Computers and Chemical Engineering 93 (2016) 118–127

Contents lists available at ScienceDirect

Computers and Chemical Engineering

journal homepage: www.elsevier.com/locate/compchemeng

Economic trade-offs in acrylic acid reactor design

William L. Luyben
Department of Chemical Engineering, Lehigh University, Bethlehem, PA, 18015, USA

a r t i c l e i n f o a b s t r a c t

Article history: The acrylic acid process using air oxidation of propylene presents many interesting design trade-offs,
Received 22 February 2016 particularly in the design of the reactor. The desired and undesired reactions are highly exothermic and
Received in revised form 7 June 2016 very temperature dependent (large activation energies), so a large flowrate of inert water is also fed to the
Accepted 9 June 2016
reactor to act as a thermal sink. Propylene conversion increases with temperature and reactor size, but
Available online 11 June 2016
acrylic acid yield decreases with increasing temperature. The heat of reaction is removed by generaing
steam, and the steam pressure is an important design optimization variable since it sets low limits on
Keywords:
reactor temperature. Using low-pressure steam gives high acrylic acid yield and lower carbon dioxide
Acrylic acid
Reactor design
generation but requires large reactors. Larger air flowrates increase reactor oxygen concentrations, which
Reactor design trade-offs reduce reactor volume but increase air compression costs.
This paper explores the effects of the many design trade-offs on capital investment, energy cost and
product selectivity.
© 2016 Elsevier Ltd. All rights reserved.

1. Introduction selectivity, required reactor size, heat-exchanger size and optimum

Acrylic acid is a commodity chemical that is widely used in the
polymers industry. The commercial petroleum-based process uses
propylene with air providing the oxygen to form acrylic acid and
byproducts of acetic acid and carbon dioxide. All of the reactions
are highly exothermic and have large activation energies, so they
are highly sensitive to temperature. There “hot” reactions present
challenging design and control problems. In addition to propylene
and air, a large amount of steam is fed to the reactor to serve as a
thermal sink for the large amount of heat generated by the reac-
tions. There is a significant cost for providing this steam. Air is free
but it requires compression to the 4.3 bar reactor pressure, which
is set by the downstream operating pressure of an absorber.
The heat of reaction is removed by using a heat exchanger inside
the fluidized-bed reactor vessel, which is assumed to act like a CSTR
in this paper. A large amount of heat must be removed, and the
required heat-transfer area depends on the temperature difference
between the reactor and the steam. Valuable high-pressure steam
can be generated when reactor temperatures are high, but high
reactor temperatures give low selectivity (more carbon dioxide for-
mation). Less valuable low-pressure steam must be used to achieve
low reactor temperatures and high selectivity (low carbon dioxide
formation).
For a given flowrate of propylene fed to the reactor, there are
optimum flowrates of both air and steam since they impact yields,
reactor temperature.
The process, which was developed many decades ago, is dis-
cussed in several books and papers. Most notably, the popular
design textbook by Turton et al. (Turton et al., 2003) presents the
process as a design project. The process was featured (AIChE, 1986)
as the AICnE Student Design Problem in 1986.
Safety is an important issue because of the upper and lower
flammability limits of propylene in oxygen. The gaseous mixture
in the reactor has high concentrations of inert nitrogen and water.
Turton et al. (Turton et al., 2003) suggest that flammability prob-
lems can be avoided by keeping the oxygen composition below
5 mol%. Of course this high limit may not be the optimum in some
cases since it entails feeding excess air with the inherent higher
compression costs.
The reactor described in Turton et al. (Turton et al., 2003) is a flu-
idized bed reactor that is essentially a CSTR. A later design project
from West Virginia (Department of Chemical Engineering, 2016)
and a recent paper (Suo et al., 2015) suggest the use of a cooled
tubular reactor. In this paper we used the Aspen RCSTR model for
the reactor. In a later section we discuss the significant differences
between these two reactor types and demonstrate that the use of a
tubular reactor in this “hot” reaction system is problematic. Reac-
tor runaways, excessive peak temperatures and low acrylic acid
yields can easily occur in a cooled tubular reatcor even with a very
large number of small tubes to provide large heat-transfer area. The
sensitivity of reactor internal temperatures down the length of the
reactor to the coolant temperature is extremely high.

