Improved plantwide control structure for extractive divided-wall columns with vapor recompression

Improved plantwide control structure for extractive divided-wall columns with vapor recompression

Accepted Manuscript Title: Improved Plantwide Control Structure for Extractive Divided-Wall Columns with Vapor Recompression Author: William L. Luyben...

1MB Sizes 0 Downloads 54 Views

Accepted Manuscript Title: Improved Plantwide Control Structure for Extractive Divided-Wall Columns with Vapor Recompression Author: William L. Luyben PII: DOI: Reference:

S0263-8762(17)30275-7 http://dx.doi.org/doi:10.1016/j.cherd.2017.05.004 CHERD 2676

To appear in: Received date: Revised date: Accepted date:

1-3-2017 18-4-2017 9-5-2017

Please cite this article as: Luyben, William L., Improved Plantwide Control Structure for Extractive Divided-Wall Columns with Vapor Recompression.Chemical Engineering Research and Design http://dx.doi.org/10.1016/j.cherd.2017.05.004 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Submitted to Chemical Engineering Research and Design (CHERD-D-17-00275)

Improved Plantwide Control Structure for Extractive Divided-Wall Columns with Vapor Recompression William L. Luyben Department of Chemical Engineering Lehigh University Bethlehem, PA 18015 USA

March 1, 2017 Revised April 14, 2017

[email protected]; 610-758-4256; FAX 610-758-5057

Highlights     

An improved control structure is proposed for a divided-wall extractive column. Vapor recompression to heat an intermediate reboiler complicates dynamic control. The modified design avoids the use of a steam-heated auxiliary side reboiler. Only a portion of the overhead vapor is compressed in the proposed process. An effective control structure is developed that handles large step disturbances.

Abstract The conventional extractive distillation flowsheet for dehydration of lowconcentration bioethanol uses three-columns: a pre-concentrator, an extractive column and a solvent-recovery column. A recent paper studied the control of a process using a single divided-wall column. The complex flowsheet uses both main and intermediate reboilers. Ethanol goes overhead, water is removed as a liquid sidestream and solvent is recovered from the base. The control challenges of such a complex multivariable interacting process are obvious. The published control structure is complex and dynamically very sensitive. Only slow ramp disturbances can be handled. The purpose of this paper is to develop a more simple control scheme that is robust and is capable of handling large step disturbances. The key feature of the modified process is to only compress what is needed in the side reboiler.

Key Words Divided-wall column; ethanol dehydration; extractive distillation; plantwide control

1. Introduction Because of its extensive industrial application, there is a vast literature dealing with the separation of ethanol/water mixtures. A number of separation techniques have been explored and applied in many plants around the world. Heterogeneous azeotropic distillation was the method of choice for many years, but extractive distillation has become the principle method in recent years. The existence of the minimum-boiling homogeneous azeotrope prevents the separation from being achieved in a single conventional binary column. The conventional heterogeneous azeotropic distillation and the extractive distillation processes both use flowsheets with two columns. When the feed is very low in ethanol, which is the case in bioethanol production, a third column is conventionally used to pre-concentrate the feed to the azeotropic-separation columns. In the last decade many non-conventional flowsheets have been explored with most of the activity involving divided-wall columns. Both the heterogeneous azeotropic system (A-DWC) and the extractive system (E-DWC) have been studied in terms of optimum economic design and dynamic plantwide control. The status of the technology is summarized by Kiss1. A paper by Kiss and Ignat2 developed a novel process to replace the conventional three-column system for dehydration of bioethanol. The 10 wt% ethanol binary mixture produced in the upstream fermentation section of the plant is conventionally separated by using a pre-concentrator column to increase the ethanol concentration to about 93 wt% before feeding into a two-column extractive distillation system. The Kiss and Ignat process achieves the separation in a single vessel. We describe this process and a proposed modification in the following sections.

2. Original Design of Single-Step Column Kiss and Ignat2 compare the economics of a conventional three-column extractive distillation process with a divided-wall column using only one vessel. Total annual cost and the energy requirements are 17% lower in the DWC process.

