This preview shows page 1. Sign up to view the full content.
Unformatted text preview: Reactions and Separations Increase Capacity and
for Existing Refinery
and Robin Smith,
UMIST This method optimizes the existing distillation system
and its heat-exchanger network simultaneously,
lowering energy consumption and freeing up
capacity at a minimum capital investment. D ISTILLATION COLUMNS ARE AMONG
the biggest energy consumers in the chemical
industries, particularly in oil reﬁneries.
Retroﬁt projects in reﬁneries mostly aim at reducing energy consumption and increasing throughput to increase
proﬁts and meet market demands. Usually, plants aim to
achieve their retroﬁt objectives by reusing the existing
equipment efficiently, rather than installing new towers
and heat exchangers, which requires a substantial capital investment. Retroﬁtting schemes
A number of modiﬁcations have been suggested to
the distillation column and the heat exchanger network
to meet these two goals. Sittig suggests changes to improve the efficiency of distillation systems, including
the installation of new internals with higher efficiency,
the use of intermediate reboilers, etc. (1). Bannon and
Marple recommend making other column modiﬁcations, such as installing pump-arounds at suitable locations in the unit and adjusting the cooling duty for each
pump-around (2). Harbert’s idea is to install preﬂash
units or prefractionators before the crude oil distillation
unit. This would save energy and increase the throughput of the column (3).
Rivero and Anaya call for installing additional trays, as
well as adding reboilers to the stripping columns (4). Fraser and Sloley propose increasing the capacity of crude-oil
units by adding pump-arounds, reducing the operating
pressure and increasing the preﬂash overhead vapor (5). 44 www.cepmagazine.org April 2003 CEP In reﬁnery distillation systems, the energy-efficiency of the process strongly depends on the heat-exchanger network design. For example, the duty and
temperature drop of each pump-around affects how
much heat can be recovered, and the connections between the heat exchangers (also known as matches)
and heat-exchanger areas determine how much heat is
Within the last decade or so pinch analysis has
been applied to identify modifications to the column
and the heat exchanger network. Linnhoff and
Dhole’s idea is to use the column’s grand composite
curve (CGCC) to identify suitable modifications that
would save energy (6). Dhole and Buckingham extended this method for energy saving and debottlenecking of refinery distillation systems. Their method
has three stages (7). First, column modifications are
made using the CGCC, then the heat-exchanger network design is changed to save energy by adding
more heat-transfer area, and, finally, design changes
are instituted to debottleneck the arrangement.
Liebmann’s approach is a two-step method for
retroﬁtting (8). The distillation column is ﬁrst modiﬁed
to reduce its energy demand, and the CGCC provides
guidelines for improving the heat-recovery potential.
Afterwards, the column is remodiﬁed, followed by a
reanalysis of the CGCC to further increase the heat recovery. Overall, two levels of modiﬁcations are proposed, those that are relatively inexpensive and those
that require larger investments. Examples of inexpen- sive modiﬁcations include piping changes to avoid mixing
unlike streams together, and adjusting the stripping-steam
ﬂowrates. Capital-intensive options include replacing internals and relocating feeds and draws.
Briones et al. (9) successfully applied Liebmann’s
retrofit approach to discover modifications for reducing
the energy consumption of the crude-oil distillation unit.