E-mail address: WLL0@Lehigh.edu

http://dx.doi.org/10.1016/j.compchemeng.2016.06.005
0098-1354/© 2016 Elsevier Ltd. All rights reserved.
 W.L. Luyben / Computers and Chemical Engineering 93 (2016) 118–127
Table 1
Kinetics for gas phase irreversible reactions.
k E (kcal/kmol)
1 4.4167 × 10−5 15,000
2 2.4528 × 10−4 20,000
3 5.0278 × 10−2 25,000
Concentration units − Pascals.
Reaction rate − kmols s−1 m−3 Pa−2.
No studies in the open literature have been found that discuss
the optimum economic design of the acrylic acid process. That is
the purpose of this paper. The study is limited to the reactor portion
of the process. This approach should be valid since there are no
recycle streams that couple the reaction and separation sections of
the process.
2. Base case design
A thorough description of the acrylic acid process is provided
by Turton et al. (Turton et al., 2003) and is used as the basis in this
paper. Turton does not address the issues of economic optimization
and dynamic controllability.
There are three reactions that are assumed to be irreversible and
first-order in both propylene and oxygen partial pressures. The first
produces the desire acrylic acid product.
3 sists of an absorber to recover the products from the gas stream,
1 : C3 H6 + O2 → C3 H4 O2 + H2 O
2 a liquid-liquid extractor and several distillation columns. The off-
The second and third reactions produce the undesired byprod-
ucts of acetic acid and carbon dioxide.
5 conversion of propylene to avoid loss of valuable raw material.
2 : C3 H6 + O2 → C2 H4 O2 + H2 O + CO2
2
9 optimized independently. In most chemical processes, this is not
3 : C3 H6 + O2 → 3CO2 + 3H2 O
2 the case since reactor performance impacts the economics of the
The acetic acid can be sold but is less valuable than acrylic acid.
The carbon dioxide represents a significant loss of propylene in
addition to adding to emission problems. Therefore the suppres-
sion of the production of carbon dioxide is particularly important
in this process. Selectivity has a strong impact on the profitability
of the process since it affects the amount of the desirable product
that is produced. Large increases in capital investment and energy
(compressor work) can be economically justified to improve the
yield of acrylic acid.
The kinetic provided by Turton use units of kmol/h and kPa. The
required units for Aspen reactor simulations must use kmol/s and
Pa. The kinetics used in this paper are given in Table 1. Note that the
reactions are gas phase and first order. The basis is reactor volume.
Aspen UNIQUAC physical properties are used.
Since the activation energies of the undesired reactions are
larger than the activation energy of the desired reaction, it is obvi-
ous that the optimum reactor temperature should be low. But of
course low temperatures produce small reaction rates, so the size
of the reactor required to achieve the desired propylene conversion
increases as reactor temperature decreases. This is an example of
the classical reactor design trade-off between yield (the value of the
products produced) and capital investment (the cost of the reactor
vessel and catalyst).
Fig. 1 gives the base-case flowsheet used in this paper. The
flowrate of vapor propylene is 127 kmol/h. The flowrate of steam is
992.3 kmol/h of saturated low-pressure steam at 160 ◦ C and 6 bar.
The flowrate of air is 1362 kmol/h. The original Turton flowsheet
used a single-stage compressor, but a two-stage system is used in
this paper to keep compressor discharge temperature less than the
heuristic maximum of 200 ◦ C. Total compressor work is 2.092 MW.
A water-cooled interstage heat exchanger is used to cool the hot
119
gas from the first compressor down to 50 ◦ C before feeding to the
second stage compressor
The reactor vessel has a total volume of 118 m3 and oper-
ates at 310 ◦ C and 4.3 bar. A large heat exchanger is installed
inside the reactor to remove the large heat of reaction (23.84 MW).
High-pressure steam is generated (254 ◦ C) giving a differential
temperature driving force of 310–254 = 56 ◦ C. Assuming an over-
all heat-transfer coefficient in this gas/fluidized solid system of
U = 0.28 kW m−2 K−1 , the required heat-transfer area is 1520 m2 .
The volume of this heat exchanger is 38 m3 . The volume of the reac-
tion phase required to achieve the specified propylene conversion
is 80 m3 , so the total volume of the vessel is 118 m3 . Using an aspect
ratio L/D = 2 gives a vessel diameter of 4.22 m and length of 8.44 m.
The gas leaving the reactor contains acrylic acide (AA), acetic
acide (HAc), carbon dioxide (CO2 ), water, inert nitrogen, unreacted
propylene (C3) and unreacted oxygen. The conversion of propy-
lene in the Turton base case is 95.25%. The selectivity of acrylic
acid relative to acetic acid is 13.5. The selectivity of acrylic acid
relative to carbon dioxide is 14.7. At the 310 ◦ C reactor temper-
ature used by Turton, there is a significant loss of propylene to
produce undesired by-products, particularly carbon dioxide. The
air flowrate of 1362 kmol/h produces an oxygen composition in
the reactor of 1.36 mol%, which is well below the flammability
limit. More air could be fed, which would decrease reactor size but
increase compression costs. This is an important design trade-off.
The downstream separation system suggested by Turton con-
gas from the absorber is sent to an incinerator, so any unreacted
propylene is lost. Therefore the reactor must be designed for high
It is important to note that there is no recycle from the sep-
aration section back to the reactor. Therefore the reactor can be
separation section.
In the following section we explore the effects of several impor-
tant design variables on the economics of the process in terms of
product value, energy costs and capital investment. The economic
objective function is profit, which is defined as the value of the
acrylic acid and acetic acid products plus the value of the steam
generated in the reactor minus the cost of the propylene and steam
feeds minus compressor energy cost minus total annual capital cost.
3. Effects of design variables
The flowrate of propylene is fixed in all cases at 127 kmol/h. The
cost of propylene and the value of acrylic acid and acetic acid are
taken from ICIS.com student historical chemical price information
(ISIS.com, 2016). The conversion of propylene is fixed at 95.25% in
all cases so that a direct comparison with the Turton design can
be made. In a later section of this paper, the effect of conversion
on profitability is quantified. Conversion is maintained constant as
other variables are changed by varying reactor volume using an
Aspen Design Spec/Vary feature. A second Aspen Design Spec/Vary is
used to achieve a specified value (to be optimized) for the selectivity
of acrylic acid relative to acetic acid. As we demonstrate below,
setting selectivity fixes reactor temperature uniquely, i.e. there is
only one degree of freedom relating selectivity and temperature.
In addition to selectivity, the feed flowrates of air and steam are
also design optimization variables. Using larger air feed flowrates
increases oxygen composition in the reactor, which results in
smaller reactor sizes for a fixed propylene feed flowrate and a
fixed propylene conversion. This reduces capital investment in the
reactor vessel and catalyst. However, using more air increases com-
pression costs, both energy and capital.
120 W.L. Luyben / Computers and Chemical Engineering 93 (2016) 118–127

