Вы находитесь на странице: 1из 21

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/235974353

Economic Optimization of the Cumene Production Process Using


Response Surface Methodology

Article  in  Chemical Engineering Communications · November 2012


DOI: 10.1080/00986445.2012.660895

CITATIONS READS

3 3,999

2 authors:

Hamid Norouzi Shohreh Fatemi


University of Tehran University of Tehran
21 PUBLICATIONS   131 CITATIONS    182 PUBLICATIONS   1,076 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Design of a Hybrid Control System for Pressure Swing Adsorption Processes View project

CFD of tubular membranes View project

All content following this page was uploaded by Shohreh Fatemi on 08 January 2016.

The user has requested enhancement of the downloaded file.


This article was downloaded by: [University of Windsor]
On: 29 July 2013, At: 03:31
Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Chemical Engineering Communications


Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/gcec20

ECONOMIC OPTIMIZATION OF THE


CUMENE PRODUCTION PROCESS USING
RESPONSE SURFACE METHODOLOGY
a a b
Hamid Reza Norouzi & Shohreh Fatemi
a
School of Chemical Engineering, College of Engineering, University
of Tehran, Tehran, Iran
b
Oil and Gas Center of Excellence, University of Tehran, Tehran,
Iran
Published online: 06 Aug 2012.

To cite this article: Hamid Reza Norouzi & Shohreh Fatemi (2012) ECONOMIC OPTIMIZATION OF THE
CUMENE PRODUCTION PROCESS USING RESPONSE SURFACE METHODOLOGY, Chemical Engineering
Communications, 199:11, 1375-1393, DOI: 10.1080/00986445.2012.660895

To link to this article: http://dx.doi.org/10.1080/00986445.2012.660895

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
Chem. Eng. Comm., 199:1375–1393, 2012
Copyright # Taylor & Francis Group, LLC
ISSN: 0098-6445 print=1563-5201 online
DOI: 10.1080/00986445.2012.660895

Economic Optimization of the Cumene Production


Process Using Response Surface Methodology

HAMID REZA NOROUZI1 AND SHOHREH FATEMI1,2


1
School of Chemical Engineering, College of Engineering, University of
Tehran, Tehran, Iran
2
Oil and Gas Center of Excellence, University of Tehran, Tehran, Iran
Downloaded by [University of Windsor] at 03:31 29 July 2013

The process of manufacturing cumene from propylene and benzene on a solid


phosphoric acid catalyst was studied. The aim of this article is to optimize this process
economically and to realize how statistical methods can be utilized to optimize the
operational conditions of the process at the industrial scale. The net present value
(NPV) that reflects all the economic aspects of the process was chosen as the objective
function for optimization. The NPV of the base process was estimated at $0.19 million.
In the first step, the full factorial design was applied to determine the significant
factors affecting the process. Results revealed that benzene recycle, reactor volume,
reactor pressure, and inlet temperature, as well as the binary interaction of recycle-
temperature and recycle-pressure influence the NPV. Benzene recycle and inlet
temperature had the most important effects on NPV. In the second step, the response
surface methodology was applied to find a reliable quadratic model for NPV. Optimi-
zation of the effective parameters was carried out. The maximum predicted value of
NPV was calculated $4.56 million at benzene recycle of 141.4 kmol=h, temperature
of 340 C, and pressure of 29 bar with 468 tubes inside the reactor.

Keywords Cumene production process; Economic optimization; Net present


value; Response surface method

Introduction
Finding the configurations and operational conditions in which a chemical is
produced in a reliable, safe, and economical manner with low energy consumption,
and low or negligible waste and byproducts is the goal of design engineers in the
synthesis and construction of chemical plants.
This requires a great amount of knowledge, experience, hard work, and a sys-
tematical strategy. The synthesis of a process is followed by its economic evaluation
as the key parameter. This type of evaluation will be conducted when the flow
sheet of the base process is available and the economic aspects including the size
of equipment, capital investment, utility requirements, and profit of the process
are evaluated (Seider et al., 2010).
In economic optimization, it is better to select manipulating variables that have
significant and contradictory effects on the economy and overall yield of the process.
In a chemical plant where valuable components are converted into products and

Address correspondence to Shohreh Fatemi, Oil and Gas Center of Excellence, University
of Tehran, PO Box 11155-4563, Tehran, Iran. E-mail: shfatemi@ut.ac.ir

1375
1376 H. R. Norouzi and S. Fatemi

byproducts, the reaction conditions and those variables that affect the conversion
and yield of reactions are in the first level of importance. Heat integration and sep-
aration optimization are less significant choices for optimization. Many objective
functions can be proposed for plant profitability measurement, such as the return
on investment, venture profit, payback period, and manufacturing cost. The net
present value (NPV) of the process is suggested for more precise evaluations. This
criterion takes account of the time value of money, the level of capital investment,
annual operating costs, and the revenue of the process (Turton et al., 1998).
In this study, the economy of the cumene production process is investigated. The
end use of cumene is bisphenol A and phenolic resins and caprolactam. NPV is chosen
as the objective function for the economic optimization of this process. The NPV of
this process is calculated based on a comprehensive economic model, which will be
introduced in detail in the following sections. Although there are some methods for
maximizing constrained or unconstrained linear or nonlinear problems (Zivorad,
Downloaded by [University of Windsor] at 03:31 29 July 2013