Bagajewicz et al. adapted Liebmann’s approach by linking the pinch analysis with rigorous simulation and optimizing the column’s operating parameters (10). Improved method
As helpful as they are, none of these retrofit methods
assesses the existing heat-exchanger network together
with the crude distillation column. Further, they do not Nomenclature
a, b, c =
m, c =
= total area required for heat exchangers, m2
ﬂooding parameters (based on stage spacing)
existing heat exchanger area, m2
area added for retroﬁtting, m2
bottom product ﬂow, kmol/h
capacity factor, dimensionless
top product ﬂow kmol/h
ratio of downcomer area to total area, dimensionless
stage diameter, m
total energy consumed after retroﬁt, MW
ﬂow parameter, dimensionless
liquid mass ﬂow, kmol/h
retroﬁt area model parameters in Eq. 12
total number of stages in column
minimum number of stages at total reﬂux
number of stages in rectifying section
number of stages in stripping section
fractional recovery of heavy key to bottom product
fractional recovery of light key to top product
minimum reﬂux ratio, dimensionless
mole fraction of heavy key in feed
mole fraction of light key in feed
design vapor velocity, m/s
ﬂooding velocity, m/s
vapor mass ﬂow, kmol/h
vapor volumetric ﬂow, m3/s Subscripts:
Fensk = Fenske
Gill = Gilliland
Kirk = Kirkbride
αLK = relative volatility of light key
αHK = relative volatility of heavy key
= liquid mass density, kg/m3
= vapor mass density, kg/m3
= ratio deﬁned in Eq. 3
= factor deﬁned in Eqs. 4 and 5
ΨGill = factor in Gilliland correlation, see Eq. 2 offer a systematic approach to retrofitting. Rather, they
propose various modifications. These methods suffer
from two more drawbacks: Some of the proposed column modifications would require a substantial capital investment, while others sometimes violate such constraints such as the maximum tray capacity. This approach aims to identify the set of operating conditions in
an existing distillation column that will allow the existing heat-exchanger network (modified by adding heattransfer area or changing the piping arrangements between exchangers, for example) to best recover heat. The
hydraulic limits of the column constrain which design
solutions can be considered.
Thus, we offer a systematic approach for retroﬁtting existing reﬁnery distillation columns. This method simultaneously considers the existing heat-exchanger network along
with reducing energy consumption and increasing the
throughput of the existing distillation unit. Framework used
The authors’s optimization framework uses models
for the column and the heat-exchanger network, as well
as pinch analysis. Optimization is carried out using a
successive quadratic programming (SQP) solver. The
solver uses an algorithm that is aimed at large, sparse
nonlinear programs. In essence, a quadratic approximation of the highly nonlinear problem is solved during
each iteration (11).
The system is represented using a column retrofit
model and a heat-exchanger network retrofit model. The
column model captures the relationships between operating conditions, product quality and column design. The
heat-exchanger model presents the results of a detailed
study of options for improving heat recovery in the network by making minor changes to the heat-exchange
hardware and configuration. For example, the sequence
of heat exchangers may be changed or a new exchanger
may be installed.
The column design is optimized using these two models. That is, its operating conditions are selected for a
column with a fixed design (the given number of stages,
diameter, feed and draw locations, etc.) to minimize the
sum of the utility costs and investment in the heat-exchanger network. The network retrofit model is used to
calculate th cost of additional heat-exchange area during
the optimization. Thus, the optimization couples the two
independent models and accounts for interactions between the column design and the heat-exchangernetwork performance.
The column retroﬁt model uses the existing parameters,
such as the number of stages and their distribution, locations of condenser, reboiler and pump-arounds, and product speciﬁcations. The heat-exchanger network retroﬁt
model takes into account the network’s details, such as the
heat-transfer areas and duties for each exchanger, the ex- CEP April 2003 www.cepmagazine.org 45 Reactions and Separations isting matches, and the existing energy consumption.
During optimization, the user can vary the column’s operating conditions, such as changing the feed preheating
temperature, steam ﬂows to each section, reﬂux ratio, and
temperature drop and ﬂow of liquid recycled by pumparounds for minimum energy consumption. The hydraulic
constraints and capacity limitations of the existing column
are taken into account during this process.
Optimization of an existing refinery distillation unit
identifies the optimum process changes for minimum
energy consumption. Since the retrofit design does not
change the dimensions or internals of the column, these
modifications do not require a major capital investment. Rather, they are simply changes to the operating
conditions. Reducing the energy consumption of the existing crude distillation unit allows the column throughput to increase, due to the resulting reduced vapor
flows. In determining the maximum increase in
throughput, the model identifies any column bottlenecks that would limit this increase, and evaluates proposed modifications for debottlenecking.
