed and the overhead product selling price.

ADDRESS Correspondence concerning this paper should be addressed to Professor P. L. Douglas, Department of Chemical Engineering, University of Waterloo, Ontario, Canada N2L 3G1. The manuscript was received 5 March 1997 and accepted for publication after revision 5 January 1998.

Trans IChemE, Vol 76, Part A, May 1998