Computers&em. Engng, Vol. 15, No. 2. pp. 133-139, 1991 009%1354/91 53.00 + 0.

00
Printed in Great Britain.All rightsreserved copyright Q 1991 PerganlonPress plc

SHORT NOTE

MODELLING OF A CRUDE DISTILLATION COLUMN

P. LANG', G. SZALP&S~, G. CHIKANY' and S. KEMI%NY'
Department
‘ of Chemical Engineering, Technical University of Budapest, Mtiegyetem rakpart 3, H- 152 1,
Budapest XI, Hungary
ZHungarian Hydrocarbon Institute, Hungary

(Received 7 January 1987; Final revision received 3 August 1990; received for publication 28 August 1990)

Abstract-An algorithm and a FORTRAN program have been developed for modelling of crude
distillation and vacuum columns on smaller computers. Wang-Henke and Sum of Rates methods are
combined in a new way. The memory requirement is low even in the case of a great number of
pseudocomponents. An atmospheric crude distillation column has been simulated. Crude oil is described
with 50 pseudocomponents chosen on the basis of the TBP curve of the feed. Calculated and experimental
results are compared.

INTRODUCYl’ION the method were demonstrated and compared. Water was

In order to successfully simulate a complex crude distillation
column (Fig. 1) the following difficulties must be overcome:
-Since the crude oil is a complex mixture containing
more than a hundred components and oil is separated
into many products, computation must be carried out
with a great number of pseudocomponents.
-Sidestrippers and pumparounds must be modelled.
-In the upper section the temperature increases down-
wards to a great extent while in the lower part tempera-
ture decreases because the main column has no reboiler.
The column operates in a very wide temperature range.
-Watersteam used for stripping must be taken into
account.
-Simple and fast VLE and enthalpy models must be
used.
Amundson et al. (1959) were the first who modelled a distil-
lation column with a sidestripper extending their algorithm
of the BP type. This method involved separate convergence
of the main column assuming compositions of the vapor-
return streams from the sidestrippers. After each sidestripper
is converged, revised vapor-return streams are used to con-
verge the main column. Dickey et uZ.(1962) and Cechetti el al.
(1963) suggested simultaneous modelling of the main
column and sidestrippers using the 8 method. In these earlier
works water was treated as a single-phase light (SPL) com-
ponent. These algorithms of the BP type may fail to con-
verge when modelling a crude oil distillation tower having
no reboiler. The t, L and V profiles of the lower part of the
main column cannot be precisely calculated by BP methods.
The “multi-8 method” was applied by Hess and Holland
(1977) and Hess et al. (1977) for modelling of absorber-type
pipestills since the 0 method had failed to converge for
columns of this type. This method belongs to the “2N-
Newton-Raphson methods” originated by Tomich (1970).
The plate temperatures and the L/l’s were considered as
independent variables. Three different numerical versions of
TStripping factors: Sj = Kb. ( VjfLr).
Withdrawal factors: RLj = 1 + Ul/Lj and R,, = 1 + Wj/ V,.
already regarded as being distributed between the vapor and
liquid phases on stages except the condenser. The rates of
side products withdrawn from sidestrippers are not specified
by Hess and Holland, which is very uncomfortable when
modeliing.
This method has the following disadvantages:
-Time consumption of one iteration step is very high
(almost (2N + 1) times greater than that of the BP
method).
-Its memory requirement is also higher than that of the
BP or SR methods.
-It needs good initial estimates in order to converge
successfully.
Its advantages are the following:
-It converges more rapidly mainly in the vicinity of the
solution than BP and SR methods, therefore fewer
iteration steps are needed.
-It can be used for the whole range of tower types
because of its generality.
Russel(l983) published a method belonging to the “inside -
our”class of methods originated by Boston and Sullivan
(1974) which was also used for solving crude-distillation-
column problems. In the inner loop, stripping and sidestream
withdrawal factors? are converged with relative volatilities
assumed (and kept constant in the inner loop) using the
quasi-Newton method to achieve all enthalpy balances.
After solving component-material balances, temperatures
are computed with the “Kb method”. Enthalpies are calcu-
lated using a simple enthalpy model. The component flow
summations give new V, (and Lj) values. In the outer loop
relative volatilities are updated and enthalpies are calculated
rigorously. (The removal of time-consuming Kand enthalpy
calculations to an outer loop gave rise to the term “inside-
out”).
This method allows a wide variety of performance
specifications and provides good speed, but it is rather
complicated.
Algorithms of the global Newton-Raphson type [e.g.
Naphtali and Sandholm (1971), Gallun and Holland

133
134 Short Note

CONDENSER

---c water

___
-5: NO. 3.

feed H heavy gas oil

crude
I----- steam

steam ---w-J
oil
Fig. 1. Crude distillation column with actual plate numbers.

