Ind. Eng. Chem. Res.

1988,27, 2293-2304 2293

Guillot, J. “Computer Simulation of Emulsion Processes for Mono- Approach”. J. Coat. Technol. 1986,58,49-60.
mers of Different Water Solubility”. In Polymer Reaction En- Poehlein, G. W.; Dubner, W.; Lee, H. C. “Steady-State Analysis of
gineering; Reichert, K., Ed.; Huthig and Wepf Verlag: Heidel- Emulsion Polymerization in a Seeded, Continuous Stirred-Tank
berg, Fed. Rep. of Germany, 1986. Reactor”. Br. Polym. J . 1982,Dec, 143-152.
Harris, B.; Hamielec, A. E.; Marten, L. Emulsion Polymerization Riddle, E. H. Monomeric Acrylic Esters; Reinhard New York, 1954.
Kinetics, Chain Entanglements and Glassy State Transition; Rudin, A.; Devon, M. J. “Monomer Chain Transfer in the Co-
Hamielec, A. E., Bassett, D., Eds.; ACS Symposium Series 165; polymerization of Styrene and Butyl Acrylate”. J. Polym. Sci.:
American Chemical Society: Washington, DC, 1981;pp 315-326. Part A 1986,24, 2191-2198.
Lee, H. C. “Emulsion Polymerization In A Seed-Fed Continuous Rudin, A.; Goldwasser, J. M. “Emulsion Copolymerization of Styrene
Stirred-Tank Reactor”. Ph.D. Dissertation, School of Chemical and Methyl Methacrylate”. J . Polym. Sci.: Polym. Chem. Ed.
Engineering, Georgia Institute of Technology, Atlanta, 1985. 1982,20,1993-2006.
Lenz, R. W. Organic Chemistry of Synthetic High Polymers; In- Rudin, A,; Samanta, M. C. “Monomer Chain Transfer Constants
terscience: New York, 1970. from Emulsion Copolymerization Data”. J. Polym. Sci.: Polym.
Litt, M.; Patsiga, R.; Stanhett, V. “Emulsion Polymerization of Vinyl Chem. Ed. 1979,17,493-502.
Acetate. 11”. J. Polym. Sci.: Part A-1 1970,8, 3607-3649. Schuller, H. “Copolymerization In Emulsion”. In Polymer Reaction
Matheson, M. S.;Ayer, E. E.; Bevilacqua, E. B.; Hart, E. J. “Rate Engineering; Reichert, K., Ed.; Huthig and Wepf Verlag: Hei-
Constants in Free Radical Polymerizations. IV. Methyl delberg, Federal Republic of Germany, 1986.
Acrylate”. J . Am. Chem. SOC.1951,73,5395-5400. Smith, W. V.; Ewart, R. H.“Kinetics of Emulsion Polymerization”.
Mead, R. N.; Poehlein, G. W. “Free Radical Transport From Latex J. Chem. Phys. 1948,16,592-599.
Particles”, submitted for publication in J. Appl. Polym. Sci. 1988. Stockmayer, W. H. “Note on the Kinetics of Emulsion
Min, K. W.; Ray, W. H. “The Computer Simulation of Batch Polymerization”. J . Polym. Sci. 1957,24, 314-317.
Emulsion Polymerization Reactors Through a Detailed Mathe- Sundberg, D. H.; Hsieh, J.; Soh, S. “Diffusion-Controlled Kinetics
matical Model”. J. Appl. Polym. Sci. 1978,22, 89-112. in the Emulsion Polymerization of Styrene and Methyl Meth-
Nomura, M.; Kubo, M.; Fujita, K. “Kinetics of Emulsion Co- acrylate”. In Emulsion Polymers and Emulsion Polymerization;
polymerization 111”. J. Appl. Polym. Sci. 1983,28, 2767-2776. ACS Symposium Series 165;American Chemical Society: Wash-
Nomura, M.; Yamamoto, K.; Horie, I.; Fujita, K. “Kinetics of ington, DC, 1981;pp 315-326.
Emulsion Copolymerization 11”. J . Appl. Polym. Sci. 1982,27, Ugelstad, J.; Hansen, F. K. “Kinetics and Mechanisms of Emulsion
2483-2501. Polymerization”. Rubber Chem. Technol. 1976,49, 536-609.
O’Toole, J. T. “Kinetics of Emulsion Polymerization”. J. Appl. Wilke, C. R.; Chang, P. “Correlation of Diffusion Coefficients in
Polym. Sci. 1965,9,1291-1297. Dilute Solutions”. AIChE J. 1955,264-270.
Odian, G . Principles of Polymerization; Wiley: New York, 1981.
Pendlis, A.; MacGregor, J. F.; Hamielec, A. E. “Mathematical Mod- Received for review March 14,1988
eling of Emulsion Polymerization Reactors: A Population Balance Accepted August 22, 1988

PROCESS ENGINEERING AND DESIGN

Design and Control of Condensate-Throttling Reboilers

Heleni S. Papastathopoulou a n d William L. Luyben*
Department of Chemical Engineering, Lehigh University, Bethlehem, Pennsylvania 18015

Two reboiler designs, vapor throttling and condensate throttling, were studied to (1) quantify the
conditions under which the condensate-throttling scheme is superior in terms of smaller area re-
quirements and (2) compare the dynamic performance of the two alternatives when used on a
distillation column. A design procedure which incorporates the process requirements at the maximum
and minimum specifications is proposed. The following factors that affect the design were explored:
heating medium, pressure level, turndown requirements, and heat-transfer coefficient. The con-
densate-throttling design saves heat-transfer area when the turndown ratio is quite large or when
the pressure level of the heating medium approaches the critical pressure. The dynamic performance
of the condensate-throttling reboiler is shown to be inferior to the vapor-throttling reboiler. Dis-
tillation column control is significantly degraded if the distillation time constant is less than about
4 times the reboiler time constant. Several control system changes and process modifications are
presented that improve the dynamics of the condensate-throttling design.

Introduction condensate throttling) affect the design and control of

The main energy input to a distillation column is sup- distillation columns. Different energy input sources (steam
plied either by steam at an appropriate pressure or by some at several pressures and several hydrocarbons) were con-
process vapor in a heat-integrated scheme. In both cases, sidered so that the results are also applicable for heat-in-
it is necessary to tightly control the energy input to the tegrated systems.
distillation column. A. Description of the Two Alternative Systems. 1.
The purpose of this paper is to investigate how two Vapor-Throttling Design. In the vapor-throttling
different reboiler configurations (vapor throttling and scheme, the control valve is placed in the vapor supply line
0888-5885/88/2627-2293$01.50/0 0 1988 American Chemical Socidty
2294 Ind. Eng. Chem. Res., Vol. 27, No. 12, 1988

I &

Figure 2. Condensate-throttling scheme.

