Chemical Engineering and Processing 108 (2016) 164–173

Contents lists available at ScienceDirect

Chemical Engineering and Processing:

Process Intensification
journal homepage: www.elsevier.com/locate/cep

Retrofitting an isopropanol process based on reactive distillation and

propylene-propane separation
Mark Wendou Niu, G.P. Rangaiah*
Department of Chemical & Biomolecular Engineering, National University of Singapore, 117585, Singapore

A R T I C L E I N F O A B S T R A C T

Article history:
Received 12 March 2016 In this study, an isopropanol (IPA) process based on reactive distillation (RD) using excess propylene and
Received in revised form 14 July 2016 propylene-propane separation column (splitter) is analyzed for retrofitting. This analysis led to
Accepted 18 July 2016 identifying two limitations of this process, namely, excess propylene feed to RD and requirement of high
Available online 9 August 2016 purity IPA product from RD. To overcome these limitations, two retrofit designs are proposed. One
modified process introduces a new RD with excess water to replace the propylene-propane splitter;
Keywords: simulation results show 14.1% reduction in manufacturing cost per unit product (MCU, decreases from
Process intensification 0.092 $/kg to 0.079 $/kg) for this retrofit design. In the other retrofit design, RD produces an azeotropic
Catalytic distillation
mixture of IPA and water, thus requiring extractive distillation and solvent recovery columns. This
Process retrofitting
modified process leads to a slight increase in MCU (from 0.092 $/kg to 0.095 $/kg). Hence, the design
Isopropyl alcohol
Propylene-propane separation involving a new RD with excess water is recommended for retrofitting the original process.
Extractive distillation ã 2016 Elsevier B.V. All rights reserved.

1. Introduction summarized other advantages of RD as follows: improved

Process intensification (PI), defined as ‘any chemical engineer-
ing development that leads to a substantially smaller, cleaner, safer
and more energy efficient technology’ [18], is an effective strategy
to achieve capital and/or operating cost savings. In recent years, it
is attracting considerable interest from both academia and
industry as a promising way for enhancing process performance,
product quality and adapting to more stringent environmental
regulations. Many PI-based unit operations such as rotating packed
bed (also known as HiGee), dividing-wall column (DWC) and
multi-functional reactors (e.g. reactive distillation, reactive ex-
traction, reactive absorption, reactive adsorption and reactive
crystallization) were studied in the literature [18,17].
One common PI-based unit operation is reactive distillation
(RD). Different from that in a conventional chemical plant, where
separate equipment are linked together by material and energy
streams, in RD, separation and reaction occur in the same vessel
[6]. This results in significant reductions in both energy and
equipment in process systems having appropriate reaction kinetics
and vapor-liquid equilibrium [15]. According to Taylor and Krishna
[24], and Sharma and Mahajani [22], an average 20% savings in
total annual cost (TAC) is achieved by using RD. Reay et al. [18]
* Corresponding author.
E-mail addresses: chegpr@nus.edu.sg, GPRangaiah@gmail.com (G.P. Rangaiah).
conversion and selectivity of the main reaction, increased
reliability, reduction of byproducts, better heat integration and
control of highly exothermic reaction, and reduced problems with
hot spots and thermal runaway.
RD is commonly used for esterification and etherification [26].
The first patent to employ RD dates back to 1920s, when RD was
proposed for homogeneous catalyzed esterification [20]. The most
well-known RD-based esterification is the ‘Eastman Process’ for
methyl acetate. Eastman’s first methyl acetate RD column was
established in 1980, producing over 200,000 metric tonnes of
methyl acetate per year [26]. Owing to its successful implementa-
tion in the industry, many researchers used this system for
experimental and modeling studies on RD. As a result, different
aspects of methyl acetate RD, for instance, characterization of the
catalyst used, effect of operating conditions on the performance
and hydrolysis, have been well studied (e.g., Kolodziej et al. [10])
For etherification, a typical process involving RD includes
synthesis of methyl tertiary butyl ether (MTBE) and ethyl tertiary
butyl ether (ETBE). Baur et al. [2] simulated the RD-based MTBE
process using both the equilibrium stage (EQ) and non-equilibrium
stage (NEQ) models. Their results show that, although both models
predict multiple steady states, the window in which these
multiplicities occur is significantly reduced in the NEQ model.
Another finding from their study is that the hardware design may
have a significant influence on the conversion and selectivity of the
reactions. This influence is further confirmed by Lee and