To run the model, a FORTRAN code was developed
that allows modeling via short-cut models, and an interface was created between the code and a rigorous simulation package. This allowed the users to obtain thermodynamic and physical data for the components and
A rigorous simulation was first performed on the existing arrangement (the base case) to initialize the shortcut calculations, e.g., specifying the key components
and their recoveries, plus matching the product
flowrates to those in the existing operation. This rigorous simulation should yield a reasonably accurate picture of the product flows and compositions, steam
flows, pump-around duties, flowrates and temperature
drops, among others. The simulation results are useful
in checking the validity of the column retrofit model, as
well as for initializing this model. The user’s manual of
the simulation software that the reader employs will
provide guidance on how to rigorously model refinery
columns. Further instructions on setting up the problem
are found in Ref. 12. Retroﬁt model for the column
The short-cut column model is based on the Underwood equation for the calculation of the minimum vapor
flow in a column. The basic model equations are those of
Fenske, Gilliland and Kirkbride together with consecutive flash calculations (13), and key-component material
balances. The retrofit equations for distillation columns
with reboilers are:
N min = N (1 − ψ Gill ) − ψ Gill 46 www.cepmagazine.org April 2003 (1) CEP where: 1 + 54.4ξ ξ − 1 ψ Gill = 1 − exp 0.5 11 + 117.2ξ ξ (2) and:
R − Rmin
The following two terms are deﬁned as:
ξ= φ Fenske φ Kirk α = LK α HK N (3) min B x fHK = D x fLK (4)
1/ 2 NR NS 2.427 (5) The recovery of the heavy key in the top product is
1/ 2 RHK 2 φ
φ Fensk − 1 + 1 Kirk
= 1− +4
2(φ Fensk – 1) φ Fensk (φ Fensk ) 2 φ Kirk (φ Kirk + 1)
2φ Kirk (φ Fensk − 1) While the recovery of light product in the bottoms is: RLK 2 (φ φ Kirk
Kirk + 1) = 1 − +φ 4(φ Fenske − 1) ( Fenske − 1) + (φ Kirk + 1)
2(φ Fenske − 1) 1/ 2 (7) A similar retrofit model can be written for distillation
columns that employ steam stripping. In this case, consecutive flash calculations are used to model the stripping
section (12). The two retrofit models relate the product
compositions to the existing number of stages, the distribution of the stages, and the existing operating conditions. The models treat the distillation column as fixed
and calculate the product flows, temperatures, compositions and duties. This provides the basis for optimizing
the existing column. Hydraulic analysis
To evaluate the column hydraulics, the diameter required for vapor flow is calculated for those stages
where there is a significant change in the vapor and liquid flows. Such stages include the top and bottom trays,
pump-around stages, and the feed stage. For sieve plates,
the diameter is calculated from the flooding limits via Water 8 Water 8 LN PA3 Existing
Retrofit Models Retrofit Shortcut
HN 6 8
PA3 PA1 Steam 2 PA2 HD 3 PA1 2
1 Steam 1 6
Feed 1 LN
Steam 2 PA2 Heat Exchanger Area 5
Network HN Steam 1 LD 4
PA1 Heat-Exchanger Network
Retrofit Energy Demand HD 5 3 Retrofit Area Steam 2
Steam 1 RES RES
Exchanger Network 7
6 Water PA2 Feed PA3 LN RES
Existing System with Maximum
Energy Recovery and Minimum
Additional Exchanger Areas Column Decomposition
and Simulation Optimizer
(Fixed) Existing Diameter
(Fixed) s Figure 1. Optimization requires decomposing the column and coupling the heat-exchanger network using the retroﬁt models.
Key: PA = pump-around; LN = light naphtha; HN = heavy naphtha; LD = light distillate; HD = heavy distillate; and Res = residue. Eqs. 8–11 (14): −c Csb = a − b × exp FLV (8) ρV
ρL (9) FLV = L
V U max = Csb
DT = ρ L − ρV
πUdes (1 − DC ) (10) (11) Fair (14) provides values for the parameters a, b and c
found in Eq. 8. Kister (15) lists ﬂooding correlations for a range of internals and operating conditions. The correlation
and the parameters used should be suited to the existing internals, and yield a reasonable prediction of the entrainment ﬂooding characteristics. In Eq. 11, the design velocity
Udes is less than the ﬂooding velocity Umax by some safety
factor. Typically, Udes is 70–80% of Umax.