(1976) and KubiEek et al. (1976)] are not applicable to solve
crude distillation problems with a great number of pseudo-
components on small computers because of their great
memory requirements.
ALCOR~M PROPOSED
In this paper only the main characteristics of the new
algorithm are summarized. In the new method, BP and SR
methods were combined for rnodelling further developing
the idea of Tim&r et al. (1979). (The above authors suggested
modelling of the upper part of the reboiled absorbers by
SR and that of the lower section by BP methods.) The
Wang-Henke algorithm (Wang and Henke, 1966) is used
for modelling the upper rectifying section (plates above the
feed plate) of the main column. For simulating the lower
stripping section of the main column and sidestrippers the
Sum of Rates (SR) method of Burningham and Otto (1967)

1 specified
_---_+ calculated

a.) usuai b.) modif ied

Fig. 2. Usual and modified specification of feed and product rates.
 Short Note 135

Give initial tj , Lj values (Vi;,5 are colcu-

toted from mater ial balances )

Calculate Q’ s from component-material balances

+
Normalize xi _ i mole-fractions

4
Compute tj bubble-points in the upper part of
the main column

Update Vj (and Lj -1) molar flows in the lower

Ipart of the main column using SR methcd (j-f....,N’ ) I

Calculate new tj temperatures in the lower part

of the main column (SR method)

Compute new Vi s for the sidestrippers L15ing

5R method
(Li_: s and rates of liquid sidestreams entering the

sidestrippers are calculated from material balances 1

1Update t’ s la the sidestriFpers (SR method) 1

(and Vi;,s) in the upper port of the main column

(from l-eat and material balances using CC method)

Yes
f
END

Fig. 3. Simplified block diagram of the algorithm.

is suggested. The SR method is applied with a modified type
of specification in order that liquid product rates leaving the
bottoms of the stripping section and sidestrippers could be
kept constant. When applying the SR method (e.g. for the
simple absorber shown in Fig. 2a) feed rates (F, and FN)
are usually specified. In this case there is no possibility
of fixing a product rate because of the lack of degrees of
freedom.
By the modified type of specification applied in the new
algorithm instead of the upper feed rate (F,) the bottom
product rate (LN) is fixed (“lower end specification”). For
the stripping section and sidestrippers the modified type of
specification is more advantageous. (Rate of a product of
the crude distillation tower is fixed instead of the rate of an
inner liquid stream.) This “lower end specification” requires
a slight modification in the original SR algorithm. Instead
of 4s the vapor flowrates (V,s) must be updated by the
summation equations (For modelling of sidestrippers the BP
method can be applied in the computer program, too, but
in our experiences the SR method usually provides better
convergence.)
For solving the almost tridiagonal set of component
material balance equations (see later in this paper) KubiEek’s
algorithm (1973) is applied. Water may be regarded as
being distributed between the vapor and liquid phases or
as an SPL component. When solving heat balance equations
by the BP method, the constant-composition method of
Weisenfelder et al. (1961) is used in a modified form which
is very advantageous in the case of a complex column
with many feeds and sideproducts. (Enthalpy and material
balances are written around the individual plates.)
In order to give more accurate initial temperature and
molar flow profiles the conception of “disturbed plates”
[e.g. Pierucci er al. (198211 is applied.
Appropriate damping technics usual at BP and SR
methods based on experience are used when updating xU
 Short Note