As shown in Figure 1, the single vessel has 42 stages with a vertical wall between Stages 17 and 35. The feed is introduced at the top of the “feed side’ of the wall, which has 17 stages. None of the liquid from the upper rectifying section is fed to this side of the wall (liquid split L = 0). At the bottom of this side of the wall a partial or “side” reboiler is used and the water product is withdrawn at this location. The vapor coming up from the lower stripping section is split between the feed side (V = 0.4) and the non-feed side of the wall in which the solvent (ethylene glycol) is separated from the water. A main reboiler in the base of the column generates vapor that rises through the stripping section before being split between the two sides of the wall. The total duty of the two reboilers is 25.77 MW. The solvent-to-feed ratio is 0.1663. The reflux ratio is 3.4. This novel configuration has two reboilers, one condenser and a liquid sidestream. Ethanol goes overhead, and water goes out in the sidestream. Solvent is recovered in the base and recycled back to Stage 4 in the rectifying section. Reflux is fed to the top of the rectifying section. Fresh feed is fed to the top of the feed-side of the wall. The authors use NRTL physical properties in their Aspen simulations. The vapor is split at the top of the stripping section, but all of the liquid leaving the bottom of the rectifying section is fed to the top of non-feed side of the wall to keep solvent from entering the feed side of the wall and adversely affecting the purity of the water sidestream product.

3. Single-Step Column with Vapor Recompression The temperature at the intermediate side reboiler (105 oC with mostly water) is much lower than the temperature in the base of the column (204 oC with mostly ethylene glcol), so Lou et al suggest in a later paper3 the use of vapor recompression (heat pump) to supply heat to the side reboiler. The flowsheet was modified by installing a compressor in the overhead vapor line from the column, compressing to 3.7 bar and using the condensing hot vapor as the heat source in the side reboiler. Figure 2 gives the Aspen simulation flowsheet in which the feed-side of the wall is fictitiously modeled as a separate vessel. There are only minor differences from the original design2 with a lower solvent-to-feed ratio, fewer total stages, a smaller reflux ratio and a larger vapor split ratio.

The authors’ economic results showed a 23% increase in capital investment because of the expensive compressor, but energy cost decreased by 21%, which resulted in an 18% decrease in total annual cost (assuming a 10-year payback period). The steady-state design of this novel system is not a trivial exercise. Likewise finding an effective control structure for the multivariable interacting system is also not a simple application of conventional distillation control wisdom. Patrascu et al4 attempted to develop a control structure but found that significant process changes were required to achieve dynamically stable operation. They reported that the dynamic simulation was very sensitive (simulation would “break” in their terminology) for even small disturbances. Patrascu et al4 modified the process in two ways. They added an auxiliary steam-heated reboiler to operate in parallel with the side reboiler driven by the hot compressed process vapor. Some of the hot vapor could be bypassed around the side reboiler. These features provide an additional degree of freedom for control. They used a “valve position controller” on the bypass to keep the use of the auxiliary steam at an average of 0.83 MW. They also added a heater in parallel with the cooler in the solvent feed to the column, claiming that the control of the solvent temperature sometimes required cooling and sometimes required heating. A split-range valve arrangement was used. Figure 2 shows the published control structure. Solvent flow is ratioed to feed with the ratio changed by a distillate ethanol composition controller. The temperature on Stage 6 on the feed side of the wall is controlled by manipulating auxiliary steam to the side reboiler, with the setpoint of the temperature controller changed by a composition controller on the water sidestream purity. The temperature on Stage 33 in the stripping section is controlled by manipulating base reboiler steam. There are two key features that we will modify in the next section of this paper that we feel are weaknesses in the published process design and control structure. 1. The pressure of the column is controlled by a valve in the overhead vapor line. Not only is a very large valve required, but any pressure drop over the valve increases the power requirement in the compressor. 2. All of the overhead vapor is compressed. We will demonstrated that duty in the side reboiler requires that only a portion of the total overhead vapor needs to be

compressed. With an overhead temperature of 78 oC at 1 bar, the remaining vapor can be condensed using cooling water in an auxiliary condenser.

4. Modified Design and Control Structure The suggested process design modifications are based on a somewhat similar vapor recompression study reported by Li et al5 of a heterogeneous azeotropic distillation system using a divided-wall column. The suggested control structure is based on a paper6 that developed a control structure for the Li et al process. The column configuration is unchanged with the number of stages in each section and feed location the same as used by Patrascu et al4. The design optimization variables reflux ratio, vapor split and solvent flowrate are also the same as those used in the published paper. The predicted results are quite closed to those given in the published paper. Note that the solvent flowrate given in the original paper2 has been modified in the control paper4.