http://dx.doi.org/10.1016/j.cep.2016.07.013
0255-2701/ã 2016 Elsevier B.V. All rights reserved.
 M.W. Niu, G.P. Rangaiah / Chemical Engineering and Processing 108 (2016) 164–173
Westerberg [12], who introduced a graphical tray-by-tray method
for the same system. This graphical method can help to determine
the optimal location for the reaction zone in RD column and also
the optimal reflux ratio. Results in Lee and Westerberg [12] are in
accordance with the values reported in other papers.
There are some studies on producing multiple products using a
RD. For example, Rix et al. [19] reported 99.9% of isobutene as a
distillate with 98% pure MTBE as a bottom product from the RD
column. Tadé and Tian [23] used a nonlinear inferential model to
simulate a 10-stage ETBE RD pilot plant. Their results from the
model are in good agreement with the experimental data. Besides
esterification and etherification, other prospective applications of
RD include liquid-liquid hydrolysis, oxidation and fermentation
[18].
The literature reviewed above covers most of the classical
examples of RD-based processes. Lutze and Gorak [13] noted that
most of the studies on RD assume high purity reactants and
analyze one-main-product-one-reaction systems. However, real
industrial systems are usually more complex in terms of number of
reactions involved and products generated. To address these,
recent studies focused on complicated systems. González R. et al.
[5] simulated a RD-based tertiary amyl ethyl ether (TAEE) process
with around 20 chemical reactions and 100 components. Following
the pre-defined criteria (e.g. pure component properties, azeo-
tropes and their presence in reactions), certain components were
lumped, and the system reduces to 20 components and 6 chemical
reactions. Even then, good agreement between experimental data
and simulation results was found by González R. et al. [5].
Keller et al. [9] applied RD to a multiple reactions system, where
diethyl carbonate and ethyl methyl carbonate are simultaneously
produced. Their results show an improved selectivity of the
product (diethyl carbonate) compared to a conventional stirred
tank reactor. Holtbruegge et al. [7] reported a study involving both
experiments and simulations for simultaneous production of
dimethyl carbonate and propylene glycol in a pilot-scale RD
column. In these experiments, selected parameters were varied to
test the feasibility of simultaneously producing the two products.
In the simulation, NEQ model was employed to analyze trends of
reactant conversion and product purities. Results in Holtbruegge
et al. [7] suggest the possibility of achieving high propylene
carbonate conversion while recovering an azeotropic mixture of
dimethyl carbonate and methanol in the distillate. Later, Holt-
bruegge et al. [8] suggested a modified process scheme, where a
membrane is added to assist the RD column, and their results show
its potential for decreasing energy required.
It is clear that the topic of RD is not new, and there have been
many studies on RD. Nevertheless, many of them are focused on
grass-roots design. In case of process retrofitting, more constraints
from the process and original design (since existing equipment
should be reused as much as possible) need to be considered. RD
related process retrofitting involves two aspects: one is introduc-
ing RD to replace the conventional reactor and downstream
distillation column, and another is to modify the existing RD
process, especially those with complex systems, so as to improve
its efficiency. Although both aspects are essential, the latter is
novel because, to the best of our knowledge, there is no previous
study to retrofit an existing RD process. Hence, the present paper
analyzes the isopropanol (IPA) process based on RD using excess
propylene and propylene-propane separation column (splitter) to
identify its limitations, and then suggests two retrofit designs. One
retrofit design involves a new RD column using excess water
(instead of excess propylene); this is the second novelty in this
study. Finally, the two retrofit schemes are compared to find the
most promising one.
IPA, also known as isopropyl alcohol and 2-propanol, is a
secondary alcohol with the chemical formula of C3H7OH. It is
165
widely used as a solvent in many industries and finds applications
in food processing, coatings to reduce flammability, as thinner and
additive in paints; it is also used as a disinfectant such as alcohol
wipes in household products and medical applications [4]. IPA is
manufactured by two major commercial processes: indirect and
direct hydration of propylene, of which the latter is common due to
less corrosion in unit operations. Direct hydration of propylene is
an exothermic, reversible reaction carried out with an acid catalyst,
which could be cation-exchange resins such as molybdophos-
phoric acid, titanium and zinc oxides [11]. The main reaction is:
CH3 CH ¼ CH2 þ H2 O Ð ðCH3 Þ2 CHOH ð1Þ
The main side product, diisopropyl ether (DIPE) forms by the
etherification reaction:
2ðCH3 Þ2 CHOH Ð H2 O þ ½ðCH3 Þ2 CH2 O ð2Þ
Conventional IPA process consists of 7 main unit operations,
namely, trickled bed reactor, propylene-propane splitter, lights
column, pre-concentrator, ether column, extractive distillation
column and regenerator for solvent recovery [11]. A novel process
was studied by Xu et al. [28], which only requires two unit
operations, namely, the RD column and the propylene-propane
splitter; rather than 7 units in the conventional process. However,
simulation of splitter is not included in the study of Xu et al. [28].
Recently, Niu and Rangaiah [16] investigated the simulation and
cost estimation of the complete RD-based IPA process, and used it
to retrofit the conventional IPA process. By this retrofitting, around
12% savings is achieved in unit production cost. Even then, cost
estimation showed that operating cost, especially steam consump-
tion, is quite high. Therefore, using the RD-based IPA process as the
base case, this study explores the retrofit designs to reduce
operating cost.
The rest of this paper is organized as follows. Section 2 presents
the IPA process based on RD using excess propylene and propylene-
propane splitter, and its cost estimation; it also includes the
analysis of this process, as the base case, to identify the
contributing units (i.e., unit operations requiring the most utilities)
and other key limitations. Based on this analysis, two retrofit
solutions are proposed, simulated and evaluated in Section 3.
Conclusions of this study are presented in Section 4.
2. Base case of RD-based IPA process
In this section, simulation and analysis of the RD-based IPA
process (consisting of RD column and propylene-propane splitter)
are carried out to identify the key limitations of this process.
2.1. Simulation of the original process
Simulation of the RD-based IPA process is conducted using
Aspen Plus 8.6 and UNIFAC model as the fluid package. Process
flowsheet with key stream data is shown in Fig. 1. As can be seen,
fresh propylene (at 47.6  C and 101.325 kPa, [1] is pressurized to
870 kPa, heated to 13.5  C in a heat exchanger (with assumed
pressure drop of 20 kPa) and then mixed with recycled propylene
in a molar ratio of 2.29:1. This mixed reactant is further pressurized
to 2013.5 kPa and fed to 5th stage of the RD column. Meanwhile,
fresh water is pressurized to 2013.5 kPa and fed to the third stage of
the RD column. The RD column has 28 theoretical stages, and the
reaction zone is from 3rd to 5th stage, with acidic proton-exchange
resin as the catalyst [11]. The product stream (99.99 mol% IPA)
leaves the RD column from the bottom, and then sent to E2 for heat
integration (severing as one heat source for the reboiler of
propylene-propane splitter). The outlet temperature of the product
stream from E2 is set at 35.4  C to ensure sufficient driving force
166 M.W. Niu, G.P. Rangaiah / Chemical Engineering and Processing 108 (2016) 164–173

Fig. 1. RD-based IPA Process Scheme with Key Stream Data.