COMPONENT-MATERIAL BALANCES
For the column shown in Fig. 1. a model column with
theoretical plates was created (sea Fig. 4). For this model
column the coefficient matrix of the set of equations of
component-material balances differs from a tridiagonal
matrix by the appearance of three non-zero elements to the
left and three non-zero elements to the right of the tridiagonal
band of elements which result from the introduction of the
sidestreams (U,, U,, and U,,) to the sidestrippers and from
the return of the vapor streams (V,, , V,, and V,) from the
sidestrippers, respectively. The abbreviated matrix display
for the set of equations of component-material balances is
shown in Fig. 5. (Component indices and off-band zeros are
omitted.)
Liquid pumparounds from one theoretical stage returning
to the same or to an adjacent stage result in no further
off-band elements. A pumparound from one theoretical
stage to a non-adjacent stage somewhere above would result
in a non-zero off-band element to the right of the tridiagonal
band.

Fig. 4. Model column with theoretical plates.
liquid mole fractions, tj temperatures and Li, Vi molar
flowrates.
A simplified block diagram of the new algorithm is given
in Fig. 3.
Instead of the Wang-Henke method it would be worth
trying to apply the Jedlovsxky (1974) algorithm. (This
method usually provides better convergence for the rectifica-
tion of hydrocarbon mixtures.)
PROPERTTRE OF PREUDOCOMPONENTS
In order to define psuedocomponents and calculate
physicochemical properties, a TBP curve of the crude oil was
determined. The crude oil was separated into 30 fractions on
a ASTM D 2892-78 standard distillation column. Boiling
temperatures, densities and molecular weights of the fractions
were measured, and correlated with a polynomial of fifth-
order as the function of the relative amount boiled off. The
possibility of assigning pseudocomponents of an arbitrarily
narrow boiling temperature range (e.g. 1O’C) was provided
by the use of these polynomials. On the bases of the boiling
points, densities and molecular weights of the pseudo-
components, their critical parameters were estimated by the
method of Riaxi and Daubert (1980). The ideal gas enthalpy
polynomial coefficients can be determined by the method of
Daubert and Danner (1983). Enthalpies of the real fluids on
the pressure and temperature of the industrial column were

r _ _ _ .
a, Xl 4
a2 x2 dz

. . . X3 d,
. . - .
.a7 ‘b, c7 --------’ --c W21 X?
.’ . I
.
1 :. .
1. * . I
.
1 aI2 bq2 c,> -----.
I
7 - -. ‘Jzs’K2. X12
1
I
I Xl3
I .
I . =
-_ f
j----, L --V2s-K25 k3
.
..
I

I
I I
1

I
0 I
0 / X20

I_a22
bzz
10
) b, c21 X21 da
x22 6,
lJ,3+____---_0 X23 63
I
I X24 d24

1I 1
I
“,a- __-__ --____.o
bzs ~25 X23 d,
02s bzs X26 4,
_
Fig. 5. Set of equations of component-material balances.
 Short Note 137

Table 1. Composition of the hydrocarbon feed pseudocompotmnts. The numbers of theoretical stages,
Pseudo- Pseudo- Pseudo- locations of f&s. sidestreams and pumparounds can be
component Mel% component Mel% component MC+% seen in Fig. 4.
No. I No. 2 No. z The hydrocarbon f& (F,, = 754.11 km01 h-i) enters the
1 8.1 16 2.8 31 1.3 column at 292°C. The steam enters the main column and
2 2.9 17 2.7 32 1.0 sidestrippers as superheated vapor at 400°C with rates
3 2.8 18 2.6 33 0.9 F,= 79.44 and Fz2= Fu= F,=38.89. The rate of top
4 3.5 19 2.0 34 0.9 product U, = 173.05 kmol h-i. The side product-rates are
5 4.2 20 1.4 35 0.6 specitied L, = I14.06, Lr., = 180.26 and r, = 79.7 km01 h-‘.
6 4.1 2.3 36 0.9
The upper pumparound stream is to be withdrawn at the rate
7 4.1 ;: 1.8 37 0.8
of F; = 63 1.92 km01 h-‘, its temperature decreases by 52°C.
8 3.0 23 1.7 38 0.5
9 5.8 24 2.1 39 0.8
The rate of the lower pumparound stream F;,, = 707.27 km01
10 3.5 25 1.2 40 0.5 h-‘, the decrease of its temperature is 78°C.
11 3.3 26 1.9 41 0.3 The initial value of the.reflm ratio L, /U, = 2. The pressure
12 3.2 27 1.1 42 0.2 in the condenser is 0.953 bar, the pressures in the bottoms
13 1.9 28 1.1 43 0.2 of the main column and the sidestrippers are 1.15 and 1.04,
14 2.4 29 1.1 44 7.6 1.lO and 1.I2 bar, respectively. (Linear pressure profiles are
15 3.4 30 1.0
assumed in the main column and in the sidestrippers.) The
temperature of the reflux is 64°C. The composition of the
hydrocarbon feed is shown in Table 1.
calculated by the Benedict-WebbRubin equation of state Initial temperature and molar flow profiles and results of
modified by Lee and Kesler (1975). The calculated enthalpies the computation are presented in Tables 2-4. In order to
were correlated with a polynomial of third-order as a function show the agreement between the experimental and calculated
of temperature. product compositions the measured and calculated cumula-
The vapor pressures of the pseudocomponents were also tive distribution curves are plotted in Fig. 6 for a similar
computed according to Lee and Kesler. Three data points operational state of the column. Calculated distribution
were correlated with the Antoine equation for column calcu- curves for this latter case can be seen in Fig. 7 (dashed lines).
lation purposes. In the program vapor-liquid equilibria may
be described with a combination of Flory-Huggins and
Hildebrand models (Hildebrand and Scott, 1950) or an ideal
liquid phase may be assumed. In the latter case water must Table 3. Product comwsitions calculated ~mol%)