Figure 1. Vapor-throttling scheme.
vapor. The higher the liquid level, the lower the heat-
(Figure 1). A trap or condensate tank must be provided. transfer rate. The condensate leaves with some subcooling.
A pump may also be required to get the condensate into Since full supply pressure is available in the reboiler, the
a storage tank. condensate-throttling design requires less heat-transfer
The rate of heat transfer is varied by changing the area than the vapor-throttling design. The control valve
pressure of the vapor in the reboiler. At the maximum is much smaller because it handles liquid. The capital cost
heat input, the control valve is fully open. Under these of the condensate-throttling design is lower than that of
conditions, the pressure drop across the valve is at its the vapor-throttling design.
minimum. A t any other heat input, the control valve is In the condensate-throttling scheme, depending on the
partially closed and the pressure drop is bigger. relative size of the process time constant and the reboiler
Because of the pressure drop across the valve, the sat- time constant, the reboiler dynamics may or may not be
uration temperature of the heating medium downstream important to the overall system performance. Any change
of the valve is lower than that at the supply pressure. This in the process requirements will cause a change in the
means that part of the driving force potential for heat opening of the condensate valve. This will affect gradually
transfer is lost. To compensate for this and to satisfy the the height of the condensate in the reboiler and thus the
energy input requirements, more heat-transfer area in the fraction of the heat-transfer area that is available for vapor
reboiler must be provided, and this increases the capital condensation. This finally changes the heat-transfer rate.
cost of the equipment. A large control valve is required. But these changes take time to occur. This is what makes
In the vapor-throttling design, the reboiler shell exhibits the dynamics of the condensate-throttling reboiler design
almost instantaneous dynamics. The big volumetric flow significant in some systems.
rate of the heating medium vapor inside the shell results In this case, the cascade system will not filter out the
in a very small residence time. Since the design is based effect of variations in the supply pressure of the heating
on the assumption that the chest is drained under all medium. This is one of the dynamic disadvantages of the
conditions, the entire heat-transfer area is available for condensate-throttling system. For the sake of making
vapor condensation. Therefore, a change in the heat-input consistent comparisons, the same controller type (PI) and
requirements of the process can almost instantaneously tuning procedure are employed in this design as were used
be satisfied. in the vapor-throttling scheme. Use of a PID controller
The cascade control scheme (Buckley et al., 1985) was is discussed later in the paper.
chosen because it compensates for the heating medium In both the vapor-throttling and the condensate-throt-
disturbances coming into the system. In both the primary tling designs, a vertical thermosyphon reboiler was used
and the secondary loops, a PI controller was implemented. with the heating medium on the shell side. The results
The tuning procedure was that of the maximum closed- of this study are, however, not limited to thermosyphon
loop log-modulus specification (Luyben, 1973). The set- reboilers. They apply to kettle and stab-in (internal) re-
tings of the vapor flow controller are typical of most flow boilers as well, since the only thing that is assumed is a
controllers: a wide proportional band and a fast reset time constant overall heat-transfer coefficient.
(Shinskey, 1979).
2. Condensate-Throttling Design. In the conden- Steady-State Design
sate-throttling design, the control valve is placed in the The steady-state designs of the two systems used the
liquid outlet line (Figure 2). No trap is required in this same heat-input requirements and the same heating me-
case. The pressure in the reboiler is always equal to that dium.
of the supply pressure. The heat-transfer rate is varied B. Design Procedure. 1. Vapor-Throttling Design.
by flooding the reboiler shell with condensate. A design procedure was developed that is not based on
The vapor flow is equal to the rate of condensation. The rules of thumb but on the basis that the maximum and
maximum heat flow will occur when the vapor chest is minimum energy input requirements of the process must
completely drained. If the condensate level is lowered too be attainable. The problem is to design both the control
far, vapor will begin blowing through the valve. This valve and the reboiler. This problem is similar to the
condition can occur when the valve can carry more con- simultaneous design of the control valve and the pump in
densate than the reboiler can condense. The flow at which a typical liquid flow loop. A perceptive discussion of the
this limit is reached will decrease as the heat-transfer area latter problem was recently presented by J. R. Connell
fouls (Shinskey, 1984). If a low-level override is used, the (1987).
system will be protected against such instances. At the maximum conditions, the control valve is wide
As the condensate outflow is reduced, the level of con- open. At the minimum conditions, the valve is designed
densate in the shell will rise, exposing less surface to the to be no less than 10% open, so as to prevent mechanical
 Ind. Eng. Chem. Res., Vol. 27, No. 12, 1988 2295

flow instabilities. The heat-transfer equations are com- the valve flow coefficient, the shell-side pressure, tem-
bined with the control valve design equations to form the perature, flow rate, and liquid enthalpy at the maximum
steady-state model of the system. and minimum conditions.
At the maximum energy input condition, the heat- First a value for the pressure inside the reboiler at the
transfer equation is maximum conditions was guessed (Papastathopoulou,
1987). Equations 1-3 were used to get a value for the valve
flow coefficient (evl).Equations 7-9 yielded another value
The vapor-side energy balance is for the valve flow coefficient (evz).If CV1and Cv2were not
equal, another value for P,, was guessed.
Qmax = F m a x ( H v - hmax) (2) There is an upper limit and a lower limit in the values
used for the reboiler pressure. P,, must be less than the
The control valve design equation is supply pressure (P,) because of the pressure drop across
Fmax = f(xmax)Cv(Pv - Pma)1'2 (3) the valve. The lower limit is imposed by heat-transfer
limitations. The shell-side (heating medium) temperature
where f(x,,) = 1. must be higher than the tube-side (process fluid) tem-
The following thermodynamic relations are also used: perature. For the solution, the initial guess for P,, was
some value near the upper limit. Then Pmax was changed
hmax = h(Pm,) (4) by using the bisection method.
H, = H(P,,T,) (5) 2. Condensate-Throttling Design. The design was
again based on the fact that the maximum and minimum
Tmax TPmax) (6) energy input requirements of the process must be attain-
able. For all operating conditions, the pressure down-
At the minimum energy input condition, the equations stream from the condensate valve was considered constant
are and specified by the process requirements. At the maxi-
Qmin = UminA(Tmin - TB) (7) mum conditions, the control valve was wide open.
At the maximum energy input condition, the equations
Qmin = F m i n ( H v - hmin) (8) describing the system are
Fmin = f ( x m i n ) C v ( P v - Pmin)"' (9) Qmax = UmaxA(Tv - TB) (12)

hmin = h(Pmin) (10) Qmax = Fmax(Hv - hv) (13)