Fig. 1. Base case acrylic acid reactor.

Table 2
Sizing and economics (propylene conversion 95.25%).

Base Case Less Water Less Air More Air Optimum HP Steam Optimum LP Steam

TR (◦ C) 310 310 310 310 262 175

Selectivtiy AA/HAc 13.5 13.5 13.5 13.5 19.9 49.6
VR (m3 ) 80.00 51.03 387.7 36.0 273.0 1068
ID (m)
Feeds (kmol/h)
Propylene 127 127 127 127 127 127
Water 992.3 508.0 992.3 992.3 508.0 508.0
Air 1362 1383 1200 1600 1111 1362
yR(mol%O2 ) 1.36 1.70 0.264 3.36 1.0 5.0
Products (kmol/h)
AA 94.79 94.79 94.79 94.79 105.79 116.86
HAc 7.016 7.016 7.016 7.016 5.318 2.356
CO2 64.54 64.54 64.54 64.54 34.91 7.618

QHX (MW) 23.84 24.62 23.99 23.46 22.13 20.34

Area HX (m2 ) 1520 1570 1530 1496 9880 4848
Compressors
Energy (MW) 2.092 2.092 1.850 2.467 1.713 2.099
(106 $/y) 1.083 1.083 0.980 1.307 0.9075 1.112
Capital (106 $)
Reactor Vessel 0.9047 0.7658 2.012 0.6732 2.278 3.812
Catalyst 0.1800 0.1148 0.8732 0.0810 0.6142 2.403
Heat Exchangers 0.8538 0.8719 0.8573 0.8449 2.862 1.935
Compressors 3.429 3.429 3.158 4.000 2.965 3.503
Total Capital 5.377 5.227 6.516 5.676 8.495 11.67

Cost Water (106 $/y) 2.921 1.496 1.496 1.496 1.496 1.496
Cost C3 (106 $/y) 61.86 61.86 61.86 61.86 61.86 61.86
Value Steam (106 $/y) 7.428 7.671 7.475 7.310 6.895 4.995
Value Products (106 $/y) 149.19 149.19 149.19 149.19 164.50 178.97