be treated as a single-phase light component.

EXAMPLES
1 34.49 0.96 0.14 0.01 0.00
The atmospheric crude distillation column shown in Fig. I 2 12.14 0.61 0.09 0.00 0.00
3 11.36 1.04 0.14 0.01 0.00
is modelled. Crude oil is described as a mixture of 44
4 13.35 2.40 0.29 0.03 0.00
13.83 5.84 0.56 0.05 0.00
Table 2. Temperature and molar flow profiles : 9.22 11.64 0.88 0.09 0.01
7 4.55 21.32 1.70 0.16 0.02
Temperature (“C) Liquid (km01 h-‘) Vapor (km01 hm’) 8 0.77 15.39 1.97 0.17 0.02
Plate 9 0.28 25.78 7.36 0.53 0.06
NO. Initial Final Initial Final Initial Final 9.89
10 0.02 8.05 0.54 0.05
1 64.0 64.0 346.1 361.9 - - 11 0.00 3.87 10.88 0.85 0.07
2 100.0 111.1 346.1 429.5 715.3 731.0 12 0.00 1.08 11.98 I.32 0.11
3 117.0 120.5 534.6 648.8 715.3 798.6 13 0.00 0.15 7.20 1.22 0.10
4 122.0 129.9 534.6 651.2 903.8 1018.0 14 0.00 0.04 8.76 2.32 0.19
5 127.0 135.3 534.6 641.0 903.8 1020.3 15 0.00 0.01 11.59 4.87 0.40
6 132.0 140.0 534.6 628.8 903.8 1010.1 16 0.00 0.00 8.59 5.77 0.49
7 137.0 144.5 398.6 421.3 903.8 998.0 17 0.00 0.00 7.09 7.69 0.69
8 146.3 149.8 398.6 413.0 843.0 865.7 18 0.00 0.00 5.48 9.71 0.95
9 155.2 153.8 398.6 402.9 843.0 857.4 19 0.00 0.00 3.12 9.17 I .03
10 165.0 157.4 775.7 796.5 843.0 847.2 0.00 0.00 1.48 7.31 0.99
11 170.0 168.5 775.7 739.2 1220.0 1240.8 z 0.00 0.00 1.49 12.65 2.20
12 188.0 176.8 775.7 686.9 1161.4 1094.3 22 0.00 0.00 0.63 9.60 2.31
13 206.0 188.9 575.7 358.4 1161.4 1041.9 23 0.00 0.00 0.30 8.10 2.81
14 226.5 209.9 575.7 282.0 1161.4 944.1 24 0.00 0.00 0.17 8.20 4.34
15 247.0 229.8 485.7 88.4 1161.4 867.7 25 0.00 0.00 0.04 3.48 2.99
16 262.0 247.8 485.7 60.4 1112.2 714.9 26 0.00 0.00 0.02 3.66 5.48
17 277.0 257.3 485.7 57.8 1112.2 686.9 27 0.00 0.00 0.00 1.25 3.51
18 292.0 286.9 694.9 248.8 1112.2 684.3 28 0.00 0.00 0.00 0.67 3.74
19 291.5 283.0 694.9 225.1 567.3 121.2 29 0.00 0.00 0.00 0.32 3.88
20 291 .O 277.6 207.0 207.0 567.3 97.5 30 0.00 0.00 0.00 0.13 3.59
21 137.5 127.9 125.0 128.4 60.8 73.0 31 0.00 0.00 0.00 0.07 4.70
22 138.0 112.1 114.0 114.1 49.9 53.2 32 0.00 0.00 0.00 0.02 3.63
23 198.0 176.8 190.1 202.4 58.6 89.3 33 0.00 0.00 0.00 0.00 3.27
24 190.0 164.6 180.3 180.3 48.8 61.1 34 0.00 0.00 0.00 0.00 3.27
25 238.5 217.3 84.9 91.2 49.2 66.7 35 0.00 0.00 0.00 0.00 2.18
26 230.0 203.6 79.7 79.7 44.0 50.4 36 0.00 0.00 0.00 0.00 3.27
37 0.00 0.00 0.00 0.00 2.91
Sidestream rates
38 0.00 0.00 0.00 0.00 1.82
(kmol h-‘)
39 0.00 0.00 0.00 0.00 2.91
Plate
40 0.00 0.00 0.00 0.00 1.82
NO. Initial Final
41 0.00 0.00 0.00 0.00 1.09
7 136.0 148.1 42 0.00 0.00 0.00 0.00 0.73
13 200.0 230.7 43 0.00 0.00 0.00 0.00 0.73
15 90.0 107.6 44 0.00 0.00 0.00 0.00 27.64