Pmin = P(Tmin) (11) Fmax = Cvf(xmax)(f'v + AI' - (14)

where f(xmin)= 0.1, and the symbols in eq 1-11 are as h, = h(P,,T,) (15)
follows: A is the heat-transfer area, Cv is the valve flow where f(x,) = 1, h, is the enthalpy of the saturated liquid
coefficient, Fm, and Fmhare the flow rates of the heating heating medium at the supply conditions, Poutis the
medium at the maximum/minimum conditions, f(x,) and pressure downstream from the control valve, and A Z' is the
f(xmin)are the valve opening fractions at the maximum/ hydrostatic head of the condensate in the reboiler shell.
minimum conditions, H, is the enthalpy of the vapor It is assumed in eq 12 and 13 that the chest is drained
heating medium at the supply conditions, h,, and hmin at the maximum conditions, so AP is zero.
are the enthalpies of the saturated liquid heating medium The more accurate relations from the Masoneilan
inside the reboiler at the maximum/minimum conditions, Handbook (1977) were used for flow of liquids through the
P, is the supply pressure of the heating medium, P,, and control valve.
Pmin are the pressures of the heating medium inside the The system of eq 12-15 can be solved sequentially
reboiler at the maximum/minimum conditions, Q, and (without requiring any iteration) to find the area, the valve
Q- are the maximum/minimum heat-input requirements size, the flow rate, and the liquid enthalpy at the maximum
of the process, TB is the saturation temperature of the conditions.
process fluid, T, and Tminare the saturation tempera- At the design and the minimum energy input conditions,
tures of the heating medium inside the reboiler at the the reboiler will be partially flooded. For these two cases,
maximum/minimum conditions, Tvis the temperature of the characteristics of the system are determined by solving
the heating medium at the supply conditions, and U,, the following system of equations at the proper conditions:
and Uminare the heat-transfer coefficients at the maxi-
mum/minimum conditions. total heat balance
Actually the more accurate relations for vapor flow
through a valve, taken from the Masoneilan Handbook for Q = Qi + Qz (16)
Control Valves Sizing (1977), were used in place of eq 3 where Q is the total heat input to the distillation column
and 9. These take into account the critical flow situation. at the design or the minimum conditions, Q1 is the heat
The system of eq 1-11 is based on the assumptions of transferred in the vapor condensation region, and Qz is the
constant heat-transfer coefficient, isothermal boiling in the heat transferred in the liquid subcooling region.
tube side of the reboiler, isenthalpic flashing in the control The total heat-transfer area ( A )in the reboiler has been
valve, and isobaric condensation in the shell side of the determined from the maximum conditions, and
reboiler.
The known parameters were U, Umh,Q,, Qmh, and
A = A1 + A2 (17)
the supply pressure and temperature of the heating me- where Ai is the area available for vapor condensation and
dium. The overall heat-transfer coefficients were selected Az is the area available for liquid subcooling.
from the literature and from actual plant data from several The heat-transfer equation in the vapor condensation
industrial sources. By use of an iterative procedure, the region is
system of 10 equations (eq 1-4 and eq 6-11) was solved
numerically for the 10 unknowns: the heat-transfer area, Q1 = UiAi(T, - TB) (18)
2296 Ind. Eng. Chem. Res., Vol. 27, No. 12, 1988

Table I. Design Seecifications Table 11. Vapor- and Condensate-ThrottlingDesign for

Process Data Steam and Heptane
process temp, "C 100 steam heptane
max heat input, GJ/h 47.0 VTQ CT" VT CT
design heat input, GJ/h 39.1
min heat input, GJ/h 19.6 A , m2 323.7 319.8 327.4 319.0
vapor heat-transfer coeff, kJ/(h m2 "C) 3066.1 cv 2050.0 50.0 5240.0 430.0
liquid heat-transfer coeff, kJ/(h m2 "C) 613.2 F,, kg/ h 22 150.0 22 180.0 165 800.0 168400.0
FdeSlgn, kg/h 18290.0 17 800.0 125 800.0 133900.0
Supply Conditions of the Heating Medium F,, kg/h 8754.0 8467.0 51 850.0 56 600.0
medium steam heptane T, "C 147.3 147.9 146.8 148.0
supply temp, "C 147.9 148.0 Tdeelgn,"c 139.4 128.5 139.0 141.6
supply pressure, atm 4.4 3.5 T-, "C 119.7 101.6 119.5 116.0
P, atm 4.33 4.4 3.41 3.5
The heat-transfer equation in the liquid subcooling re- PdWlgn, atm 3.54 4.4 2.86 3.5
gion is Pmm, atm 1.96 4.4 1.78 3.5
f(Xm-1 1.00 1.00 1.00 1.00
Q2 = u2A2ATh (19) f(xdeSlgn) 0.26 0.61 0.32 0.72
f(Xmn) 0.10 0.25 0.10 0.26
where
* VT = vapor throttling; CT = condensate throttling.
(20) of 48 "C is available as the driving force for heat transfer.
The turndown ratio is defined as the ratio of the heat-
input requirements at the maximum conditions to the
where Toutis the temperature of the heating medium at heat-input requirements at the minimum conditions. The
the exit of the reboiler. base-case turndown ratio was 1.2/0.5 = 2.4 since Q, was
The energy balance in the condensation region is 20% greater than Qdesign and Qmin was 50% of the Qdesign.
Note (Table 11) that there is only a slight decrease in
Qi = F i W , - h,) (21) the required area when changing the design from steam
The energy balance in the liquid subcooling region is to condensate throttling because the heat-transfer area is
primarily determined by the maximum conditions. In the
Q2 = F z C P ( T -
~ Tout) (22) steam-throttling design, at the maximum specifications,
where C p is the specific heat of the liquid heating medium. the pressure drop across the valve is only 4.4-4.33 = 0.07
At steady state, the shell-side mass balance yields atm which causes a 147.9-147.3 = 0.6 "C loss in the satu-
ration temperature of the heating medium. This small loss
F1 = Fz (23) does not result in big savings in the heat-transfer area:
Equations 16-23 can be solved iteratively for the values 323.73/319.8 = 1.01 (1% area savings). The control valve
of the unknowns Q1, Q2, Al, A2, Tout,Fl, and F2. The is much bigger in the vapor-throttling design.
solution has physical meaning only if it also satisfies the In Table 11, the vapor- and condensate-throttling designs
following relations: for heptane are also presented. The supply pressure was
3.5 atm (Table I) and was chosen so that the same 48 "C
Qi > 0 (24) temperature difference was available for heat transfer. The
Q2 >0 (25) savings in the heat-transfer area because of the conden-
sate-throttling reboiler design was only 3% (327.4/319.0
Tout > TB (26) = 1.03).
A value for the heating medium temperature leaving the 2. Effect of the Heating Medium. Six different
reboiler (Tout)was guessed (Papastathopoulou, 1987). heating media were investigated steam, heptane, cyclo-
Equations 16-22 were used to calculate F1 and F2. If Fl hexane, methanol, isobutane, and benzene. One of the
was not equal to F,, the procedure was repeated assuming reasons for exploring different heating media was to see
another value for To,,. When the iterations converged, the if the reduction in area of condensate-throttling design over
percent valve openings at the design and at the minimum vapor-throttling design was increased as the slope of the
heat-input conditions were determined by using the con- vapor pressure curve increased. We found that this was
trol-valve sizing equation. not the case. Even for the best feasible case, the savings
Note that in the condensate-throttling design Pout is in the heat-transfer area was less than 8%. The larger
fixed and the percent valve opening is the variable. To savings occur when the reboiler is operating at higher
avoid flow instability problems with the valve at the pressures (where dT/dP is smaller) and for the heating
minimum conditions, the valve should be at least 10% medium that has the lowest critical conditions (in this case
open. The pressure downstream from the condensate valve heptane). Isobutane was tested at supply pressures near
can be adjusted, if necessary, to keep f(x& greater than its critical so that this trend could be checked.
10%* The significant parameter seems to be the reduced
C. Results. A methanol/water distillation column was pressure. As shown in Figure 3, the higher the reduced
used as a typical industrial example. The flow rates inside pressure the bigger the area savings become. For all the
the column were such that a 10-ft column diameter was cases, the difference T, - TBwas constant, 48 "C.
required. The reboiler heat duty was 39.1 GJ/h at design 3. Effect of the Turndown Ratio. The previous
conditions. The temperature in the base of the column conclusions are valid for a turndown ratio of 2.4. The
was 100 "C. design of the vapor- and the condensate-throttling schemes
1. Base Case. The base-case steam and condensate- for other turndown ratios indicates that the condensate-
throttling design is presented in Table 11. The design throttling scheme becomes more favorable in terms of the
specifications are summarized in Table I. The steam savings in the heat-transfer area as the turndown ratio
supply pressure is 4.4 atm (the corresponding saturation becomes larger. In Figure 4, this trend is presented. The
temperature is 147.9 "C). A total temperature difference heating medium is steam.
 Ind. Eng. Chem. Res., Vol. 27, No. 12, 1988 2297