between the product stream and the inlet stream of propylene-
propane reboiler (25.4  C). On the other hand, the top stream from
the RD column, which mainly consists of unreacted propylene and
propane (inert component), is depressurized to 950 kPa and sent to
the propylene-propane splitter. Thus, the RD column in the base
case process (Fig. 1) is operating under excess propylene condition.
The propylene-propane splitter has 150 theoretical stages (due
to the close boiling point of these components), where propane is
discharged from the bottom and propylene, with a mole fraction of
0.95, leaves at the top. Note that the temperature of the propylene
stream is 13.5  C, which means chilled water is required as the
cooling agent in the condenser. As shown in Fig. 1, heat integration
is applied between the condenser of propylene-propane splitter
and the propylene feed pre-heater, E1. As a result, cost of condenser
duty reduces from $101.5/h to $91.7/h (given in Table 2). Apart from
the heat integration of condenser and E1, heat integration is also
applied between E2 and propylene-propane splitter reboiler,
where the bottom stream is vaporized at a low temperature of
27.4  C. With this heat integration, cost of steam required by the
reboiler is $87.4/h (Table 2), compared to $174.0/h without heat
To verify the simulation results, several results are compared
with those in the literature. For instance, temperature of top and
bottom streams of the RD column are 52.0  C and 186.6  C
compared to their respective values of 49.5  C and 186.5  C in Xu
et al. [28]. Bottom stream pressure is 2013.5 kPa while its value in
Xu et al. is 2000 kPa. Hence, it can be concluded that the simulation
results are acceptable for using as the base case for retrofitting.
2.2. Analysis of base case simulation results
The performance of the base case and retrofitted processes is
evaluated by MCU ($/kg), which is the total annual cost (TAC, $/yr)
divided by annual production rate (kg/year). TAC consists of two
parts, namely, annual production cost (APC) and amortized capital
cost (ACC), where the latter is equal to the total capital investment
(TCI) multiplied by an amortization factor. The following equations
Table 1
Cost Estimation of RD-based IPA Process (Base Case).

integration. Item Cost

Cost estimation for the RD-based IPA process is given in Table 1. Cost of Utilities ($/yr) 10,825,472
It follows the procedure described in Seider et al. [21]. As can be Cost of Operations ($/yr) 943,380
seen, the manufacturing cost per unit product (MCU) is 0.092 $/kg. Cost of Maintenance ($/yr) 1,560,650
Utilities used are steam (LP: 3.5 bar and HP: 31 bar), coolants Cost of Operating Overhead ($/yr) 384,266
Property Taxes and Insurance ($/yr) 339,272
(cooling water and chilled water) and electricity; their unit prices Cost of Depreciation ($/yr) 1,357,087
are $16.9/1000 kg (LP steam) and $22.3/1000 kg (HP steam), $0.02/ General Expenses ($/yr) 1,306,207
1000 kg (cooling water) and $4/GJ (chilled water), as well as $0.06/ Total Production Cost ($/yr) 16,716,333
kWh of electricity [21,27]. Note that the unit price of steam is the Annual IPA Production (kg/yr) 182,264,068
Manufacturing Cost per Unit Product ($/kg) 0.092
average of that in Seider et al. and Turton et al.
 M.W. Niu, G.P. Rangaiah / Chemical Engineering and Processing 108 (2016) 164–173
Table 2
Summary of Utilities Consumed in the Unit Operations and their Costs.
Unit Operation Cost of Steam
Reboiler Duty (kW) Steam Type
RD Column 25478.0 HP
Propylene- 6149.6 LP
Propane Splitter
Cost of Electricity
Unit Operation Power Consumption (kW)
P1
10.5
P2
24.0
P3
10.1
show MCU calculation [21].
MCU ¼ ðCAPC þ a  CTCI Þ=P
CAPC ¼ C U þ C OP þ C M þ C OPO þ C P þ C D þ C GEN
CTCI ¼ C BM þ C S þ C C&C þ C L þ C SC þ C W
i  ð1 þ iÞn
a¼
ð1 þ iÞn  1
In Eq. (3), a is the amortization factor (1/year), and P is the
annual production rate of the desired product (kg/year). In Eq. (4),
CU is the cost of all utilities (steam, electricity and coolants, $/year)
used; COP is the cost of operations ($/year); CM refers to cost of
maintenance ($/year); COPO, CP and CD are respectively cost of
operating overhead ($/year), cost of property, taxes and insurance
($/year), and cost of depreciation ($/year); and CGEN represents
general expenses ($/year). In Eq. (5), CBM is the bare module cost
($); CS, CC&C and CL represent site and service cost ($), contingencies
and contractor’s fee ($), and land cost ($), respectively; CSC is the
startup cost ($) while CW is the working capital ($). In this study,
CBM is estimated using cost correlations in Seider et al. (2010), and
then updated to Chemical Engineering Plant Cost Index (CEPCI) of
576; CS, CC&C, CL and CSC can be calculated from CBM (although CL is
zero in most retrofit cases); and finally CW is related to production,
once the production is fixed, CW becomes a fixed value. In Eq. (6), n
is the plant life (years), and i is the expected annual return ($/($.
year)); assuming n = 15 and i = 0.15, a = 0.171, which is used in this
study.
According to Seider et al. [21], COP, COPO and CGEN in Eq. (4) are
related to direct wages and benefits, and they are assumed to be
fixed in this study, any discussion of reducing such costs are
beyond the scope of this study. Moreover, CM and CP are in
proportion to CTCI, which is a fixed value in the base case. Therefore,
CM and CP are both fixed, and no need to conduct further analysis.
Hence, analysis of simulation results focuses on CAPC, and CTCI is
considered for retrofit designs involving new unit operations. After
screening, CU is the only term requiring further investigation.
According to Fig. 1, utilities consumed in the IPA process are LP and
MP steam, cooling and chilled water, and electricity.
In the base case, the plant is assumed to be in operation for
many years and that its amortized capital investment (a CTCI) is
zero, and only CAPC contributes to MCU. Table 2 presents utility cost
for all the unit operations in the base case process. Each item in this
table is a contributor (i = 1 to n) to CU; a less important contributor
167
Cost of Cooling Agents
Steam Cost ($/h) Condenser Duty (kW) Cooling Agent Cooling Cost($/h)
1140.0 18128.8 Cooling Water 31.2
87.4 6366.8 Chilled Water 91.7
Electricity Cost ($/h)
0.63
1.44
0.61
could be neglected if it satisfies:
ð3Þ di < 0:08dmax i ¼ 1; 2; . . . ; n; and i 6¼ maxÞ ð7Þ
A conservative factor of 0.08 is used in this equation as
ð4Þ compared to 0.1 suggested by [3]. This is mainly to avoid missing
any possible contributor. As can be seen from Table 2, reboiler
steam cost of RD column has the highest utility consumption
ð5Þ among all the contributors; hence, it is set as the reference to
evaluate the remaining ones. Following Eq. (7), it is clear that cost
of cooling agent for RD column and cost of electricity for all pumps
(P1–P3) are negligible. In contrast, besides for HP steam cost in the
ð6Þ RD column, this RD-based IPA process has high steam cost and high
cooling cost for propylene-propane splitter.
After figuring out the contributors, it is essential to identify the
limitations related to the identified contributors and also the
causal relationship among different contributors. Steam consump-
tion in the RD column is firstly considered; as can be seen in Fig. 1,
mole ratio of propylene to water in the RD inlet streams is 1.43:1,
whereas the stoichiometric ratio of these two reactants is 1:1
(Eq. (1)). Excess propylene (mixed with propane), although
unreacted, is also evaporated and raised to the reaction tempera-
ture, resulting in high steam consumption in the reboiler of RD.
Actually, excess propylene is inevitable as long as pure IPA (99.99
mol%) is required from the bottom of the RD column. This is
because IPA and water form an azeotrope of 87.4 wt% alcohol,
having a boiling point of 80.3  C at atmospheric pressure [11]. On
the other hand, it can be seen that 174.4 kmol/h propylene-propane
mixture (with propylene mole fraction: 0.95) is recycled out of the
RD column inlet flow rate of 574.4 kmol/h (with propylene mole
fraction: 0.95), suggesting propylene conversion of only 69.6%.
Considering propylene is a more expensive reactant compared to
water, this conversion is relatively low and has potential to be
improved.
Propylene-propane splitter is then considered. As can be seen in
Table 2, chilled water is required for the condenser of this column.
This is because the boiling points of propylene and propane are
close to each other, requiring depressurization of the mixture from
1750 kPa to 950 kPa [11] for improving separation by distillation.
This lower pressure results in a low outlet temperature (13.5  C) of
the top stream.
Fig. 2 summarizes the observed limitations and the causal
relationship between them. As can be seen, ‘high separation cost in
C3H6-C3H8 splitter’ is resulting from ‘excess propylene feed to RD’
and ‘close boiling points of propylene and propane’; ‘high steam
cost in RD column’ is partly caused by ‘large amount of propylene
evaporated in RD’; and ‘high purity IPA required from RD’
contributes to ‘excess propylene feed to RD’. Generally, process
modifications should be proposed to address the factors causing
168 M.W. Niu, G.P. Rangaiah / Chemical Engineering and Processing 108 (2016) 164–173