4.1 Design: Figure 3 shows the flowsheet of the modified process. The divided-wall column is modeled in Aspen by two Radfrac units, both with only a reboiler and no condenser. The feed-side of the wall is modeled as Column C1 with 14 stages. Fresh feed enters on the top stage after being preheated to 61 oC in an economizer by the hot solvent from the base of Column C2. The stripping section, the “non-feed” side of the wall and the rectifying section are modeled as Column C2 with 38 stages. The key features are compression of only a portion of the overhead vapor and use of auxiliary condensers. The vapor leaving the top of the column (54,210 kg/h) is split with 73% fed to the compressor. The hot compressor discharge vapor at 143 oC and 3.64 bar is fed to the side reboiler in which the heat duty is 9.328 MW. The vapor is condensed and leaves at 112 oC. The base temperature in C1 is 105oC and a log-mean temperature driving force is used with an overall heat-transfer coefficient U = 1.2 kW m-2 K-1 giving an area of 449 m2. A more accurate steady-state design approach would be to take into account the different heat-transfer regimes in the heat exchanger (de-

superheating and condensing, which have different heat-transfer coefficients). To keep the model simple, the constant U and log-mean temperature was used, which should have negligible dynamic effect (slightly larger area). The side reboiler is modeled in the Aspen Dynamics simulation by using Flowsheet Equations as discussed below. The liquid is cooled to 70 oC in a water-cooled trim condenser (1.548 MW) and fed through a control valve into the reflux drum operating at 0.724 bar. The rest of the overhead vapor is condensed to 70 oC in an auxiliary condenser (3.571 MW) and fed through a control valve into the reflux drum, which is adiabatic. Reflux is pumped back to the top of C2 (41,870 kg/h) and distillate ethanol product (12,346 kg/h) is removed at a purity of 99.87 wt%. The reflux ratio is 3.39. The liquid water product is withdrawn from the side reboiler (the base of C1) at 112,866 kg/h with a purity of 99.66 wt% water. The vapor leaving the stripping section of C2 is split between a vapor stream fed to the base of C1 (modeled in Aspen as a sidestream from C2) such that the fraction of vapor to C1 is 0.548 of the vapor on Stage 32. At design conditions the sidestream flowrate is 9826 kg/h with a composition 2.15 wt% ethanol, 93.42 wt% water and 4.43 wt% ethylene glycol. The overhead vapor from C1 is fed to Stage 17 of C2 at 21,960 kg/h with composition 56.13 wt% ethanol and 41.87 wt% water. The reboiler duty in C2 is 12.20 MW with a base temperature of 204 oC. The bottoms flowrate is 16,210 kg/h with a purity of 99.99 wt% ethylene glycol. A makeup stream of ethylene glycol (210 kg/h) is added to account for the very small loss of solvent in the product streams. A comparison of the published design with the modified design shows that the compressor power is reduced from 1.808 MW to 1.217 MW. On the other hand, the reboiler duty in the column base is increased from 10.04 MW to 12.20 MW. However, the published design uses addition energy in the auxiliary side reboiler. The range of this additional heat is given4 as 5% of the combined reboiler duties, which would amount to 1.103 MW.

4.2 Control Structure

Figure 4 shows the initial regulatory base-level control structure tested for the modified process after exporting to Aspen Dynamics as a pressure-driven simulation. We first want to see if good control can be attained without the use of on-line composition measurement. The temperature profiles in the two sections are shown in Figure 5. A temperature in the feed side of the wall on Stage 21 (Stage 4 in fictitious column C1) is controlled by manipulating compressor power, and a temperature in the stripping section on Stage 33 is controlled by manipulating the heat input to the reboiler in the base of the column. Temperature loops have 1-minute deadtime. Level controllers are proportional with KC = 2 (ranges of the level transmitters and control valves are twice the design values). The control loops are enumerated below; 1. Fresh feed is flow controlled. 2. Solvent flowrate is ratioed to fresh feed flowrate. 3. Stage 33 temperature in C2 is controlled at 125 oC by manipulating reboiler duty. 4. Stage 4 temperature in C1 is controlled at 99.9 oC by manipulating compressor power. 5. Base level in C1 is controlled by manipulating bottoms flowrate B1. In the physical system, this is the level in the side reboiler and the flowrate of the water sidestream product. 6. The level in the reflux drum is controlled by manipulating the reflux flowrate R2 in Column C2 in accordance with “Richardson’s Rule” for high reflux-ratio columns. 7. The distillate flowrate is ratioed to the reflux flowrate. 8. The C2 column pressure is controlled by the valve in the auxiliary condenser line. 9. The pressure in the high-pressure side of the side reboiler (the compressor discharge pressure) is controlled by the control valve in the line after the trim condenser. Note that the pressure in the reflux drum is not controlled 10. The temperatures of the streams leaving the trim and auxiliary condensers are controlled at 70 oC by manipulating cooling water flowrates. 11. The temperature of the solvent entering the column is controlled at 50 oC by manipulating cooling water in the cooler. 12. Base level in C2 is controlled by the makeup of fresh solvent.