Cm.3 15,2-E
138 Short Note

Table 4. E&halpies and heat duties calculated: (a) the feed enthalpk REMARKS
hydrocarbon: hF,,s= 1.439 x loSkJkmol-‘, watcrsteam = hr M
=h,,=h,,=hF.,= 4.644 x 104 kJ kmo1-‘; (b) molar entha1pika The method suggested usually provides good convergence.
of liauid and vabor flows However, in the vicinity of the solution convergence becomes
Molar _
entbduics 03 kmo1-‘) slower as is usual with methods of total decomposition.
Generally 50-60 iteration steps are enough to achieve a
Plate No. Liquid Vapor
result of sutiicient accuracy for a chemical engineer.
1 -7.662 x 10’ -
When water is treated as being distributed between liquid
2 1.1666 x Iti 4.625 x 10’ and vapor phases at least one product rate must not be fixed,
3 1.485 x 10’ 4.895 x 104
4 1.821 x 10’ 5.194 x lti
therefore in most of the calculations water was considered
2.048 x 10’ 5.4cKl x lti an SPL component.
: 2.260 x 104 5.580 x lti The program was also successfully used for modelling a
7 2.479 x lOa 5.757 x Iti complex vacuum unit without condenser.
2.734 x 10’ 5.961 x 10’ The algorithm proposed can also serve for providing
f 2.944 x 10’ 6.118 x 104 appropriate initial values for other computation methods
10 3.148 x 10’ 6.258 x LO4 which am more sensitive to the starting estimates.
11 3.704 x 104 6.839 x 104 A detailed algorithm including the damping techniques
12 4.218 x Iti 7.239 x lti
applied and the characteristics of the pseudocomponents
13 5.109 x 104 7.756 x IO
14 4.855 x Iti 8.678 x 10’ will be sent upon request.
15 8.779 x 104 9.560 x lti
16 1.092 x 105 1.033 x 10s Acknowiedgemenrs-The authors are indebted to Dr A.
17 1.248 x 10s 1.058 x 10’
Deak, Mr Gy. Vancsura of the Technical University,
18 1.892 x lo5 1.156 x 10’
19 1.942 x 10J 8.642 x 104
Budapest and Z. Merth of the Hungarian Hydrocarbon
20 1.946 x 10’ 7.285 x 10’ Institute for their help in making parts of the computation
21 1.946 x 10’ 5.244 x lo* work.
22 1.449 x 104 4.982 x 10’
4.675 x 10’ 6.714 x 10’
:: 4.214 x 10’ 6.019 x lti NOMENCLATURE
25 8.373 x 10’ 7.570 x Iti
26 7.774 x 10’ 6.440 x lo* II, b, c = Coefficients in the set of component
material
balance equations
the enthalpies of the pumparound streams after cooling d = Right-hand side elements in the set of com-
Enthaluy ponent material balance equations
Plate (kJ LmOizl) f = Number of feed plate of crude oil
3 3.587 x 102 F = Total flowrate of a feed
10 5.158 x 103 F’ = Molar flowrate of a liquid pumparound
h = Molar liquid enthalpy
(c) heat duties of the condenser and pumparound heat exchangers U = Molar vapor enthalpy
Heat duties K = Vapor-liquid equilibrium constant
Plate (LJ h-‘) L = Liquid molar flowrate
-3.325 x 10’ N = Number of trays
: -9.155 x 106 R = Withdrawal factor
10 - 1.862 x 10’ S = Stripping factor
U = Liquid sidestream
U’ = Water stream
Figure 7 illustrates the modification of calculated distribution V = Vapor molar flowrate
curves due to the change in product rates when an LGO x = Molar liquid fraction
fraction containing pseudocomponents from 8 to 15 had to y = Molar vapor fraction
be produced in a larger quantity. z = Molar fraction in a feed