Fig. 2. Links between Limitations and the Factors Causing Them.

the limitations. Therefore, in the RD-based IPA process, modifi-
cations on any of the following factors, namely, ‘excess propylene
feed to RD’, ‘close boiling points of propylene and propane’, ‘large
amount of propylene evaporated in RD’ and ‘high purity IPA
required from RD’ are expected to improve the process perfor-
mance. On the other hand, ‘close boiling points of propylene and
propane’ is a physical phenomenon, and ‘large amount of
propylene evaporated in RD’ is due to the quantity of propylene
feed to the RD column. Hence, factors available for addressing
reduces from four to two, namely, ‘high purity IPA required from
RD’ and ‘excess propylene feed to RD’ (shown in dashed box in
Fig. 2). Thus, they are the starting point for exploring the retrofit
designs in the next section.
3. Retrofit designs and their comparison
Using the analysis summarized in Fig. 2, two retrofit designs are
proposed to overcome the process limitations; both designs do not
require the propylene-propane splitter. The first retrofit scheme is
aimed at reducing the amount of propylene feed to the RD column;
the core idea in this scheme is to split the fresh propylene feed into
two streams, where one of them still goes to the original RD
column while the other stream is sent to a new RD column. The
second retrofit scheme shifts the (final) product stream from the
original RD column to another unit operation, where an extractive
distillation column (azeotropic distillation) and a regenerator are
introduced. In this retrofit solution, instead of pure IPA, IPA-water
mixture leaves the RD column, and the extractive distillation takes
care of IPA purification. The proposed two retrofit designs are
discussed in detail in this section.
3.1. Retrofit Design with a New/Second RD Column
The retrofitted process with a new/second RD column (RD-W,
where water is fed in excess compared to propylene) and its
simulation results are shown in Fig. 3. As can be seen, fresh
propylene liquid (47.6  C and 101.325 kPa, [1] is pressurized to
2033.5 kPa, heated to 24.2  C in a heat exchanger (with assumed
pressure drop of 20 kPa) and then split into two streams. The first
stream is fed to the RD column operating under excess propylene
condition (RD-P column), where 99.99 mol% IPA is collected as the
bottom product, and the gaseous mixture discharged from the top
mainly consists of unreacted propylene and propane. This gas
stream, after condensation to liquid at 52.1  C and re-pressuriza-
tion to 2013.5 kPa, is mixed with the remaining fresh propylene
feed and then fed to the RD-W column (which is introduced to
replace the propylene-propane splitter in the base case); another
reactant (water) is also fed to this column. Note that the RD-W
column operates under excess water condition and converts most
of propylene entering into it.