Calculation of diameters of the distillation column allows the analysis of the hydraulic performance of the column. The diameter profile along the column is obtained
by plotting the diameters for various stages vs. the stage
number. This profile allows identifying the column bottlenecks that limit throughput enhancement. Column bottlenecks occur on those stages in which the required diameter is larger than the existing one. The diameter calculations allow the existing hydraulic limitations of the
tion column to be considered in the optimization framework. Therefore, during this process, the diameter is calculated for the key stages. Then, the calculated diameters CEP April 2003 www.cepmagazine.org 47 Reactions and Separations Water 2 14 13
1 4 6 8 10 12 3 5 7 9 11 LN 9 11
12 13 6 21 PA3
8 HN 1 22 18 15 7
5 7 PA2
10 6 15
18 25 3
26 5 PA1 16 Steam
4 22 27 9 24 HD Crude 23 5 17 Steam 9
10 Res 28 s Figure 2. The atmospheric crude unit and its heat-exchanger network before optimization was carried out.
Key: PA = pump-around; LN = light naphtha; HN = heavy naphtha; LD = light distillate; HD = heavy distillate; and Res = residue. are compared with the existing ones to guarantee that the
existing diameters are not exceeded; otherwise a penalty
is imposed. Hence, the optimum distillation column will
not have bottlenecks. Retroﬁt model for the heat-exchanger network
This model calculates the required area of the retroﬁtted
heat-exchanger network, while considering the ﬁxed parameters of the network (e.g., heat-transfer areas, duties,
matches, stream splits). The retrofit model and the associated parameters, m and c, are obtained from an extensive
retrofit study on the existing heat exchanger network.
The model, although simple, incorporates the details of
the existing Heat-exchanger network in the process optimization framework. The model allows the benefits of
energy savings to be weighed against the capital investment required to modify the heat-exchanger network.
Details on the model are found in Ref. 12, and these
should aid the reader in performing his or her own analysis.he optimization considers the details of the existing
heat exchanger network simultaneously with the existing
crude distillation column and accounts for the hydraulic 48 www.cepmagazine.org April 2003 CEP constraints of the column. Figure 1 illustrates the approach. The retrofit curve is obtained from an extensive
retrofit study on the heat exchanger network. The model
relates the exchange area required for reducing energy
consumption, Aret, to the reduced energy consumption,
Eret. The model may take different forms; the power law
form has been found suitable for a number of case studies investigated.
A = m( Eret )c (12) The additional area requirement for the retroﬁtted network is related to the existing area of the network, Aexist,
and the total area requirement, A:
Aret = A − Aexist (13) The heat-exchanger retroﬁt model mathematically describes a retroﬁt curve of an existing heat exchanger network (e.g., Figure 1). The retroﬁt curve is a graphical representation of the capital-energy trade-offs in an existing
heat-exchanger network; it consists of a plot of retroﬁt area Liquid
Liquid Vapor Steam Vapor 1. Increase Temperature
Drop Across Pump-around 2. Reduce Steam Flow Vapor
3. Increase Liquid Flow
Through Pump-around 4. Adjust Feed Preheating s Figure 3. Column modiﬁcations that can overcome bottlenecks.
vs. energy demand. The model is based on the network
pinch developed by Asante and Zhu (16).
The amount of heat recovery that can be achieved is
limited by the network pinch. Network pinch analysis
Table. Base case vs. optimum case for constant feed flowrate.
Feed preheat temperature, °C Liquid Base case Optimum case 360 363 PA1 liquid ﬂow, kmol/h 1,228 1,233 PA2 liquid ﬂow, kmol/h 2,396 3,989 PA3 liquid ﬂow, kmol/h 5,868 3,953 PA1 temperature difference, °C 40
50 28.1 PA3 temperature difference, °C 20 58.9 1,200 1,088 The optimization framework
The optimization considers the
details of the existing heat-exchanger
network simultaneously with the existing
crude distillation column and accounts
for the hydraulic constraints of the column. Figure 1 illustrates the approach. Note that the column conﬁgurations before and after optimization are the same.
The overall strategy consists of:
1. Decomposing the crude distillation column into an
equivalent sequence of simple columns, which is simulated using the retrofit model.