Profit (106 $/y) 88.96 90.68 88.72 88.52 104.3 115.6

Feeding more steam has two undesirable effects on the eco-
nomics. It reduces oxygen concentration, which increases reactor
size. It also requires the generation of more steam, which increases
boiler-feed-water purchase costs and steam boiler energy con-
sumption to vaporize the water. However, the WVa Design Project
(Department of Chemical Engineering, 2016) reports that there is
a low limit of 4 for the ratio of steam to propylene fed to the reac-
tor to prevent coke formation. So we assume that the minimum
steam flowrate is (127)(4) = 508 kmol/h. The effects of this reduc-
tion on process design and economic parameters are given in the
next section.
An additional design optimization variable is the heat sink used
in the internal heat exchanger in the reactor. If high-pressure steam
(254 ◦ C) is generated, the reactor temperature must be greater than
 W.L. Luyben / Computers and Chemical Engineering 93 (2016) 118–127
254 ◦ C by some reasonable differential temperature driving force
for heat transfer. As the reactor temperature approaches the steam
temperature, the area of the heat exchanger becomes very large,
which translates into excessively large capital investment. It also
means that the internal heat exchanger becomes larger in vol-
ume and occupies a larger fraction of the total reactor volume. We
assume that mixing and fluidization issues limit this fraction to 50%.
Generating steam at lower pressure permits operating the reac-
tor at lower temperature and thus achieving higher selectivity.
However, the value of steam decreases as pressure decreases. The
use of high or low pressure steam generation is considered later in
this paper.
All of these considerations make it clear that economic optimum
design of the acrylic acid reactor involves a number of interesting
and important design trade-offs.
4. Basis of economics
Prices of chemicals are obtained from the ICIS.com web site
(ISIS.com, 2016): propylene (C3) $0.60 per pound, acrylic acid (AA)
$1.09 per pound and acetic acid (HAc) $0.68 per pound. The costs
of producing the boiler feed water and vaporizing it to produce
the low-pressure steam fed to the reactor are taken from Tur-
ton: $2.45 per 1000 kg of BFW purchased and $16.22 per 1000 kg
of steam generated in a boiler. Vessel, heat exchanger and com-
pressor sizing equations and capital cost are taken from Luyben
(Luyben, 2013).The reactor overall heat-transfer coefficient used is
0.28 kW m−2 K−1 .
The value of high-pressure steam at 42 bar and 254 ◦ C is assumed
to be $9.88 per GJ. The value of low-pressure steam at 6 bar and
160 ◦ C is assumed to be $7.78 per GJ. Compressor work cost is $16.8
per GJ. Catalyst cost is $2.25 per kg, and catalyst properties are 0.5
void fraction and 2000 kg/m3 particle density.
The economic objective function is profit, which take into con-
sideration the value of the products and steam produced, the cost
of raw materials, compressor energy cost and the annual cost of
capital (using a 3-year payback period).
Costpropylenefeed = (127 kmol/h)(42.07 kg/kmol)(1lb/0.454 kg)($0.60/lb)(8760 h/y)
Coststeamfeed = (FSkmol/h)(18 kg/kmol)($2.45 + $16.22)(8760 h/y)/1000 kg
Valueacrylicacid = (FAAkmol/h)(72.06 kg/kmol)(1lb/0.454 kg)($1.09/lb)(8760 h/y)
Valueaceticacid = (FHAckmol/h)(60.05 kg/kmol)(1lb/0.454 kg)($0.68/lb)(8760 h/y)
Valuereactorsteam = (QRMW)(GW/1000 MW)(Value$/GJ)(3600 s/h)(8760 h/y)
Costcompressorwork = (WcompMW)(GW/1000 MW)($16.8/GJ)(3600 s/h)(8760 h/y)
Totalcapital($) = Reactorvessel + catalyst + heatexchangers + compressors
Annualcapitalcost($peryear) = totalcapitalcost/3years
Profit = Valueproducts + valuereactorsteam − costofpropylene − coststeamfeed – costcompressorenergy−annualcapitalcost
Using these economic factors, the profit generated in the base-
case Turton design is given in the first column in Table 2. Notice that
the value of the products ($149,190,000 per year) is much greater
than energy and capital costs, so the optimum design will feature
high selectivity.
The base-case reactor temperature TR is 310 ◦ C, which gives
the AA/HAc selectivity of 13.5. The base-case propylene conver-
sion (held constant in all the cases) is 95.25%. The required reaction
121
volume VR is 80 m3 . Remember that this is the volume of the reac-
tive solid catalyst/gas portion of the reactor. The total volume of
the vessel is this reaction volume plus the volume occupied by the
internal heat exchanger. Heat exchanger volume is calculated from
the heat-transfer area (0.025 m3 of volume per m2 of heat-transfer
area, assuming typical heat exchanger designs with 0.05 m tube
diameter).
Table 2 shows that the total capital investment for the base
case is $5,377,000. The cost of energy (compressor work) is only
$1,083,000 per year, which is much smaller than the cost of the
propylene feed ($61,860,000 per year) and the value of the acrylic