Fig. 3. Process Retrofit Scheme with a Second RD Column (RD-W) in place of Propylene-Propane Splitter.
 M.W. Niu, G.P. Rangaiah / Chemical Engineering and Processing 108 (2016) 164–173
After reaction and separation, the gas mixture containing
mostly propane from propylene feed is discharged from the top of
RD-W column. As can be seen in Fig. 3, small amount of propylene
(7 kmol/h out of 380 kmol/h entering the process) is lost in the
propane purge from the top of RD-W column. The azeotropic
mixture of IPA and water in the bottom stream is sent to the RD-P
column to react with propylene feed. The retrofit design in Fig. 3
is strategically developed such that the original RD column can be
reused and only one new RD column is required. Design
constraints to fulfill this requirement are discussed in the later
part. Further, plant space occupied by the propylene-propane
splitter can be used for the RD-W column. The existing propylene-
propane splitter can be re-used elsewhere or sold; this possible
benefit from retrofitting the base case is not considered in the
analysis.
As discussed in Section 2.2, the retrofitted process in Fig. 3 is
evaluated by MCU ($/kg). While generating the retrofit design,
process optimization is conducted to find the optimal retrofit
design under this scenario. As can be seen in Fig. 3, the new RD-W
column is interlinked with the original RD-P column. To avoid
convergence problems in the simulation of these two RD
columns, simplification is necessary for optimization. For this,
based on preliminary simulations and analysis, CTCI is calculated
and fixed during optimization. Specifically, split fraction of fresh
propylene stream going to RD-P column (SPRD-P) is varied from 0
to 1 (in steps 0.05 and 0.001 near the optimal value), CTCI is
calculated for all solutions satisfying the process constraints (Eqs.
(11)–(14) given later), and the mean value of CTCI of all feasible
solutions is used. This simplification is reasonable because,
according to Table 2, energy consumption (contributing to CAPC)
are large in the base case. Once CTCI is calculated and fixed, MCU is
determined by CAPC, which is the sum of CU, COP, CM, COPO, CP and
CD (Eq. (4)). Following the definition of each of them in
Section 2.2, COP, CM and COPO are fixed once the unit operations
in the process are determined; CP and CD are directly in
proportion to CTCI and hence they are fixed once CTCI is fixed as
outlined above. Therefore, CAPC now varies solely with CU. In other
words, after simplification, optimization of CU is likely to lead to a
near optimal retrofit design.
For minimization of CU, decision variables are chosen after
analyzing the degrees of freedom of the relevant unit operations,
and sensitivity to the total energy consumption. In the proposed
retrofit scheme, after analysis, distillate rate of RD-P column (DRD-
P), reflux ratio of RD-W Column (RRD-W) and SPRD-P are selected as
decision variables. While other important operating parameters
such as reflux ratio of RD-P column (RRD-P), feed flow rate of water
(FH2 O ), and distillate rate of RD-W column (DRD-W) are used to fulfill
the process requirements.
Constraints in the optimization are based on process knowledge
and retrofitting considerations. Bounds on decision variables and
process constraints are:
DRDP  1:1  D0RD ð8Þ
RRDW > 0 ð9Þ
0 < SPRDP < 1 ð10Þ
Eq. (8) sets the upper bound on the distillate rate of RD-P
column, which should not exceed 10% of its original value in the
base case (D0RD ) [25]. Eqs. (9) and (10) ensure a positive value for
the reflux ratio of RD-W Column and SPRD-P is between 0 and 1,
respectively. Further, in order to reuse the original RD column as
the RD-P column in the retrofitted process, the following
169
constraints are included:
0
FRDP
IPA  1:1  FRD
IPA ð11Þ
0
FAZOMIX  1:1  FRD
H2 O ð12Þ
0
FRDP RD
TOP  1:1  FTOP ð13Þ
PRDP  1:1  P0RD ð14Þ
In Eq. (11), FRDP
IPA refers to flow rate of the product stream
(99.99 mol% IPA) from the RD-P column, and it should be within
110% of that from the RD column in the base case. Similarly,
FAZOMIX , flow rate of IPA-water mixture from the bottom of the RD-
W column and feed into RD-P column as one reactant, should be
within 110% of the flow rate of water feed into the original RD
column; this requirement is expressed in Eq. (12). Eq. (13)
guarantees flow rate of the top stream from the RD-P column
(FRDP
TOP ) in the retrofit scheme does not exceed that in the original
RD from the base case by more than 10%. Finally, Eq. (14) constrains
the operating pressure of RD-P (PRDP ) within 110% of that of the
original RD column (P0RD ). In the retrofitted process, where feeds,
product and reactions remain the same/similar to those in the
original process, the operating temperature is expected to be
within the safe operating range. Hence, there is no constraint on
the operating temperature. Finally, equality constraints, which
include mass and energy balances for each unit operation in the
process, are taken care by Aspen Plus simulation and its
convergence.
Results from the simplified optimization along with using the
built-in optimizer show that the optimal SPRD-P = 0.745, optimal
DRD-P = 100.9 kmol/h and optimal RRD-W = 5.1, and the correspond-
ing MCU is 0.079 $/kg. Cost estimation of the retrofit scheme
compared to the base case is summarized in Table 3. In the
retrofitted process, utilities used are steam (MP: 10.3 bar and HP:
31.0 bar), cooling water and electricity. Unit price of MP steam is
$19.5/1000 kg [21,27], and unit prices of the other utilities are the
same as those in the base case. As can be seen in Table 3, owing to
reusing the original RD column, the amortized capital cost of the
new unit operation (RD-W column) is merely 290,309 $/yr
whereas utility cost decreases from 10,825,472 $/yr to
7,527,872 $/yr, mainly due to reduced steam consumption. Given
these numbers, it is clear that savings of 3,297,600 $/yr from utility
consumption overrides the capital expense of RD-W column for
retrofitting. As a result, MCU drops from 0.092 $/kg to 0.079 $/kg,
achieving a saving of 14.1%.
Moreover, feeding part of fresh propylene into the RD-W
column reduces overall steam consumption, which indicates that