12 5 L
D steam
o Hepcane
b Benzer'le
v netnanol
0 Isobutsne
E Cyclohexane
r Table 111. Effect of the Turndown Ratio (Heating Medium:
Heotane)
________~

turndown ratio
A , m2
~

vapor throttling
1.2/0.5
327.4
condensate
throttling
1.5/0.3
464.3
1.2/0.5
319.0
1.5/0.3
398.7
C" 5240.0 2820.0 430.0 540.0
Fmam kg/h 165 800.0 193 700.0 168450.0 210 600.0
Fdesign, kg/h 125800.0 111600.0 133 900.0 126 300.0
Fmim k / h 51 850.0 28370.0 56610.0 31 310.0
2 t ,,T
, "C 146.8 141.2 148.0 148.0
Tdeslgnr O C 139.0 127.5 141.6 133.1
Tmint "C 119.5 108.3 116.0 102.0
Pmam atm 3.41 3.01 3.5 3.5
pdesign,atm 2.86 2.17 3.5 3.5
Pmin, atm 1.78 1.32 3.5 3.5
Yma, m 0.0 0.0 0.0 0.0
ydesign, 0.0 0.0 1.0 2.0
Ymin, m 0.0 0.0 3.2 4.2
f (xmin) 0.1 0.1 0.26 0.09

1 ' I ' I ' 1-1

12 5
I I I '
0
I ' I
Omax-l a n d e s i g n
0 Omin-0 5 Q d e s ~ o n
/

- sol
0 Qmax/Qmln 1 2 / 0 5
0 omax/omin 1 2/0 3
A Osax/omln i 4/0 5
1
l o o L
i 1
4

2'5 t //
I , I , l , 0 0
0 0 0 0 2 s so 7 5 100 125
0 0 2 0 4 0 6 0 8 0 100

Turndown R a t i o
FIXMINI ( % 1

Figure 4. Effect of the turndown ratio on the heat-transfer area Figure 5. Effect of the minimum valve opening on the heat-transfer
savings. area savings.

In the vapor-throttling scheme, when designing for larger pressure of the condensate tank, and it may or may not
turndown ratios, the heat-transfer area increases, whereas be possible to change it to a new level of operation. A split
the valve flow coefficient decreases (Table 111). The size range control valve system (one big and one small valve
of the valve is determined by the minimum heat-input in parallel) can be used with the condensate-throttling
specifications. Notice that a smaller valve will result in scheme to overcome this problem. A t normal operating
higher pressure drops, which implies bigger losses in the conditions, the small valve remains wide open and the flow
available temperature differences for heat transfer. This rate is adjusted by changing the opening of the large valve.
is an intrinsic characteristic of the vapor-throttling design. When the opening of the large valve drops below lo%, the
I t implies that the savings in the area because of the control of the flow rate is maintained through manipula-
condensate-throttling design become larger as the turn- tion of the opening of the small valve. The flow rate
down ratio increases. becomes zero when both valves are entirely closed.
In the condensate-throttling scheme, the design of the Figure 5 shows how the percent valve opening at the
reboiler for larger turndown ratios results in larger heat- minimum conditions affects both the vapor-throttling and
transfer areas and bigger valves (Table 111). In this design, the condensate-throttling designs. The heating medium
the area and C, are determined by the maximum heat- is steam. As the percent valve opening at the minimum
input requirements. The increased heat-transfer area conditions decreases, the savings in the heat-transfer area
makes possible the transfer of the required amount of heat reduce toward zero. When a split range control valve is
at the minimum and the design conditions, while using used, the minimum valve opening is not a limitation. In
smaller flow rates and higher liquid heights. such cases, there is less incentive for choosing the con-
There can be a rangeability problem in the conden- densate-throttling reboiler design.
sate-throttling design. Table I11 shows that for a turndown 4. An Industrial Application. Table IV gives spec-
ratio of 1.510.3 = 5 the valve opening at the minimum ifications for an actual commercial reboiler that was de-
conditions is less than lo%, which for control purposes is signed for a very large turndown ratio. The reboiler of this
undesirable. Redesign of this system for a higher pressure tower was designed for a heat-transfer rate of 55 X lo6
downstream of the condensate valve (that is, for a smaller Btu/h and a turndown ratio of 1.2710.18 = 7.1. The
pressure drop across the valve) cures this problem and heating medium was steam at a supply pressure of 2.36
gives a percent valve opening at the minimum conditions atm. The discharge pressure was 2.02 atm. The saturation
bigger than 10%. In most cases, the pressure downstream temperature of the process fluid was 110 " C . Since the
from the condensate valve is determined by the operating design was for a big turndown ratio, the condensate-
2298 Ind. Eng. Chem. Res., Vol. 27, No. 12, 1988

Table IV. Design Specifications of the Industrial

Application
Process Data
process temp, "C 110
max heat input, GJ/h 73.7
design heat input, GJ/h 58.0 reboller process
min heat input, GJ/h 10.4
vapor heat-transfer coeff, kJ/(h m2 "C) 3066.1
liquid heat-transfer coeff, kJ/(h m2 "C) 613.2 Iransmmer
transminer