Note that the pressure in C1 is not controlled. It floats on the pressure in C2. The small compressor shown on the gas stream leaving Stage 31 in C2 and entering the bottom of C1 is a fictitious unit whose function is to keep the vapor split constant during the dynamic simulations. The position of the wall is set at design, so the fraction of the total vapor leaving the stripping section and going to the feed side of the wall is held constant by a fictitious flow controller whose setpoint is adjusted by the output of a multiplier block with an input that is the vapor flowrate on Stage 32.

4.3 Aspen Dynamics Simulation Issues The method used to model the side reboiler involves the use of Aspen Dynamics Flowsheet Equations. The stream leaving the compressor is fed into a fictitious heat exchanger block that is an Aspen Heater model labeled “SIDEREB” as shown in Figure 6. The heat transfer rate in calculated as discussed earlier in the description of the side reboiler. The base temperature of C1 (105 oC) is the cold sink and a log-mean temperature differential driving force using the compressor discharge temperature (143 o

C) at the hot end of the heat exchanger and the bubblepoint temperature of the distillate

product (112 oC at 3.64 bar) at the cold end. Duty is calculated knowing the heat-transfer area (449 m2) and U = 1.2 kW m-2 K-1. Heat is transfer out of this “SIDEREB” block, so it is a negative number using Aspen notation. Then the reboiler duty in C1 is set equal to the negative of the duty in the fictitious “SIDEREB” block. Figure 7 gives the Flowsheet Equations used in the simulation. Note the convenient use of intermediate temperature variables “dt1”, “dt2” and “tlog” by declaring them AS Temperature. Note also the natural log function loge. The heat-transfer rates in these calculations must have units of GJ/h to satisfy Aspen conventions.

4.4 Dynamic Performance The modified process with its control structure exhibits very robust dynamics. No tendency to “break” occurred. The two temperature controllers can be put on manual and

the process remains at the steady state. The control structure displays no fragility and can handle large 20% step changes in throughput. With the two-temperature control structure shown in Figure 4, both products purities remain close to their specification for a 20% increase in feed flowrate. However, the ethanol purity drops to 95 wtl% when the feed flowrate is decreased by 20%. Installing a composition controller on the ethanol product cascaded to temperature took care of this problem, but a disturbance in feed composition (ethanol increased from 10 to 12 wt%) caused the water product to go off-specification. These results are in line with those of Patrascu et al4 who found that dual composition control is required. As shown in Figure 2 these authors use the following composition loops: 1. Composition controller on ethanol distillate purity changes the solvent-to-feed ratio. 2. Composition controller on water purity changes the setpoint of the temperature controller on Stage 6 in Column C1. The temperature controller output manipulates the auxiliary steam to the side reboiler. In the modified process there is no auxiliary steam in the side reboiler, but the available manipulated variable is compressor power. Two alternative dual-composition control structures are studied in this paper.

CS1: 1. A composition controller on the D2 ethanol distillate purity changes the setpoint of the temperature controller on Stage 4 in Column C1. The temperature controller output manipulates the compressor power. 2. A composition controller on the B1 water purity changes the solvent-to-feed ratio.

CS2: 1. A composition controller on the D2 ethanol distillate purity changes the solventto-feed ratio. 2. A composition controller on the B1 water purity changes the setpoint of the temperature controller on Stage 4 in Column C1. The temperature controller output manipulates the compressor power.