vaporizing large amount of unreacted propylene in the RD column
is the leading cause of high energy consumption and key process
limitation in the base case. Thus, the second RD (RD-W) column
eliminates this limitation and improves the process performance
substantially.
Note that annual IPA production in the base case is slightly
different from that in the retrofit design (Table 3). This is mainly
because some propylene (1.85% of total propylene feed) is lost
together with propane from the top of RD-W column in the retrofit
design (without the propylene-propane splitter). This propylene
loss is considered in the cost estimation for the retrofit scheme
(Table 3) as follows: amount of propylene discharged is
2,356,800 kg/year, unit price of propylene ($ 0.639/kg; [1]), and
hence cost of propylene loss is 1,585,188 $/year.
170 M.W. Niu, G.P. Rangaiah / Chemical Engineering and Processing 108 (2016) 164–173
Table 3
Cost Estimation of the Process Retrofitted with a Second RD (RD-W) Column Compared to the Base Case.
Item
Capital Investment for Added Unit Operations ($)
Cost of Utilities ($/yr)
Cost of Operations ($/yr)
Cost of Maintenance ($/yr)
Cost of Operating Overhead ($/yr)
Property Taxes and Insurance ($/yr)
Cost of Depreciation ($/yr)
General Expense ($/yr)
Cost of Propylene Loss ($/yr)
Cost of Water Reduced/Increased ($/yr)
Summary
Total Production Cost ($/yr)
Amortized Capital Investment ($/yr)
Annual IPA Production (kg/yr)
Manufacturing Cost per Unit Product ($/kg)
3.2. Retrofit design with an extractive distillation column and a
regenerator
The retrofit process involving an extra RD column, an extractive
distillation column and a regenerator is simulated using Aspen Plus
8.6, and the simulation results with key process data are presented
in Fig. 4. As can be seen, liquid fresh propylene (47.6  C and
101.325 kPa, [1], after pressurizing to 2033.5 kPa and heating to
24.2  C, is split into two streams, where one of them goes to the RD-
1 column (the original RD using excess propylene) and the other to
the RD-2 column (new RD column also using excess propylene).
The reason for splitting the fresh propylene is to re-use existing RD
column while ensuring that it meets the new design specifications
(e.g. column diameter and liquid level on each tray) when excess
water is used. Accordingly, split ratio of propylene is optimized
later. Meanwhile, excess water stream, after pressuring to
2013.5 kPa, is fed to both RD-1 and RD-2 in water to propylene
Fig. 4. Process Retrofit Scheme with an Extractive Distillation Column and a Regenerator.
Base Case Retrofitted Process with a Second RD (RD-W) Column
0 1,697,714
10,825,472 7,527,872
943,380 943,380
1,560,650 1,031,769
384,266 318,731
339,272 224,298
1,357,087 897,191
1,306,207 1,282,376
0 1,585,188
0 (195)
16,716,333 13,810,610
0 290,309
182,264,068 179,040,785
0.092 0.079
ratio of 2.05. After reaction and separation, the gas mixture
containing mostly propane from the propylene feed stream is
discharged from the top of both RD columns. On the other hand, as
water is fed in excess, bottom product from both the columns is a
mixture of IPA and water. Note that DIPE is minimized through
choosing appropriate column operating conditions. The bottoms
streams from RD-1 and RD-2 are mixed, depressurized to 135 kPa,
and then sent to the downstream units for separation.
The downstream separation units consist of the azeotropic
distillation column (Extractive distillation column) and the
regeneration column (Regenerator). The former is introduced
because IPA and water form an azeotrope; to break this azeotrope,
an entrainer, namely, dimethyl sulfoxide (DMSO) is employed [14].
The regenerator is used to recover the entrainer by separating/
removing water. Simulation results of the extractive distillation
column and regenerator are shown in Fig. 4; IPA-water mixture
(with IPA mole fraction of 0.455) is fed to the 26th stage of the
 M.W. Niu, G.P. Rangaiah / Chemical Engineering and Processing 108 (2016) 164–173
extractive distillation column; while the recycled (lean entrainer/
DMSO), after cooling to 71.8  C [14], enters at 6th stage of the
extractive distillation column. IPA with molar purity of 0.9999 is
collected from the extractive distillation column top as the main
product, whereas practically pure DMSO (0.99999 mol fraction)
leaving the column bottom is returned to the regenerator. Make-up
solvent in very small quantity is added to recycle DMSO.
The retrofitted scheme in Fig. 4 is optimized following the steps
and simplification outlined in Section 3.1. Decision variables for
optimization are chosen after analyzing the degrees of freedom of
the unit operations, and their sensitivity to the total energy
consumption. In this retrofitted process, decision variables are split
fraction of the fresh propylene stream going to RD-1 (SPRD-1),
distillate rate of RD-1 (DRD-1), reflux ratio of RD-2 (RRD-2), reflux
ratio of extractive distillation column (RExtr) and reflux ratio of
regenerator (RReg). Other important operating parameters: feed
flow rate of water for both RD-1 and RD-2 (FH2 O1 andFH2 O2 ),
distillate rate of RD-2 (DRD-2), distillate rate of extractive
distillation column (DExtr) and of regenerator (DReg), and reflux
ratio of RD-1 (RRD-1) are used to fulfill the process requirements.
For instance, RExt is used to meet the purity specification for IPA.
Bounds on decision variables and process constraints are as
follows:
0 < SPRD1 < 1 ð15Þ
DRD1  1:1  D0RD ð16Þ
RRD2 > 0 ð17Þ
RExtr > 0 ð18Þ
RReg > 0 ð19Þ
Here, Eq. (15) ensures SPRD-1 is between 0 and 1. Eq. (16)
provides the upper bound of the distillate rate of RD-1 column;
similar to that in Eq. (8), this rate should not exceed 110% of its
original value (D0RD ) [25]; Eqs. (17)–(19) ensure positive reflux
ratios.
To ensure reuse of the original RD column as RD-1 in the
retrofitted process, the following constraints are considered:
0
FRD1 RD
VG1  1:1  FTOP ð20Þ
0 Table 4
FRD1 RD
H2 O1  1:1  FH2 O ð21Þ
0
FRD1 RD
BTM  1:1  FIPA ð22Þ
PRD1  1:1  P0RD ð23Þ
In Eq. (20), FRD1
VG1 refers to the flow rate of top stream of RD-1,
and it should not exceed 110% of flow rate of that in the original RD
column from the base case. Also, flow rate of water, which feeds to