Supply Conditions of the Heating Medium

medium steam
supply temp, "C 126.1 I
supply pressure, atm 2.36

Table V. Vapor- a n d Condensate-Throttling Design for

rtl
Industrial Application [Effect of f(xmin)] wniroiier 1

VT CT VT 6 1~

0.10 0.06 0.06

2070 1490 1630 Figure 6. Block diagram representation of the system.
2140 110 3600
33 520 33 810 33 720 Table VI. Transfer Functions
26 380 25 930 26 470 TIS +1
4660 5330 4670 PI controller B ( s ) = Kc-
121.6 126.1 124.7 TIS
119.2
111.6
126.1
126.1
121.6
112.1
TIS +1 r@ 1 +
PID controller B ( s ) = Kc- ___
2.06 2.36 2.26 TIS CYrDs +1
1.93 2.36 2.08 1
1.51
1.00
2.36
1.00
1.53
1.00
temp transmitter GT1 = KTlrTIS+l
0.75 0.38 0.54

throttling design resulted in significant savings (28%) of

the heat-transfer area (Table V). 1
control valve G, = K-,
The choice of the valve opening at the minimum con- (TYS + 112
ditions of the steam-throttling design has a large impact where 01 = 0.1, rT1= 0.25 min, T T ~= 0.083 min, r y = 0.167 min
on the design. The steam-throttling scheme requires a
heat-transfer area of 2070 m2 when the valve opening is Table VII. PID Controller Settings (T,, = 16 min, T~ = 1.6
10% at the minimum conditions. The area reduces to 1630 min, rB = 5.83 min)
m2when the design is for a minimum valve opening of 6%.
Z-N T n oDtimum T n
On the other hand, the size of the valve increases (C,
increases from 2140 to 3600). Thus, it may be quite ben- temp loop
KC1 0.0602 0.0965
eficial, in terms of the heat-transfer area, to use two vapor TI1 14.3 6.75
control valves (one large and one small) for large turndown flow loop
designs with vapor throttling. Kc, 0.46 10.5
Conclusions of the Steady-State Design. The key 712 1.20 1.20
point throughout the entire steady-state design is the TD 0.18 0.70
consistent comparison of the two cases on a similar basis.
Generally, the use of the condensate-throttling design is constant and the reboiler time constant could be found.
more effective when the heating medium is some hydro- A. Transfer Functions. The two control schemes,
carbon. vapor- and condensate-throttling design, can be repre-
As far as the heat-transfer area is concerned, there are sented by the same block diagram shown in Figure 6.
significant savings in the condensate-throttling design Note that the only difference between the two configura-
when the turndown ratio is big and/or the reboiler oper- tions is in the G2transfer function, the reboiler dynamics
ating pressure is close to the critical pressure of the heating transfer function.
medium. The control valve in the condensate exit pipe Transfer functions for transmitters, valves, and con-
is drastically smaller than the valve in the vapor line. trollers are given in Table VI.
The vapor-throttling design requires a condensate trap 1. Process Transfer Function. The process transfer
and sometimes a condensate pump. There is no such need function is assumed to be of the form
in the condensate-throttling design. However, there is need
for a low-level override controller to prevent vapor blowing
through the condensate valve.
where F, is the flow rate of the vapor heating medium, G1
Dynamics is the process transfer function, K pis the steady-state gain
Dynamic simulations were used to explore how the de- of the process, T,, is the temperature on a tray of the
terioration in performance of the condensate-throttling distillation column, 7d is the process dead time, and T~ is
case affects the composition control of the column and to the process time constant.
develop methods (control schemes) that improve the dy- Several combinations of the process time constant and
namic response. Different distillation columns with a dead time were examined so that the dynamics of different
range of time constants and dead times were investigated. types of columns could be approximated. Two cases for
It was hoped that a relationship between the process time the process dead time were investigated:
 Ind. Eng. Chem. Res., Vol. 27, No. 12,1988 2299

7d = 0.17, (28) Table VIII. Time Constants of the Condensate-Throttling

Reboiler
7d = 0.57, (29)
heating medium steam steam heptane
For each one of them, the process time constant was Q,GJ/h 39.1 39.1 39.1
taken to be rP = 2, 4, 8, 16, 32, and 64 min. turndown ratio 1.210.5 1.210.5 1.210.5
U, kJ/(h m2 " C ) 3066.1 5110.2 3066.1
The range of the process time constant was chosen so U, Btu/(h ft2 O F ) 150 250 150
that it represented typical industrial applications: columns 2117.2 2117.2 278.8
with fast, medium, or slow responses. 100 100 100
2. Transfer Function of the Reboiler. In the va- 147.9 147.9 148
por-throttling scheme, because of the very big flow rates, 319.8 191.9 319.0
the residence time of the heating medium in the reboiler 4,
cm 2.54 2.54 2.54
no. of tubes 822 657 819
shell was calculated to be 1 s or less. Since this value is 4.9 3.7 4.9
L, m
small compared to the other time constants of the system, D,,m 1.07 0.89 1.07
it was assumed that the shell dynamics are almost in- V , m3 2.33 1.05 2.33
stantaneous. T R , min 5.83 2.63 0.57
G, = 1 for vapor-throttling (30) gives data for several reboilers under different conditions.
where G 2 is the reboiler transfer function. Time constants can range from 1 to 10 min depending
In the condensate-throttling scheme, the proper transfer strongly on the heat-transfer coefficient.
function was found by using a simplified model of the As the heat-transfer coefficient decreases, the reboiler
system. time constant increases by almost the same factor. In
The mass balance for the liquid in the shell side of the practice, the heat-transfer coefficient often decreases with
reboiler yields time because of fouling. The subsequent change of the
p dV2/dt = Fa- F,
reboiler time constant could give control problems because
(31) it may require retuning of the controllers.
where F, is the condensate flow rate, t is the time, V2 is Changing the heating medium from steam to heptane
the volume of the condensate in the shell, and p is the results in a smaller reboiler time constant. There is a
density of the condensate. reduction by almost a factor of 10 for the same heat-
The heat transfer rate in the vapor-condensation region transfer coefficient. Heptane has a much smaller heat of
is vaporization than steam. Therefore, higher flow rates are
Q = F,AH = U,A,(T, - TB) (32) required in order to satisfy the heat-input requirements.
The residence time of the heating medium in the shell is
where AI is the heat-transfer area in the steam conden- smaller, and the reboiler time constant is smaller. From
sation region, Q is the heat input to the distillation, TBis this point of view, the condensate-throttling scheme could
the temperature of the process fluid, T, is the saturation be more favorable in a heat-integrated environment where
temperature of the heating medium at the supply condi- the heating medium is not steam. Actually, the U for
tions, VI is the heat-transfer coefficient in the steam organics is lower and thus the change in the reboiler time
condensation region, and AH is the latent heat of vapor- constant is not that big.
ization. B. Tuning of the Controller. The tuning of the
Ai = A ( 1 - V,/V) (33) cascade control scheme started from the secondary (inner)
loop. Once the settings of the secondary controller were
where V is the total volume of the reboiler shell and V , determined, the master controller could be tuned. In this
is the volume of the liquid in the reboiler shell. study, the maximum closed-loop log modulus tuning pro-
If we define cedure was used for both loops. The resonant peak was
PVAH chosen to be +2 db (db = decibel). This corresponds to
7R = (34) a magnitude ratio of 1.3 and is approximately equivalent
UiA(Tv - TB)
to a closed-loop damping coefficient of 0.4.
where 7R is the reboiler time constant, then 1. Tuning the PI Controller. The reset time was
7R dF,/dt + Fa = F , (35) determined from the Ziegler-Nichols (Z-N) settings. Then
the proportional gain was set so that the resonant peak
Laplace transforming eq 35, we get the transfer function of the closed-loop log-modulus Bode plot was +2 db.
expression in perturbation variables: 2. Tuning of the PID Controller. In order to see if
the system dynamics could be improved, a PID controller
(36) was used in the secondary loop of the condensate-throttling
scheme in some cases. The reset time (q)was set equal
Equation 36 gives the reboiler transfer function for the to the Z-N value. Initially the derivative time (7D) was
condensate-throttling scheme. also set equal to the Z-N value. The proportional gain was
The previous derivation has been based on the as- adjusted to give +2-db maximum closed-loop log modulus.
sumptions that there is no accumulation of mass in the Then TD was varied over a wide range, finding the value
vapor phase, that the dynamics of the vapor phase are very of K , at each rDthat gave L , = +2 db. The value of 7D
fast so that the energy balance is represented by an alge- that resulted in the biggest proportional gain (Kc,Table
braic equation, and that the heat-transfer occurring in the VII) was considered as the optimum (Cheung and Luyben,
condensate subcooling region is not significant (it was 1979).
calculated to be less than 10%). C. The Dynamic Performance of the Two Schemes.
Equation 34 reveals that the time constant of the re- The base case for all the following comparisons is a dis-
boiler is affected by the density of the condensate, the heat tillation column with a time constant (7,) of 16 min and
of vaporization of the heating medium, the size of the a dead time (rd)of 1.6 min. The reboiler time constant
reboiler, the heat-transfer coefficient, and the temperature for the condensate-throttling design was 5.83 min. The
difference that is available for heat transfer. Table VI11 heating medium was steam.
2300 Ind. Eng. Chem. Res., Vol. 27, No. 12, 1988