2. Modeling the existing heat-exchanger network using
the retroﬁt area model.
3. Simultaneously optimizing the operating conditions
of the existing distillation operation to minimize the sum
of the utility costs and additional exchanger area costs.
4. Taking as fixed, the existing number of stages, the
column configuration and diameters. 44.1 PA2 temperature difference, °C suggests that topology changes (e.g.,
resequencing of exchangers, installation of new exchangers) are necessary
to overcome the limits caused by the
pinch. For no topology changes, the
network is optimized by adding area
to the existing exchanger units. The
white bar under the curve in Figure 1
indicates the additional heat-exchanger capacity needed for the retrofit.
retrofit model simply considers the
total energy demand and total area requirement, which greatly simplifies
the characterization of the heat-exchanger network during process optimization. The model allows the benefits of energy savings to be weighed
against the capital investment required to modify the heat-exchanger
network. Details on the model are
found in Ref. 12 and these should aid
the reader in performing his or her
Note that the column configurations before and after optimization
are the same. Main steam ﬂow, kmol/h
HD-stripper steam ﬂow, kmol/h 260 247 R/Rmin 1.2 1.11 Example
The approach was tried, using data from an actual unit
(Figure 2). The numbers inside of the column represent the
number of stages in each section; note that this notation
differs from that used in Figure 1. The grid in Figure 2 represents the process streams, the heat exchangers connecting them, and any other heaters or coolers, including the CEP April 2003 www.cepmagazine.org 49 Reactions and Separations furnace and steam heaters. The hot streams (i.e., those that
require cooling) run from left to right, while the cold
streams run from right to left. The numbers in the circles of
the heat-exchanger grid stand for individual exchangers in
the network. The vertical lines represent the heat exchange
between process streams. This ﬁgure shows the base case,
that is, before optimization.
Two aims of the retroﬁt design are considered: (1) to improve the energy-efficiency of the process; and (2) to increase the throughput by 20% over the current capacity. The
atmospheric tower is fed 100,000 bbl/d of crude oil. Before
the retroﬁt, the scheme was consuming power at 99.5 MW,
with a total operating cost of $28.4 million/yr. The heat-exchanger network retroﬁt model was found to be:
A = 6.75 × 10 6 Eret −1.61 (14) Literature Cited
1. Sittig, M., “Petroleum Reﬁning Industry Energy Saving and Environmental Control,” Noyes Data Corp., Park Ridge, NJ (1978).
2. Bannon, R. P., and S. Marple, “Heat Recovery in Hydrocarbon
Distillation,” Chem. Eng. Progress, 74 (7), pp. 41–45 (July 1978).
3. Harbert, W. D., “Preﬂash Saves Energy in Crude Unit,” Hydrocarb. Proc., 57 (7), pp. 23–125 (1978).
4. Rivero, R., and A. Anaya, “Exergy Analysis of a Distillation Tower
for Crude Oil Fractionation,” and “Computer Aided Energy Systems
Analysis,” Proc. of Winter Annual Meeting of ASME, 1 (11), pp.
25–30 and 55–62, Dallas, TX (1990).
5. Fraser, A. C., and A. W. Sloley, “Consider Modeling Tools to Revamp Existing Process Units,” Hydrocarb. Proc., 79 (6), pp. 57–63
6. Dhole, V. R., and B. Linnhoff, “Distillation Column Targets,”
Computers Chem. Eng., 17 (5/6), pp. 549–560 (1993).
7. Dhole, V., and P. Buckingham, “Reﬁnery Column Integration for
De-bottlenecking and Energy Saving,” paper presented at ESCAPE
IV Conf., Dublin, Ireland, sponsored by IChemE, Rugby, U.K.
8. Liebmann, K., “Integrated Crude Oil Distillation Design,” PhD thesis, UMIST, Manchester, U.K. (1996).
9. Briones, V., et al., “Pinch Analysis Used in Retroﬁt Design of Distillation Units,” Oil & Gas J., No. 6, pp. 41–46 (June 1999).
10.Bagajewicz, M., et al., “Energy Savings Horizons for Crude Fractionation,” Computers Chem. Eng., 23 (1), pp. 1–9 (1998).