acid and acetic acid products ($149,190,000 per year). The cost of
the steam fed to the reactor is $2,921,000 per year, which is three
times the cost of energy. The value of the high-pressure steam
gneratd in the reactor is $7,428,000 per year.
The selectivity of acrylic acid relative to adetic acid is 13.5 at
the base-case reactor temperature of 310 ◦ C. Profit at the base-case
conditions is $88,960,000 per year.
5. Effect of design optimization variables
With the propylene feed flowrate fixed at 127 kmol/h, the opti-
mum flowrates of the other two feed streams (air and steam) need
to be found. The optimum values of the reactor temperature and
reactor size must be determined.
5.1. Effect of reactor temperature
Temperature is the most important variable in any reactor
design. An important fundamental understanding ot the acrylic acid
reactor is;
Specifying the selectivity fixes the reactor temperature. Spec-
ifying both the selectivity and the propylene conversion fixes the
production rates of all the products.
The effect of varying reactor temperature is quantified in Fig. 2.
Decreasing reactor temperature increases selectivity, as expected,
since the activation energies of the undesirable reactions are
greater than the activation energy of the desired reaction. As reactor
temperature decrease, reactor volume VR must increase to main-
tain propylene conversion. Higher selectivity means more acrylic
acid, less acetic acid and much less carbon dioxide (since the carbon
dioxide reaction has the largest activation energy). Note the sharp
rise in AA/CO2 selectivity that can be achieved if reactor temper-
ature can be lowered below 200 ◦ C. However, also note the sharp
122 W.L. Luyben / Computers and Chemical Engineering 93 (2016) 118–127
Fig. 2. Temperature effects.
rise in reactor size, which means increased capital investment in
both vessel and catalyst.
It is clear that there is a significant potential to improve the
reactor design by finding the optimum selectivity for the process.
5.2. Effect of steam fed to reactor
The second column in Table 2 gives results when the steam fed to
the reactor is reduced from 992.3 to 508 kmol/h, which corresponds
to using a steam-to-propylene feed ratio = 4. Propylene conversion
is constant at 95.25% (by adjusting reactor volume) and selectivity
is held constant at 13.5 (by maintaining the same 310 ◦ C reactor
temperature as used in the base case). Air flowrate is held at the
base-case value (1362 kmol/h).
The smaller steam feed flowrate increases the oxygen concen-
tration in the reactor gas (yR) from 1.36 to 1.7 mol% O2 , which
reduces the required reactor volume from 80.00 to 51.03 m3 . Reac-
tor capital cost drops from $904,700 to $765,800. Catalyst cost
drops from $180,000 to $114,800.
The big effect is the cost of supplying the steam, which drops
from $2,921,000 per year to $1,496,000 per year. The result is an
increase in profit from $88,960,000 to $90,680,000 per year.
These results demonstrated that feeding the minimum amount
of steam, subject to the steam-to-propylene ratio of 4, is the opti-
mum economic condition.
5.3. Effect of air fed to reactor
The third and fourth columns in Table 2 show the effect of reduc-
ing or increasing flowrate of the air fed to the reactor. To provide
a direct comparison with the base case, the steam feed, conversion
and selectivity (temperature) are set at the base-case values.
Feeding less air decreases compression costs (energy and cap-
ital) but decreases oxygen composition in the reactor, which
increases reactor size. The decrease in air flowrate from 1362.9 to
1200 kmol/h reduces compressor power from 2.092 to 1.850 MW
but drastically increases reactor volume from 80.0 to 387.7 m3
because the oxygen concentration in the reactor drops from a rea-
sonable 1.36 mol% O2 to a very low 0.264 mol%. The final result is a
reduction in profit because of the increase in capital investment in
the reactor vessel and catalyst.
The increase in air flowrate from 1362.9 to 1600 kmol/h
increases compressor power from 2.092 to 2.467 MW but decreases
reactor volume from 80.0 to 36.0 m3 because of the higher oxygen
concentration in the reactor (1.36 mol%–3.36 mol% oxygen). The net
result is again a reduction in profit because the increase in compres-
sion costs (energy and capital) outweigh the decrease in reactor
capital cost.
These studies indicate that the air flowrate should be kept high
enough to keep a reasonable oxygen composition in the reactor
so that the size of the reactor is reasonable. At the 310 ◦ C base-
case reactor temperature, the oxygen composition does not affect
reactor size very significantly because the high temperature keeps
reactor size small. However, if we move to lower temperatures in
order to improve selectivity, reactor size increases drastically and
the impact of air flowrate becomes more important. The results