RD-1 as excess reactant following retrofitting, should not exceed
110% of water feed into the original RD column in the base case.
This restriction is expressed in Eq. (21). Similarly, Eq. (22)
guarantees flow rate of the bottom stream in RD-1 column does
not exceed 110% of that in the IPA stream from the base case.
Eq. (23) constrains the operating pressure of RD-1 (PRD1 ) within
171
110% of that of the original RD column (P0RD ). For the reason stated
for the previous retrofit scheme, there is no constraint on operating
temperature. Finally, equality constraints, which include mass and
energy balances for each unit operation in the process, are taken
care by Aspen Plus simulation and its convergence.
Results from the simplified optimization along with using the
built-in optimizer show that the optimal SPRD  1 = 0.506, optimal
DRD  1 = 14.03 kmol/h, optimal RRD  2 = 25.0, optimal RExtr = 0.787
and optimal RExtr = 0.451, and the corresponding MCU is 0.095 $/kg.
Cost estimation of the current retrofit scheme compared to the
base case is summarized in Table 4. In the retrofitted process in
Fig. 4, utilities used are steam (MP: 10.3 bar and HP: 31.0 bar),
cooling water and electricity. Unit prices of these utilities have
been provided in the previous sections. As can be seen in Table 4,
the three new columns (RD-2, extractive distillation column and
regenerator) totally cost $ 9,700,884, which equals to amortized
capital cost of 1,658,851 $/yr; in addition, operations cost increases
(3rd row in Table 4) from 943,380 $/yr to 1,824,260 $/yr. As a result,
savings from the utility cost (3,815,430 $/yr or 35.2%) is not able to
cover the extra cost, leading to a slight increase in MCU from
0.092 $/kg to 0.095 $/kg. Note that energy recovered from
propylene feed (818.6 kW in E1) is assumed to be used to produce
chilled water, and the revenue from the chilled water produced is
deducted from the total utility cost. On the other hand, in the
current retrofit, propylene loss is 2,303,183 kg/yr (1.8% of fresh
propylene feed); following the similar calculation in the previous
retrofit scheme, cost of this propylene loss is 1,549,194 $/yr. This is
also included in MCU in Table 4.
3.3. Comparison of the retrofit designs
The two retrofit schemes proposed, namely, the retrofit design
with a new/second RD (RD-W) column, and the retrofit design with
an extractive column and a regenerator, are investigated in
Sections 3.1 and 3.2, respectively. In the former, capital investment
for retrofitting is small (with amortized capital cost of only
290,309 $/yr), and this is only 17.5% of that in the latter retrofit
scheme (with amortized capital cost as high as 1,658,851 $/yr).
Further, when more unit operations involved, cost of operations
tend to be higher, this cost is 943,380 $/yr for the former retrofit
and 1,824,260 $/yr for the latter retrofit. In terms of utility
consumption, both retrofit designs achieve significant savings;
the former reduces utility cost by 30.5% while the latter brings
35.2% savings. Another contributing factor to MCU (Tables 3 and 4)
is the cost of propylene loss. Both the retrofit schemes incur small
and similar loss of 1.85% of fresh propylene feed, whose cost is
Cost Estimation of the Process Retrofitted with an Extractive distillation column and
a Regenerator Compared to the Base Case.
Item Base Case RD-ED
Capital Investment for Added Unit Operations ($) 0 9,700,884
Cost of Utilities ($/yr) 10,825,472 7,010,042
Cost of Operations ($/yr) 943,380 1,824,260
Cost of Maintenance ($/yr) 1,560,650 1,502,410
Cost of Operating Overhead ($/yr) 384,266 567,931
Property Taxes and Insurance ($/yr) 339,272 326,611
Cost of Depreciation ($/yr) 1,357,087 1,306,443
General Expense ($/yr) 1,306,207 1,283,134
Cost of Propylene Loss ($/yr) 0 1,549,194
Cost of Water Reduced/Increased ($/yr) 0 12,635
Summary
Total Production Cost ($/yr) 16,716,333 15,382,659
Amortized Capital Investment ($/yr) 0 1,658,851
Annual IPA Production (kg/yr) 182,264,068 179,155,696
Manufacturing Cost per Unit Product ($/kg) 0.092 0.095
172 M.W. Niu, G.P. Rangaiah / Chemical Engineering and Processing 108 (2016) 164–173
Table 5
Steam cost for unit operations in the base case and two retrofit designs.
Base Case RD-P&RD-W
Unit Operation Steam Cost ($/h) Unit Operation
RD Column 1140 RD-P Column
Propylene- propane Splitter 87.4 RD-W Column
Total 1227.4
Table 6
Status of the Columns in the Two Retrofit Processes.
RD-P&RD-W Column Status RD-ED Column Status
RD-P Column Existing Unit RD-1 Column (RD-P) Existing Unit
RD-W Column New RD-2 Column (RD-P) New
\ \ ED column New
\ \ Regenerator New
1.6 million $/yr compared with no propylene loss in the base case.
In summary, the retrofit scheme with a new/second RD reduces
MCU from 0.092 $/kg to 0.079 $/kg (or 14.1%), and the scheme with
an extractive distillation column and regenerator increases MCU
slightly to 0.095 $/kg (or 3.2%) despite significant reduction in
utility cost.
From Section 2.2, it is clear that cost of steam is the main
contributing factor to the high MCU in the base case. To ensure that
this key limitation has been addressed in the retrofit scheme,
steam cost for unit operations in the base case and two retrofit
designs is summarized in Table 5. As can be seen, cost of reboiler
duty in the RD-P column is reduced 854 $/h, as compared to that in
the original column (1140 $/h). The most significant reduction is in
RD-ED scheme; with water in excess, steam cost is only 217.0 $/h
for the two RD columns. However, this low cost is offset by steam
cost in the ED column and regenerator, making the overall steam
cost of RD-ED scheme (853 $/h) comparable to that in RD-P&RD-W
scheme (914.9 $/h).
Finally, there is one more factor not shown in either Tables 3 or
4 but need to be considered when making the decision on the
retrofitting. Table 6 summarizes the status of unit operations for
both retrofit schemes. In the retrofit design with a second RD
column, only one new unit operation (RD-W column) is introduced
to replace the propylene-propane column. Therefore, there is
probably no need for extra land. In the alternate retrofit design,
there are three new columns involved; assuming one column can
be built in the site of the propylene-propane column eliminated,