Table 1 gives the controller tuning constants and other parameters for the two alternative control structures. The composition loops have 3-minute deadtimes and are tuned sequentially (the loop with compressor power tuned first) using relay-feedback testing and Tyreus-Luyben parameters. The gains in the composition loops that change the setpoint of the Stage 4 temperature controller in C1 (manipulating compressor power) are detuned slightly because of the large sensitivity of the system to changes in compressor power. Following distillation control wisdom, both composition controllers measure and control the impurity not the purity because of the high product purities. The ethanol product impurity is essentially only water. However, in the water product both ethanol and ethylene glycol are impurities as shown in Figure 3 (0.15 wt% water and 0.19 wt% ethylene glycol). Notice in Table 1 that the solvent-to-feed loop is slower than the other loop with large integral times. In the CS1 structure I is 92 minutes while in the CS2 structure it is 63 minutes. As the results given below demonstrate, the faster CS2 structure gives better performance.

CS1: Figure 8 shows the CS1 control structure. Figure 9 gives the dynamic responses to 20% step changes in the setpoint of the feed flow controller at 0.5 hours. The solid lines are for a 20% step increase, and the dashed lines are for a 20% step decrease. The two product purities xD2(E) and xB1(W) are well controlled. Note that the ethanol composition controller changes the setpoint of the temperature controller on Stage 4, so T4 is not controlled at a constant value in this composition/temperature cascade structure. The S/F ratio produces an immediate change in the solvent flowrate, and the water composition controller then adjusts the ratio to drive the water composition to its specification. As expected, larger feed flowrates require larger compressor power, larger reboiler duties and larger solvent and reflux flowrates. Figure 10 gives the dynamic responses to changes in feed composition. The solid lines are for a step increase in feed ethanol at 0.5 hours from 10 to 12 wt% with a corresponding decrease in water concentration. The dashed lines are for a step decrease in feed ethanol at 0.5 hours from 10 to 8 wt% with a corresponding increase in water concentration.

The responses are slow, taking over 10 hours. Ethanol purity is fairly well controlled, but water purity is not back to its setpoint even after 10 hours. Notice the slow ramp down in the solvent flowrate due to the very large 92 minute reset time.

CS2: Figure 11 shows the CS2 control structure and Figures 12 and 13 show the dynamic responses to step disturbances in feed flowrate and feed composition. The CS2 structure is clearly superior to the CS1 structure as revealed by comparing Figure 9 and Figure 12 for feed rate disturbances and by comparing Figure 10 and Figure 13 for feed composition disturbances. Figure 14 provides a direct comparison of the product purities for the two structures when these large step disturbances enter process.

5. Conclusion A significant modification of the bioethanol dehydration extractive distillation divided-wall column with vapor recompression is demonstrated to greatly improve the dynamics of the system by removing the sensitivity and fragility of the process. Compression of only a portion of the overhead vapor from the column and the use of auxiliary condensers improve the dynamic robustness of the system. An effective plantwide control structure is developed and tested for large step disturbances in feed flowrate and composition.

References 1. Kiss, A. A. Advanced Distillation Technology – Design, Control and Applications (2013), Wiley 2. Kiss, A. A., Ignat, R. M. “Innovative single step bioethanol dehydration in an extractive dividing-wall column” Sep.Pur. Tech. 98 (2012) 290-297. 3. Luo, H., Bildea, C. S., Kiss, A. A. “Novel heat-pump-assisted extractive distillation for bioethanol purification” Ind. Eng. Chem. Res. 54 (2015) 22082213. 4. Patrascu, I., Bildea, C. S., Kiss, A. A. “Dynamics and control of a heat-pumpassisted extractive dividing-wall column for bioethanol dehydration” Chem. Eng. Res. Des. 119 (2017) 66-74. 5. Li, R., Ye, Q., Suo, X., Dai, X., yu, H., Feng, S., Xia, H. “Improving the performance of heat-pump-assisted azeotropic dividing-wall distillation” Ind. Eng.Chem. Res. 55 (2016) 6454-6464. 6. Luyben, W. L. “Control of an azeotropic DWC with vapor recompression” Chem. Eng. Proc. 109 (2016) 114-124.

Table 1 – Composition Controller Parameters CS1 CS2 xB1

xD2

Manipulate

S/F

SP TC4

KC I (minutes)

0.050 92

0.028 26

Manipulate KC I (minutes)

SP TC4 0.20 33

S/F 0.7 63

(1) Composition of ethanol impurity controlled; transmitter span 0.005 wt. fraction (2) Composition of water impurity controlled; transmitter span 0.005 wt. fraction (3) Controller output span for S/F = 0.267 (4) Controller output span for SPTC4 = 100 oC