1i2.0 7
t
I

0 20 0 40 0 60 0 80 0 0 20 0 40 0 60 0 80 0
T I M E (MINI TIME ( M I N )
Figure 8. Temperature on a tray of the column. Disturbance in the
I supply pressure of the steam (T,, = 16 min, 7 d = 1.6 min, 7R = 5.83
1 min).

1 is 5.83 min and the process time constant is 16 min, a fairly

i significant deterioration in performance occurs.
In the vapor-throttling design, the steam flow rate is
always equal to the condensate flow rate (the reboiler
exhibits almost instantaneous dynamics). In the conden-
sate-throttling design, the two flow rates change in a dif-
ferent fashion during the transient period but both finally
reach the same steady-state value (the dynamics of this
reboiler are important). The condensate flow rate changes
first because the flow controller acts on the condensate
valve. But it takes time to adjust the level in the shell side
of the reboiler and thus the steam flow rate changes later.
The behavior of the system to a disturbance in the
supply pressure of the heating medium is shown in Figure
8. In the vapor-throttling scheme, the system remains
3 20 0 40 0 60 3 80 0 almost undisturbed from a change in the supply pressure
TIME (MIN) of the heating medium. The secondary loop of the cascade
configuration takes corrective actions before the disturb-
303 0' ' 7 ance is even sensed by the column. Although a cascade
control configuration also exists in the condensate-throt-
tling design, the reboiler dynamics prevent tight control
-
5 296 0
I of flow. The transition period is much longer and the
\I overshoot bigger.
-0
Y

292 0
1. Effect of the Column Characteristics. Up to now,
a certain distillation column design was considered with
1
i 7, = 16 min, 7 d = 1.6 min, and reboiler time constant of
7~ = 5.83 min.
An increase in the process dead time results in a slower
W
t \ and more sluggish response of the system. The overshoot
becomes bigger and the settling time longer. The effect
of the process dead time is the same for both reboiler
designs (vapor and condensate throttling).
The effect of the process time constant on the overall
performance of the system is presented in Figure 9. The
process dead time satisfies the relation 7 d = 0.17,. The
27601 ' ' ' ' ' ' ' I
reboiler time constant is 5.83 min.
3 20 0 40 0 60 0 80 0 When the process time constant is 8 min, the effect of
T I M E (MINI a 5.83-min reboiler time constant is very significant. The
Figure 7. Comparison of vapor (VT) and condensate (CT)throt- deterioration in the performance of the condensate-
tling. Disturbance in the feed flow to the column ( T =~ 16 min, 7 d throttling design is large. The settling time becomes about
= 1.6 min, Q = 5.83 min). 6 times longer and the overshoot more than twice that of
the corresponding quantities for the vapor-throttling
In Figure 7 the behavior of both the steam-throttling scheme. When the time constant is 32 min, the perform-
and the condensate-throttling designs is presented. The ance of the condensate-throttling scheme differs only
disturbance is a step change in feed flow coming into the slightly from that of the vapor-throttling scheme. In this
column. case (7, = 32 min, rR= 5.83), either of these schemes can
In the vapor-throttling design, the transition is handled be used effectively.
easier and the system moves quickly to the new steady The main point is that, when the reboiler time constant
state. The deviation of the column temperature from its is in the order of magnitude of the process time constant,
final value is smaller in absolute magnitude and the set- the performance of the condensate-throttling scheme is
tling time is less. Thus, when the reboiler time constant poor. But when T, becomes 4-5 times bigger than TR the
 Ind. Eng. Chem. Res., Vol. 27, No. 12, 1988 2301
112.0

-G' 108.0 If \ CT I
108.0

w H \

96.0 96.0
0. 40.0 80.0 120.0 160.0 0 20.0 40.0 60 0 80.0
TIME (MINI TIME (MINI

112.0 112.0 I I l ) I I I

7~=5.83 min

I
108.0 108.0 1A

96.0 96,O
0. 20.0 40.0 60.0 80 0 0. 20.0 40.0 60.0 80.0

TIME (MINI TIME (MIN)

Figure 10. Effect of the reboiler time constant. Disturbance in the
112.0 feed flow to the column (T,, = 16 min, Td = 1.6 min).