given in the following section show that air flowrate should be set
at the 5 mol% O2 flammability limit when reactor temperatures are
low.
Fig. 3 illustrates the interplay between reactor oxygen compo-
sition (as set by air flowrate) and reactor size at several reactor
temperatures. Reaction volume increases as reactor temperature
or oxygen composition decrease. There are important design trade-
offs among these variables. For high selectivity with low reactor
 W.L. Luyben / Computers and Chemical Engineering 93 (2016) 118–127
Fig. 3. Effect of air flowrate.
temperatures and large reactor volumes, the optimum air flowrate
will be large to give a large reactor oxygen composition.
5.4. Effect of specified selectivity
With propylene conversion fixed and with steam feed set to
give a steam-to- propylene ratio of 4, the important effect of the
specified AA/HAc selectivity is explored in the section. Selectivity is
uniquely related to reactor temperature. We consider two cases in
which either high-pressure steam at 254 ◦ C or low-pressure steam
at 160 ◦ C is generated in the reactor. Obviously reactor tempera-
ture has to be somewhat greater than the temperature of the steam
being generated.
A: HP steam generation
We first consider the case in which high-pressure steam at
254 ◦ C is generated in the reactor so reactor temperature must be
greater than 254 ◦ C. The closer the reactor temperature is to the
temperature of the steam, the larger the internal heat exchanger
must be. As the results presented below demonstrate, the limit-
ing condition on reactor temperature is the volume of the heat
exchanger becoming too large for the required reaction volume.
Since reactor temperatures must be fairly high when using
high-pressure steam, reactor volume will not have to be large.
So reducing reactor volume by using more air to increase oxygen
composition is not necessary. Therefore we fix the reactor oxygen
composition at 1 mol% and study the impact of changing the spec-
ified selectivity, which is uniquely related to reactor temperature.
This oxygen composition is way below the safety limit of 5 mol%
oxygen for flammability.
Results are shown in Fig. 4. The upper left graph shows how the
selectivity of acrylic acid to acetic acid increases as reactor tem-
perature is reduced. Lower temperatures require larger reactors
and larger heat exchangers. However, lower temperatures require
somewhat smaller air flowrates to maintain the same reactor oxy-
gen composition. This is due to the lower production of carbon
dioxide, which reduces the consumption of oxygen. The resulting
lower compression costs are less significant than the higher capital
cost of the reactor and catalyst.
123
The net effect is an increase in total capital investment as reactor
temperature is decreased (middle right graph in Fig. 4). The lower
right graph shows that the profit is maximized when a reactor tem-
perature of 260 ◦ C is selected. This is the point where the optimum
trade-off between capital investment and product value is achieved
with the fixed 1 mol% oxygen composition.
However, the bottom left graph in Fig. 4 shows an interesting
effect. The volume of the reactive portion of the vessel VR is the
solid line. The volume occupied by the internal heat exchanger VHX
is dashed line. At a reactor temperature of 262 ◦ C, these two vol-
umes are approximately equal. We assume that a practical limit on
the fraction of the total vessel volume that can be occupied by the
heat exchanger is 50%. Above this fraction, adequate mixing and
fluidization may be problematic.
Details of the equipment sizes and economic variables are given
in the fifth column in Table 2. A comparison of this optimum design
using high-pressure steam with the base case shows:
1. Capital investment is higher ($8,495,000 versus $5,377,000)
because the reactor and heat exchanger are much larger.
2. Less air and less steam are fed to the reactor, which reduces
compression costs and BFW and steam generation costs.
3. The value of the products is increased from $149,190,000 per
year to $164,500,000 per year because the lower reactor tem-
perature (262 ◦ C versus 310 ◦ C) increases selectivity.
4. The generation of green-house gas carbon dioxide is reduced by
a factor of almost 2 (64.54 kmol/h to 34.91 kmol/h).
5. Profit increases by 11%.
The large improvement in the economics can justify the
increased capital investment in reactor, catalyst and heat
exchanger. Fig. 5 gives the flowsheet for the optimum reactor when
generating high-pressure steam.
B: LP steam generation
Now we consider the case in which low-pressure, less-valuable
steam at 160 ◦ C is generated in the reactor. Reactor temperatures
can be lower to achieve higher selectivity but cannot approach too
close to the low limit of 160 ◦ C because heat-exchanger area will
124 W.L. Luyben / Computers and Chemical Engineering 93 (2016) 118–127