Figure Captions Figure 1 – Original DWC design with side reboiler Figure 2 – Published control structure with VRC and auxiliary side reboiler Figure 3 – Modified process flowsheet Figure 4 – New control structure Figure 5 – Temperature profiles Figure 6 – Aspen Dynamics PFD Figure 7 – Flowsheet equations for heat integration Figure 8 – Dual composition control structure CS1 Figure 9 – 20% feed flowrate CS1 Figure 10 – Feed composition disturbances CS1 Figure 11 – Dual composition control structure CS2 Figure 12 – 20% feed flowrate CS2 Figure 13 – Feed composition disturbances CS3 Figure 14 – Comparison of CS1 and CS2

Figure

Fig. 1 – Original DWC Design with Side Reboiler 1 bar

Solvent 20,793 kg/h

1

0.0001 W 0.9999 EG

4

RR = 3.4

17

Feed

1

18

17

34

2.964 MW

L = 0

125,000 kg/h 0.1 E 0.9 W

Sidestream 112,500 kg/h 0.002 E 0.998 W

35

Distillate V = 0.4

12,500 kg/h 0.998 E 0.002 W

41

Solvent Recycle

Bottom + Side Reboilers 25.77 MW

Fig. 2 – Published Control Structure With VRC and Auxiliary Side Reboiler PC

1 bar

1

Solvent 16,737 kg/h

4

TC

CC RR = 3.2

1.808 MW

17

PC 1

3.7 bar

Feed

C2

C1 SP

TC S/F SP

FC

FC

SP

VPC 31

FT

6

 V = 0.548

13 R/D

FC 33

Feed

TC

TC 1.10 MW

37

10.9 MW

LC LC

FT

LC

CC Makeup

10.04 MW

Water

Ethanol

Fig. 3 – Modified Process Flowsheet 1.01 bar; 78 oC

14,630 kg/h 0.9987 E 0.0014 W 41,870 kg/h 70 oC

Solvent 16,420 kg/h 0.0001 W 0.9999 EG 50

1 4

1.217 MW

oC

21,960 kg/h; 93 oC 0.5613 E 0.4187 W

17

39,580 kg/h 3.64 bar 143 oC

1 Feed

C2

Auxiliary Condenser

C1

79 oC

61 oC

3.571 MW 70 oC

Feed

31

125,000 kg/h 50 oC 0.10 E 0.90 W

9826 kg/h; 111 oC 0.0215 E 0.9342 W 0.0443 EG

13 105 oC

Trim Condenser

1.548 MW

9.328 MW 37

70 oC

112 oC

12.20 MW 0.724 bar 70 oC

1.23 bar 204 oC

Water B1

Makeup 210 kg/h

Solvent B2 16,210 kg/h; 0.0001 W; 0.9999 EG

Base Rebolier

112,866 kg/h 0.0015 E 0.9966 W 0.0019 EG

Distillate D2 12,346 kg/h 0.9987 E 0.0014 W

Fig. 4 – New Control Structure

1

Solvent SP

S/F

4

FC

PC 17

TC

1

C2

C1 4 TC TC

FC

VS1/V32

31

FT

FT

SP

13

FC TC

Feed 33

TC

LC D/R

37 PC

Makeup

SP

LC FC

LC

Water B1

Ethanol D2

Fig. 5 – Temperature Profiles

C2

C1 (Feed Side of Wall)

Fig. 6 – Aspen Dynamics PFD

QR1

QSIDEREB

Fig. 7 – Flowsheet Equations for Heat Integration

Fig. 8 – Dual Composition Control Structure CS1

1 SP

S/F

CC

4

FC

PC 17

TC

1

C2

C1 4 SP

FC

TC TC

VS1/V32

31

FT

FT

SP

13

FC TC 33

TC

LC D/R

37 PC

SP

LC FC

LC

CC

Fig. 9 – 20% Feed Flowrate CS1

Fig. 9 – Continued

Fig. 10 – Feed Composition Disturbances CS1

Fig. 10 – Continued

Fig. 11 – Dual Composition Control Structure CS2

CC

1 SP

S/F

4

FC

PC

17

TC

1

C2

C1 4 TC TC

SP

FC

VS1/V32

31

FT

FT

SP

13

FC TC 33

TC

LC D/R

37 PC

SP

LC FC

LC

CC

Fig. 12 – 20% Feed Flowrate CS2

Fig. 12 – continued

Fig. 13 – Feed Composition Disturbances CS2

Fig. 13 – continued

Fig. 14 – Comparison of CS1 and CS2

CS1

CS2

CS2

CS1

CS1