11. Biegler, L. T., et al., “Systematic Methods of Chemical Process Design,” pp. 761–763, Prentice Hall, Englewood Cliffs, NJ (1997).
12.Gadalla, M., “Retroﬁt Design of Heat-Integrated Crude Oil Distillation Systems,” PhD thesis, UMIST, Manchester, U.K. (2003).
13.Suphanit, B., “Design of Complex Distillation Systems,” PhD thesis, UMIST, Manchester, U.K. (1999).
14.Fair, J. R., “How to Predict Sieve Tray Entrainment and Flooding,”
Petro/chem. Engr., 33 (10), pp. 45–62 (1961).
15.Kister, H., “Distillation Design,” McGraw-Hill, New York, Chap. 6
16.Asante, N. D. K., and X. X. Zhu,“An Automated and Interactive
Approach for Heat Exchanger Network Retroﬁt,” Trans. IChemE,
75, Part A, pp. 349–360 (Mar. 1997). 50 www.cepmagazine.org April 2003 CEP Table 1 compares the base case with the optimized arrangement when the feed flowrate is unchanged. The optimum energy consumption of the crude unit is 77.4 MW,
a reduction of 22% and a savings of $6.3 million/yr.
Some investment is needed to improve the performance
of the heat-exchanger network; the energy savings arise
from changing process operating conditions (e.g., furnace inlet temperature, pump-around duties, feed temperature) such that more heat recovery is possible. The
payback for these modifications is 4 mo.
However, to increase the throughput would require
using larger diameters than exist in some sections inside
the column. Therefore, to increase its capacity, the column had to be debottlenecked. The column model and
hydraulic analysis identified the bottlenecked sections.
Proposed modifications that can eliminate the bottleneck
are shown in Figure 3. Parts 1 and 3 in the figure show
increasing the amount of heat removed during a pumparound; Part 2 represents reducing the flow of stripping
steam and, hence, the total vapor flow; and Part 4 illustrates reducing the temperature and, therefore, the vapor
fraction of the column feed. The unshaded arrows that
point toward each other mean that the required diameter
would be decreased.
The optimized distillation unit has a 20% increase in
throughput and requires 94.8 MW of heating. The operating cost saving is $1.9 million/yr, relative to the base
case, with a payback of less than 1 yr.
MAMDOUH GADALLA recently completed his PhD at UMIST, Dept. of Process
Integration (P .O. Box 88, Manchester M60 1QD, U.K. He has four years of
research and teaching experience with the Atomic Energy Authority of
Egypt and the Chemical and Petroleum Engineering Dept. of the United
Arab Emirates Univ. Gadalla designs equipment for the retrofitting of
crude-oil distillation units. He holds a bachelor’s and master’s in chemical
engineering from Cairo Univ.
MEGAN JOBSON is a lecturer in the Dept. of Process Integration at UMIST
(Phone: 44 161 200 4381; Fax: 44 161 236 7439; E-mail:
firstname.lastname@example.org). She carries out research, teaches and
undertakes industrial studies on the synthesis and design of distillation,
absorption and reactive separations. Previously, she worked as a
process engineer in the food industry. She did her undergraduate work
in chemical engineering at the Univ. of Cape Town, South Africa and
holds a PhD in the same field from the Univ. of the Witwatersrand,
Johannesburg, South Africa.
ROBIN SMITH is a professor and head of the Dept. of Process Integration at
UMIST (Phone: 44 161 200 4382; Fax: 44 161 236 7439; E-mail:
email@example.com). He has extensive industrial experience with Rohm &
Haas and ICI. Smith has consulted extensively for process integration
projects. He is widely published in process integration and is the author of
“Chemical Process Design,” (McGraw-Hill). He is a Fellow of the Royal
Academy of Engineering and of the Institution of Chemical Engineers in
the U.K., as well as being a chartered engineer. In 1992, he was awarded
the Hanson Medal of the Institution of Chemical Engineers for his work on
waste minimization. His main research activities include the design of
reaction and separation systems, site utility systems, waste minimization
and water-system design. Smith holds BSc, MSc and PhD degrees in
chemical engineering from the Univ. of Bradford, U.K. ...
View Full Document