Fig. 4. Results for HP steam generation.

Fig. 5. Flowsheet for HP steam generation.

become very large and increase capital investment. The closer the
reactor temperature is to the temperature of the steam, the larger
the internal heat exchanger becomes.
Low reactor temperatures require large reactor volumes, so the
volume of the internal heat exchanger will not be a limiting con-
dition. On the other hand, reducing reactor volume by using more
air to increase oxygen composition becomes very important. We
fix the reactor oxygen composition at the 5 mol% maximum oxy-
gen flammability limit so as to keep reactor volume as small as
possible. The economics were studied over a range of reactor tem-
 W.L. Luyben / Computers and Chemical Engineering 93 (2016) 118–127 125

Table 3 of 95.25–97% is 82% (Profit/Investment). The return on incre-

Effect of conversion.
mental capital investment in going from a propylene conversion of
Propylene Conversion 95.25 97 98 99 97–98% is 16%. Fig. 7 gives the flowsheet for the 98% conversion
TR (◦ C) 175 175 175 175 process.
VR (m3 ) 1068 1739 2650 5386
ID (m) 9.11 10.58 12.09 15.19
6. Dynamic control
Feeds (kmol/h)
Propylene 127 127 127 127
Water 508 508 508 508 A vital component of any process design is quantitative assess-
Air 1362 1383 1395 1408 ment of the dynamic controllability of the process. The concept
yR(mol%O2 ) 5 5 5 5
of “simultaneous design” has been discussed for many decades in
Products (kmol/h)
AA 116.86 119.0 120.23 121.46
the design and control literature. Many companies include control
HAc 2.356 2.400 2.424 2.449 studies as an integral part of their design activities. Unfortunately
CO2 7.618 7.756 7.838 7.918 very few universities adhere to this important concept in their cap-
QHX (MW) 20.34 20.73 20.95 21.17 stone senior design course.
Area HX (m2 ) 4848 4935 4988 5041 In this section we present a brief discussion of the dynamic
Compressors controllability of the acrylic acid reactor.
Energy (MW) 2.099 2.133 2.151 2.170
(106 $/y) 1.112 1.130 1.140 1.150
Capital (106 $) 6.1. CSTR system
Reactor Vessel 3.812 5.142 6.466 9.913
Catalyst 2.403 3.910 5.963 12.12 The reactor studied in this paper is assumed to behave as a
Heat Exchangers 1.935 1.974 1.988 2.001
Compressors 3.503 3.549 3.579 3.600
CSTR with perfect mixing (temperature and compositions are uni-
Total Capital 11.67 14.48 18.00 27.63 form throughout the vessel). The dynamic controllability of CSTR
systems has been extensively studied. Problems with thermal run-
Cost Water (106 $/y) 1.496 1.496 1.496 1.496
Cost C3 (106 $/y) 61.86 61.86 61.86 61.86 aways are known to occur in systems with exothermic irreversible
Value Steam (106 $/y) 4.995 5.085 5.140 5.194 reactions.
Value Products (106 $/y) 178.97 182.1 184.1 186.0 However, temperature instability does not occur in CSTR sys-
Profit (106 $/y) 115.6 117.9 118.8 117.5 tems that are designed for high conversion. There is little “fuel” in
the reactor if the concentration of the limiting reactant is low. Such
is the case in the acrylic acid reactor, which is designed for 95.25%
peratures, and the temperature that gives the maximum profit is conversion of the propylene “fuel.”
175 ◦ C. The AA/HAc selectivity at this temperature is 49.6. The dynamics of the low-pressure design (Fig. 6) were explored,
The right-most column in Table 2 gives the sizing and economic and stable temperature control was found to be easily achieved. The
results for the optimum low-pressure steam case. The flowsheet is control structure used consisted of the following loops.
shown in Fig. 6.
A comparison of the low-pressure steam design with the high- 1. Propylene feed is flow controlled by manipulating the control
pressure steam design shows: valve in the feed line.
2. Steam feed is flow controlled by manipulating the control valve
1. Capital investment is higher ($11,670,000 versus $8,495,000) in the steam feed line. It is ratioed to propylene feed by using
because the reactor is larger. a multiplier whose two inputs are propylene flowrate and the
2. Slightly more air is fed, which increases compression costs. constant “4.” The output of the multiplier is the setpoint of the
3. The value of the products is increased from $164,500,000 per steam flow controller, which is on “cascade.”
year to $178,970,000 per year because the lower reactor tem- 3. Air feed is flow controlled by manipulating the power sent to
perature (175 ◦ C versus 262 ◦ C) increases selectivity. both air compressors. It is ratioed to propylene feed by using
4. The generation of green-house gas carbon dioxide is reduced by a multiplier whose two inputs are propylene flowrate and the
78% (from 34.91 kmol/h to only 7.618 kmol/h). constant “10.72.” The output of the multiplier is the setpoint of
5. Profit increases by 11%. the steam flow controller, which is on “cascade.”
4. Pressure is controlled by manipulating control valve in the reac-
The improvements from the original base-case design are tor effluent line.
dramatic. By investing more capital ($5,377,000 increased to 5. Reactor temperature is controlled by manipulating the pressure
$11,670,000), profits have been increased 30% ($88,960,000 per of the steam in the internal heat exchanger. In the Aspen sim-
year to $115,600,000 per year). Thus the incremental return on ulation, the Heat Transfer option of the Aspen RCSTR model is
investment is over 400%. selected to be Constant Temperature. This means that the tem-
perature of the cooling medium is adjustable in the dynamic
5.5. Effect of specified conversion simulation.