- Measured

0 Calculated

Pseudocomp. No.
Fig. 6. Measured and calculated product compositions.
 Short Note 139

25

--- Bask

- Modified

15

10

Pseudocamp. No.

Fig. 7. Modification of product distribution curves.

Subscripts Gallun S. E. and C. D. Holland, Hydrocarbon Process. 55,

b = Base component I37 (1976).
F = Feed F. E. Hess, C. D. Holland, Hydrocarbon Process. 56(S), 241
i= Component (1977).
j = Tray Hess F. E. er al. Hydrocarbon Process. 56(6), 181 (1977).
L = Liquid Hildebrand J. H. and R. L. Scott, The Solubility of Non-
V = Vapor Electrolytes. Reinhold, New York (1950).
Superscripts
Holland C. D., G. P. Pendon and S. E. Gallun, Hydrocarbon
Process. 54(l), 101 (1975).
I = Main column
II, III, = Side-stripper Jedlovszky P., Clrem. Engng. Sci. 29(l), 287 (1974).
KubiEek M.. Commun. ACM 16(121. 760 (1973).
KubiEek M., V. Hlav%ek and F.‘P&chask& Chkm. Engng.
Sci. 31, 277 (1976).
REFERENCES Lee B. I. and M. G. Kesler, AIChE JI 21, 510 (1975).
Naphtali L. M. and D. P. Sandholm, AlChE JI 17(l), 148
Amundson N. R., A. J. Pontinen and J. W. Tiernay, AIChE (1971).
\-~
Jf 5, 295 (1959). Pieru&S. et al.. Computers &em. Engng 6, 39 (1982).
Boston J. F. and S. L. Sullivan Jr, Can. J. Chem. Engng Feb, Riazi M. R. and T. E. Daubert, Hydrocarbon Process. 59(3),
52 (1974). 115 (1980).
Burningham D. W. and F. G. Otto, ffydrocarbon Process. Russel R. A., Chem. Engng Ott 17, 53 (1983).
Ott, 163 (1967). Tim&r L., Z. Csermely and S. BBcskai, Hung. J. Ind. Chem.
Cechetti R. C. er al., Hydrocarbon Process. 42(g), 159 (1963). 7(4), 377 (1979).
Daubert T. E. and P. P. Danner, API Technical Data Book Tomich J. F., AZChE Jl 16, 229 (1970).
Perroieum Refining, 3rd Edn. Washington, D. C. (1983). Wang J. C. and G. E. Henke, Hydrocarbon Process. 45, 155
Dickey B. R.. C. D. Holland and R. Cecchetti, Petrol. (1966).
Refiner 41(2), 143 (1962). Weisenfelder A. J. et al., Petrol. Refiner 40(10), 175 (1961).