'7d=3.2 m'n
s
..-. 1 0 8 . 0
- 108.0

96 0
o 40 o ao c 120 o 160 o
TIME (MINI
Figure 9. Effect of the process time constant. Disturbance in the
feed flow to the column ( T R = 5.83 min).
behavior of the condensate-throttling scheme is compa-
rable to that of the vapor-throttling design.
2. Effect of the Reboiler Design. The effect of the
reboiler characteristics on the overall performance of the
-z
13
ZI
\
296 0
R
system is shown in Figure 10. When 7R = 2.63 (almost (II
Y 28a o
6 times smaller than 7 J , both reboiler designs behave very
-W
I
similarly and the dynamic performance is not a restriction U
a:
in choosing either one of them for such an application. g 280 0
A reboiler time constant of 2.63 min is calculated for the
base steady-state reboiler design case (Table VIII) when W
+
the heat-transfer coefficient is 5110 kJ/(m2 h "C) (250 u
m
272.0
Btu/(ft2 h OF)). This big heat-transfer coefficient can be z
W
achieved in practice after a reboiler cleaning and start-up. z
0
8
However, after some time of the reboiler operation, the
heat-transfer coefficient starts decreasing because of
264.0
tll
1 V.eR=5.83min
I
fouling. In Figure 11, the behavior of such a system is 256 0 I I I I I I I I I ,
presented. The same base case distillation column is used. 0 10 0 20 0 30 0 40 0 50 0 60 0
Three values for the reboiler time constant are considered. TIME (MINI
The 7R = 2.63 min describes the reboiler behavior during
the start-up period and is the value used for the tuning Figure 11. Effect of variable reboiler time constant. Disturbance
in the feed flow to the column (iP = 16 min, Td = 1.6 min).
of the controller. The 7R = 4.38 min represents the be-
havior of the reboiler after some time of operation when value of industrial applications, 3066 kJ/(m2 h " C ) (150
the heat-transfer coefficient has been reduced to the typical Btu/(ft2h OF)). The 7R = 5.83 min correspondsto an even
2302 Ind. Eng. Chem. Res., Vol. 27, No. 12, 1988

96 0 1 , I , . , , ,
1
Figure 12. Condensate-throttling scheme with auxiliary vapor-
throttling reboiler. 0 20 0 40 0 60 0 E0 0 100 C 12J 0
TIME (MINI
Figure 14. Comparison between the PI and the PID control of the
condensate-throttling design. Disturbance in the feed flow to the
/
1 ~ 16 min, T~ = 1.6 min, T R = 5.83 min).
column ( T =

I I
optimum ' 4

Figure 13. Valve position control of the condensate-throttling re-

boiler.

smaller value of the heat-transfer coefficient.

As the heat-transfer coefficient decreases and the re-
boiler time constant increases, the performance of the
system deteriorates and the response of the process be-
comes oscillatory. This implies that the system progres-
sively becomes tightly tuned, and it may be a good practice
to retune it.
D. Alternative Control Structures. In the next few 8 0
paragraphs, alternative control structures are presented
that improve the dynamic performance of the conden-
Ea
ci i I \
i
sate-throttling scheme.
1. Split Range Valve Control. A split range control
valve scheme may be used with either the vapor-throttling
or the condensate-throttling design, as has already been
discussed. It could also be used when only part of the
heat-input requirements of the distillation column is
supplied from the condensate-throttling reboiler, the rest
being provided from an auxiliary steam/vapor-throttling
* O t I -1

reboiler (Figure 12). The second control valve handles 0 I I I I I ,

the heating medium flow rate at the desired level when 0 0 8 I C 2 4 3 2 4 0

the condensate valve is saturated. DERIVATIVE TIME
2. Valve Position Control. Valve position control
Figure 16. Variation of the proportional gain of the flow PID con-
(Shinskey, 1971; Yu and Luyben, 1986) can be used to troller with the derivative time.
improve the dynamic performance of the condensate-
throttling reboiler (Figure 13). portional gain was set to give +2-db maximum closed-loop
During the transient periods, when the process is away log modulus. With PID control, the overshoot and the
from the steady state, the vapor-throttling valve controls settling time became smaller.
the system. As it moves toward the steady state, the In addition to using the Z-N values for the derivative
condensate-throttling valve starts controlling the process. time, an optimum value of T D was determined. Figure 16
Near the steady state, the vapor-throttling valve is nearly presents the dependence of the proportional gain of the
wide open and the control is maintained through the flow controller on the value of the derivative time. For
condensate-throttling valve. each TD, the proportional gain was determined by the
3. PID Control. Use of a PID controller in the flow +2-db specification.
loop improved significantly the dynamic performance of Figure 15 compares responses using Z-N and optimum
the condensate-throttling system. Figure 14 compares the TD. The improvement over the PID with Z-N settings is
temperature response when PI and PID control is imple- significant. The overshoot and the settling time have been
mented. The derivative and the integral time of the PID reduced by almost a factor of 2. Although derivative action
controller were set equal to the Z-N values. The pro- can improve the performance of the system, in the presence
 Ind. Eng. Chem. Res., Vol. 27, No. 12, 1988 2303
The condensate throttling is dynamically inferior, par-
ticularly in rejecting disturbances of the heating medium.
But if the column time constant is more than 4 times
greater than the reboiler time constant, both schemes

A I behave essentially the same.

Nomenclature
A = heat-transfer area, m2
A , = heat-transfer area in the condensate-throttling design,
m2
A, = heat-transfer area in the vapor-throttling design, m2
B(s) = controller transfer function
B1,B2 = transfer function of the controller in the primary/
secondary loop
C p = specific heat of the liquid heating medium, kJ/(kg "C)
C, = valve flow coefficient
D,= inside diameter of the reboiler shell, m
D,= outside diameter of the reboiler tubes, cm
F = flow rate of the heating medium, kg/h
Figure 17. Condensate feedback scheme. F, = condensate flow rate, kg/min
F, = flow rate of the vapor heating medium, kg/min
of noisy control signals it deteriorates rapidly. In practice, f ( x ) = fraction valve opening
the flow transmitter signals are often noisy and as a con- Gd = transfer function of the heating medium disturbance
sequence the use of derivative action may not be effective. coming into the system
E. Alternative Process Structure. During normal Gf, G, = transfer function of a feed flow rate/composition
operating conditions, the condensate-throttling reboiler is disturbance coming into the column
partially flooded. If there is need for increased heat input GT1, G T = ~ transfer function of the transmitter in the pri-
to the process, the condensate valve opening increases and mary/secondary loop
G, = transfer function of the control valve
the condensate level in the reboiler shell drops. If a large GI = process transfer function
condensate valve is used, area can be exposed quickly. So G2 = reboiler transfer function
more heat can be provided rapidly. If the need is for H = enthalpy of the vapor heating medium, kJ/kg
decreased heat input to the process, the condensate valve h = enthalpy of the liquid heating medium, kJ/kg
can be completely shut, but it still takes time to accumu- K , = proportional gain of the controller
late condensate in the reboiler shell and reduce the heat Kp = steady-state gain of the process
input. Thus, the condensate-throttling system could ex- KT1 = steady-state gain of the temperature transmitter
hibit nonlinear dynamic responses: fast increases in heat KT2 = steady-state gain of the flow transmitter
input but slow decreases in heat input. K, = steady-state gain of the valve
The condensate feedback strategy, shown in Figure 17, L = total height of the reboiler, m
might be useful in some processes to permit more rapid P = pressure of the heating medium, atm
covering of the tubes. The condensate level rises faster Q = desired heat input to the distillation column, kJ/h
because additional make-up condensate is fed into the t = time, min
reboiler through the auxiliary valve from a feed water T = saturation temperature of the heating medium, O C
Pump. TB = temperature of the process fluid, "C
Conclusions of Dynamic Study. The dynamics of the Ttr = temperature on a tray of the distillation column, O C
condensate-throttling scheme can be slow and result in U = heat-transfer coefficient, kJ/(h m2 O C )
poorer performance than the vapor-throttling design. The V = total volume of the reboiler shell, m3
difference becomes even larger when the system is dis- Y = height of the condensate in the reboiler shell, m
turbed by a change in the supply pressure of the heating Greek Letters
medium. The dynamics of the condensate-throttling re- a = parameter of the PID controller
boiler become less and less important as the ratio T J T R A H = latent heat of vaporization, kJ/h
reaches or exceeds the value of 4 or 5. Thus, the con- AP = hydrostatic head of the condensate in the reboiler shell,
densate-throttling scheme can be used effectively with slow atm
columns. Several alternative control or process structures p = density of the condensate, kg/m3
can be used to improve the dynamic performance of the TD = derivative time, min
condensate-throttling scheme. 7 d = process dead time, min
T~ = reset time, min
Conclusions T~ = process time constant, min