Up to this point we have used the base-case propylene conver-
sion of 95.25%. It is important to see if this conversion is really the
economic optimum. Table 3 gives results over a range of conver-
sions up to 99% with reactor temperature held constant at 175 ◦ C
and generating low-pressure steam in the reactor. Reactor gas com-
position is held constant at 5 mol% O2 .
Reactor size increases as conversion increases, which increases
capital investment. But the amount of acrylic acid increases, which
increases the value of the products generated. The return on incre-
mental capital investment in going from a propylene conversion
A 1 min temperature measurement lag was inserted in the tem-
perature loop.
Relay-feedback testing and Tyreus-Luyben tuning rules were
used to tune the temperature controller. The process gave stable
robust control for large 20% changes in propylene feed flowrate
and large 10 ◦ C changes in the temperature controller setpoint.
These dynamic studies demonstrated that the acrylic acid reac-
tion can be effectively controlled when a CSTR system is used.
Despite the high heat of reaction and the large activation energies
that give “hot” reactions, good temperature control is achievable
126 W.L. Luyben / Computers and Chemical Engineering 93 (2016) 118–127

Fig. 6. Flowsheet for LP steam generation.

Fig. 7. 98% conversion with LP steam generation.

because of the high conversion of propylene. There is little propy-
lene inside the reactor to fuel a temperature runaway. As the
propylene is fed into the perfectly mixed reactor, it is immediately
reacted.
6.2. Tubular reactor system
The suggestion is made in both the WVA design project
(Department of Chemical Engineering, 2016) and in Suo et al. (Suo
et al., 2015) that a cooled tubular reactor can be used in the acrylic
acid system. We explored the steadystate design of such a configu-
ration using the base-case conditions for the three feed flowrates.
We were unable to achieve a design that produced the desired high
conversion of propylene and high selectivity.
Even when an enormous number of small diameter tubes were
used, the system was extremely sensitive to the coolant tempera-
ture on the steam side of the reactor.
Fig. 8 illustrates the problem. These results are for a case using
the base-case feed conditions and using the tubular reactor config-
uration suggested by Suo et al. (Suo et al., 2015) (16,000 tubes with
0.026 m diameter, 10 m length and U = 1 kW m−2 K−1 ).
 W.L. Luyben / Computers and Chemical Engineering 93 (2016) 118–127
Fig. 8. Tubular reactor temperature profiles.
With a steam temperature (Tmed) of 208 ◦ C, the peak temper-
ature is 262 ◦ C at the front end of the reactor very close to the
inlet. Raising the steam temperature to 208.4 ◦ C produces a peak
of 279 ◦ C. Then raising Tmed by only 0.1 ◦ C results in a sharp spike
in the peak temperature up to 737 ◦ C. The carbon dioxide in the
reactor effluent jumps dramatically from 21.31 kmol/h at 208 ◦ C to
101.6 kmol/h. Oxygen is complete consumed. Propylene conversion
is lower (4.06 kmol/h in the reactor effluent instead of 2.30 kmol/h).
Acrylic acid produced decreases from 115 to 84.6 kmol/h.
Therefore our preliminary studies indicate that designing and
operating the acrylic acid reaction in a cooled tubular reactor may
not be feasible. We plan to study this issue further in future work.
7. Conclusion
The design of the acrylic acid reactor features several interest-
ing trade-offs. The results of this study indicate that the reactor
should be operated at low temperature using large reactors and
generating low-pressure steam. The steam feed should set to give a
127
steam-to-propylene ratio of 4 (coke formation constraint). Air feed
should be set to give a reactor gas composition of 5 mol% oxygen
(flammability constraint).
The dynamic controllability performance of a CSTR for produc-
ing acrylic acid is excellent. The use of a tubular reactor for acrylic
acid production appears to be problematic.
References
Turton, R., Bailie, R.C., Whiting, W.B., Shaelwitz, J.A., 2003. Analysis, Synthesis and
Design of Chemical Processes, 2nd edition. Prentice Hall (appendix B.9).
AIChE , 1986. Student Design Problem.
Department of Chemical Engineering , West Virginia University, ChE 655–Plant
Design Project #4 (Summer 20140 “Design of an acrylic acid production
process”. 2016.
Suo, X., Zhang, H., Ye, Q., Dai, X., Yu, H., Li, R., 2015. Design and control of an
improved acrylic acid process. Chem. Eng. Res. Des. 104, 346–356.
ISIS.com web site 2016.
Luyben, W.L., 2013. Distillation Design and Control Using Aspen Simulation, 2nd
ed. Wiley.