enough space is required for the other two columns, which
increases the difficulty in implementing this retrofit scheme. In
summary, the retrofit design with a new/second RD column (RD-
W) is recommend for modifying the base case process.
4. Conclusions
In this study, a RD-splitter based IPA process is investigated and
its two key limitations, namely, ‘high purity IPA required from RD’
and ‘excess propylene feed to RD’ are identified through process
simulation, analysis and cost estimation. To improve this process,
two retrofit designs are proposed accordingly; they are retrofit
design with a new RD (RD-W) column, and the retrofit design with
an extractive distillation column and a regenerator. The former
scheme has an amortized capital cost for retrofitting of only
290,309 $/yr, and also reduces MCU from 0.092 $/kg to 0.079 $/kg
(or 14.1%). The latter scheme has a significantly large amortized
RD-ED
Steam Cost ($/h) Unit Operation Steam Cost ($/h)
854.0 RD-1 Column 110.0
60.9 RD-2 Column 107.0
ED Column 230.0
Regenerator 406.0
914.9 853.0
capital cost (1,658,851 $/yr) for retrofitting, increases MCU from
0.092 $/kg to 0.095 $/kg, and also requires additional site area for
two columns. Hence, the retrofit design with a new RD-W column
is recommend for modifying the original process.
Acknowledgments
The first author thanks the National University of Singapore for
supporting his doctoral research. Both authors acknowledge
Yonghui Lin for the initial study on reactive distillation options
for the IPA process.
References
[1] American Chemistry Council, Propylene: Production Stewardship Guidance
Manual, (2007) . accessed 16.11.15 http://www.lyondellbasell.com/techlit/
techlit/3303.pdf.
[2] R. Baur, A.P. Higler, R. Taylor, R. Krishna, Comparison of equilibrium stage and
nonequilibrium stage models for reactive distillation, Chem. Eng. J. 76 (1)
(2000) 33–47.
[3] J.M. Douglas, Cost diagrams and the quick screening of process alternatives,
Ind. Eng. Chem. Res. 24 (1985) 970–976.
[4] P. Dutia, Isopropyl Alcohol: A Techno-commercial Profile, (2012) . accessed
16.11.15 http://www.chemicalweekly.com/Profiles/Isopropyl_Alcohol.pdf.
[5] C.A. González R, T. Keller, J. Pilarczyk, W. Sałacki, A. Górak, TAEE synthesis from
isoamylenes and ethanol by catalytic distillation: pilot plant experiments and
model validation, Fuel Process. Technol. 102 (2012) 1–10.
[6] D.W. Green, R.H. Perry, Perry’s Chemical Engineers’ Handbook, 8th ed.,
McGraw-Hill Companies, Inc., 2007.
[7] J. Holtbruegge, S. Heile, P. Lutze, A. Górak, Synthesis of dimethyl carbonate and
propylene glycol in a pilot-scale reactive distillation column: experimental
investigation, modeling and process analysis, Chem. Eng. J. 234 (2013) 448–
463.
[8] J. Holtbruegge, M. Wierschem, P. Lutze, Synthesis of dimethyl carbonate and
propylene glycol in a membrane-assisted reactive distillation process: pilot-
scale experiments, modeling and process analysis, Chem. Eng. Process.:
Process Intensif. 84 (2014) 54–70.
[9] T. Keller, B. Dreisewerd, A. Górak, Reactive distillation for multiple-reaction
systems: optimisation study using an evolutionary algorithm, Chem. Process
Eng. 34 (1) (2013) 17–38.
[10] A. Kolodziej, M. Jaroszynski, A. Hoffmann, A. Gorak, Determination of catalytic
packing characteristics for reactive distillation, Catal. Today 69 (1–4) (2001)
75–85.
[11] J.I. Kroschwitz, Isopropyl Alcohol in Kirk-Othmer Encyclopedia of Chemical
Technology, 5th ed., Wiley-Interscience, New York, 2004.
[12] J.W. Lee, A.W. Westerberg, Graphical design applied to MTBE and methyl
acetate reactive distillation processes, AIChE J. 47 (6) (2001) 1333.
[13] P. Lutze, A. Gorak, Reactive and membrane-assisted distillation: recent
developments and perspective, Chem. Eng. Res. Des. 91 (10) (2013) 1978–1997.
[14] W.L. Luyben, I.L. Chien, Design and Control of Distillation Systems for
Separating Azeotropes, John Wiley & Sons, 2011.
[15] W.L. Luyben, C.C. Yu, Reactive Distillation Design and Control, John Wiley &
Sons, 2008.
[16] W.M. Niu, G.P. Rangaiah, Process retrofitting via intensification: a heuristic
methodology and its application to isopropanol process, Ind. Eng. Chem. Res.
55 (12) (2016) 3614–3629.
[17] D.P. Rao, Process intensification in process retrofitting and revamping, in: G.P.
Rangaiah (Ed.), Chemical Process Retrofitting and Revamping: Techniques and
Applications, John Wiley, Chichester, 2016.
[18] D. Reay, C. Ramshaw, A. Harvey, Process Intensifi Cation: Engineering for
Efficiency, Sustainability and Flexibility, Butterworth-Heinemann, Oxford,
2013.
[19] A. Rix, G. Grund, W. Bueschken, Process for preparing highly pure raffinate II
and methyl tert-butyl ether, US Patent 6 (657) (2003) 090.
 M.W. Niu, G.P. Rangaiah / Chemical Engineering and Processing 108 (2016) 164–173
[20] J.G. Segovia-Hernández, S. Hernández, A.B. Petriciolet, Reactive distillation: a
review of optimal design using deterministic and stochastic techniques, Chem.
Eng. Process.: Process Intensif. 97 (2015) 134–143.
[21] W.D. Seider, J.D. Seader, D.R. Lewin, S. Widagdo, Product and Process Dsign
Principles: Synthesis, Analysis and Evaluation, 3rd ed., John Wiley & Sons,
2010.
[22] M.M. Sharma, S.M. Mahajani, Reactive Distillation: Status and Future
Directions, Wiley-VCH, Weinheim, 2002.
[23] M.O. Tadé, Y.C. Tian, Conversion inference for ETBE reactive distillation, Sep.
Purif. Technol. 19 (1) (2000) 85–91.
[24] R. Taylor, R. Krishna, Modelling reactive distillation, Chem. Eng. Sci. 55 (2000)
5183–5229.
173
[25] G. Towler, R. Sinnott, Chemical Engineering Design: Principles Practice and
Economics of Plant and Process Design, Butterworth-Heinemann, Oxford,
2013.
[26] A. Tuchlenski, A. Beckmann, D. Reusch, R. Dussel, U. Weidlich, R. Janowsky,
Reactive distillation: industrial applications: process design and scale-up,
Chem. Eng. Sci. 56 (2001) 387–394.
[27] R. Turton, R.C. Bailie, W.B. Whiting, J.A. Shaeiwitz, D. Bhattcharyya, Analysis
Synthesis and Design of Chemical Processes, 4rd ed., Prentice Hall, NJ, 2012.
[28] Y. Xu, K.T. Chuang, A.R. Sanger, Design of a process for production of isopropyl
alcohol by hydration of propylene in a catalytic distillation column, Trans.
IChemE 80 (2002) 686–694 Part A.