This study was concerned with the design and the con- T R = reboiler time constant, min
T~~ = time constant of the temperature transmitter, min
trol aspects of the condensate-throttling reboiler scheme.
T~~ = time constant of the flow transmitter, min
It was found that the condensate-throttling design results
7, = time constant of the valve, min
in significant savings of the heat-transfer area when
turndown ratios are large or when pressures are near the Subscripts
critical pressure of the heating medium. B = of the process fluid
The valve opening a t the minimum conditions was design = at the design conditions
critical for the overall design. If smaller values of f(xmin) max = at the maximum conditions
can be used, the savings in the heat-transfer area of the min = at the minimum conditions
condensate-throttling design become very small. In a split out = downstream from the condensate-throttling control
range control valve scheme, the minimum valve opening valve
is not a limitation. v = at the supply conditions of the heating medium
2304 Ind. Eng. Chem. Res. 1988, 27, 2304-2322
1 = in the stream/vapor condensation region Papastathopoulou, H. S. "Design and Control of Condensate-
2 = in the condensate subcooling region Throttling Reboilers". Master's Thesis, Department of Chemical
Engineering, Lehigh University, 1987.
Literature Cited Shinskey, F. G. "When to Use Valve Positioners". Instrum. Control
Syst. 1971, Sept, 11.
Buckley, P. S.; Luyben, W. L.; Shunta, J. P. Design of Distillation Shinskey, F. G. Distillation Control for Productivity and Energy
Column Control Systems; Instrument Society of America: New Conservation, 2nd ed.; McGraw-Hill: New York, 1984.
York, 1985. Shinskey, F. G. Process Control Systems: Applications, Design,
Cheung, T. F.; Luyben, W. L. "PD Control Improves Reactor Adjustment; McGraw-Hill: New York, 1979.
Stability-. Hydrocarbon Process. 1979, Sept., 215-218. Yu, C. C.; Luyben, W. L. "Analysis of Valve-Position Control for
Connell, J. R. "Realistic Control-Valve Pressure Drops". Chem. Eng. Dual-Input Processes". Ind. Eng. Chem. Fundam. 1986, 25,
1987, Sept, 123-127. 344-350
Luyben, W. L. Process Modeling, Simulation and Control for
Chemical Engineers; McGraw-Hill: New York, 1973. Received for review March 15, 1988
Masoneilan Handbook for Control Valves Sizing; Masoneilan In- Revised manuscript received July 13, 1988
ternational: New York, 1977. Accepted July 31, 1988

Studies in Chemical Process Design and Synthesis. 8. A Simple

Heuristic Method for the Synthesis of Initial Sequences for Sloppy
Multicomponent Separations?
Shueh-Hen Chengj and Y. A. Liu*
Department of Chemical Engineering, Virginia Polytechnic Institute a n d S t a t e University, Blacksburg,
Virginia 24061

A simple heuristic method for the synthesis of initial sequences for sloppy multicomponent separations
is proposed. In this type of separation, some components being separated may simultaneously appear
in two or more product streams. Included in the proposed method are (i) effective and flexible tools
for representing the synthesis problem, called the component assignment diagram (CAD), and for
analyzing the technical feasibility of separation tasks or product splits, called the separation
specification table (SST); (ii) practical design guidelines for the shortcut feasibility analysis of product
splits; and (iii) simple rank-ordered heuristics for the synthesis of initial separation sequences. Of
particular significance in the method is the quantitative consideration of the feasibility of product
splits. The proposed method has been applied to a number of industrial separation problems. The
results show that the heuristic method offers a simple and effective means for design engineers to
generate several good initial sequences for sloppy multicomponent separations prior to the final heat
integration and separator optimization.

1. Introduction A typical chemical process usually comprises a reaction

An important process-design problem in multicompo- system and a separation system. The separation system
nent separations is separation sequencing, which is con- generally involves reactant recovery, product separation,
cerned with the selection of the best method and sequence and waste and byproduct separations. As was pointed out
for the separation system. This problem has been covered by Douglas et al. (19851, the problem of flow-sheet syn-
in textbooks by King (1980, pp 710-720) and Henley and thesis can be posed as an optimization problem in which
Seader (1981, pp 527-555). Reviews of recent studies on the reaction conversion is a key parameter for the optimum
the subject can be found in Westerberg (1985) and Liu flow sheet. Further, there is an economic trade-off between
(1987). high reactor costs and significant selectivity losses asso-
Most of the published work on multicomponent sepa- ciated with high reaction conversions, which are balanced
ration sequencing has been limited to high-recovery or against large separation costs at low reaction conversions.
sharp separations, in which each component being sepa- This implies that it may not be necessary to purify un-
rated appears in one and only one product stream. In converted reactants to a high level prior to recycling them
industrial practice, however, it is often useful to permit back to the reactor system. In other words, sloppy sepa-
some components to simultaneouslyappear in two or more rations may be sufficient for many processes. The ob-
product streams. This type of separation is called non- jective of this work is to present a simple heuristic method
sharp or sloppy separation. Figure 1 compares two al- for the systematic development of initial sequences for
ternative schemes (sharp and sloppy) for the separation sloppy multicomponent separations.
of a three-component mixture. We see that one more Section 2 introduces an effective and flexible framework
separator is needed in the sharp scheme than in the sloppy for representing the problem of multicomponent separa-
sequence. tion sequencing. In particular, we propose a new repre-
sentation, called component assignment diagram (CAD),
* To whom correspondence should be addressed. to facilitate the identification of alternative separation
' P a r t I of this series appears in AIChE J. 1985, 31, 487. tasks or product splits. In section 3, we analyze the
Present address: Glitsch, Inc., P. 0. Box 660053, Dallas, TX technical feasibility of product splits based on component
72566. recovery specifications. Key developments in this section
0888-5885/88/2627-2304$01.50/0 1988 American Chemical Society