2004), response surface methodology (RSM), as one of the most efficient methods
in the statistical design of experiments, is used to find the optimal conditions of the
process. RSM is a statistical technique for designing experiments, building models
for the systems, evaluating effects of manipulating variables, and finding optimal
values for objective functions (Vining, 1998). This method has been successfully
applied in many fields of science such as optimization of sugar production from
the medium of bacteriocin production by Lactococcus lactis (Li et al., 2002) and
the optimization of sugar production from sago starch (Wee et al., 2011) in the bio-
logical field; optimization of the conditions of lindane biodegradation in the environ-
mental field (Rigas et al., 2005); optimization of the microwave extraction of phenol
and methyl phenol isomers from soil (Llompart et al., 1997); constructing a signifi-
cant model describing the thermal stability of lysozyme in the pharmaceutical field
(Branchu et al., 1999); and modeling the phenol adsorption process in fixed bed
reactors in the chemical engineering field (Oughlis-Hammache et al., 2010).
In this research, the aim is to determine how the method of RSM works for the
economic optimization of operational conditions in the cumene production process
at the industrial scale and how a step-by-step procedure can be introduced to find the
best operating conditions with the highest economic attractiveness.

Process Description
Cumene is produced by alkylation of benzene with propylene in the presence of a
solid phosphoric acid (SPA) catalyst that is a composite of phosphoric acid and a
binder such as kieselguhr or diatomaceous earth (Prinsloo, 2007). The following
reactions occur in a fixed bed of catalyst pellets at the temperature of 350 C and
pressures ranging from 25 to 32 bar in the vapor phase.

k1
C3 H6 þ C6 H6 ! C9 H12 ð1Þ
Propylene benzene cumene

k2
C3 H6 þ C9 H12 ! C12 H18 ð2Þ
Propylene cumene DIPB

The kinetic rates are listed in Table I. As can be seen in this table, the activation energy
of the cumene reaction (Equation (1)) is lower than that of the diisopropylbenzene
Economic Optimization of Cumene Production 1377

Table I. Kinetics of reactions in the reactor (Turton et al., 1998)

Reaction name Rate expression Rate constant


 
Cumene reaction rcumene ¼ k1 CB Cp 4 24:9
k1 ¼ 3:5  10 exp
RT
 
DIPB reaction rDIPB ¼ k2 CB Cp 6 35:08
k2 ¼ 2:9  10 exp
RT

E is in kcal  mol1, concentration in mol  l1, and temperature in Kelvin.


The rate units are in mole=(gr cat. sec).

(DIPB) reaction (Equation (2)). Thus, the lower temperature of the reactor enhances
Downloaded by [University of Windsor] at 03:31 29 July 2013

the selectivity of the cumene reaction over the DIPB reaction. But the lower reaction
temperature will result in lower conversion per pass of reaction. Therefore, there must
be a tradeoff between the conversion per pass and the selectivity.
The process flow diagram of a cumene production plant taken from Turton et al.
(1998) is illustrated in Figure 1. Fresh and pure benzene at the molar flow rate of
105 kmol=h at 25 C is fed to V-801 and mixed with the recycled benzene coming from
the separation section and pumped to 31.5 bar. The recycled benzene increases the
benzene-to-propylene ratio in the reactor and consequently minimizes the production
of DIPB. Refinery-grade propylene at the molar flow rate of 110.27 kmol=h and
5 mol% propane impurity is pumped to 31.5 bar and mixed with the benzene stream,

Figure 1. Process flow diagram and stream table of cumene process at the base conditions.
1378 H. R. Norouzi and S. Fatemi

Table II. Operating conditions and sizing data for main equipment in the process

Exchangersa
Uoveral Construction Design
(W  m2  K1)b materialc pressure (bar)
E-801 (evaporator) 900 CS=CS 32
E-802 (condenser) 150 CS=CS 32
E-803 (condenser) 450 CS=CS 3
E-804 (re-boiler) 750 CS=CS 3
E-805 (condenser) 450 CS=CS 2
E-806 (re-boiler) 900 CS=CS 2

Vessels
Downloaded by [University of Windsor] at 03:31 29 July 2013

Residence Configuration
time (min)d (mater.) L=D
V-801 10 Horizontal (CS) 3
V-802 5 Vertical (CS) 5
V-803 10 Horizontal (CS) 3
V-804 10 Horizontal (CS) 3

Towers
e
T-801 Reflux ratio ¼ 0.44, pressure ¼ 1.75 bar, vessel material: carbon steel
27 sieve trays with 50% efficiency, 24-in. tray spacing, 75% active area, tray
material: stainless steel
T-802e Reflux ratio ¼ 0.63, pressure ¼ 1.2 bar, vessel material: carbon steel
37 sieve trays with 50% efficiency, 24-in. tray spacing, 75% active area, tray
material: stainless steel

Other equipment
R-801 Ntube ¼ 342, Dtube ¼ 3 in., Ltube ¼ 6 m
Uov ¼ 60 W  m2  K1, HXarea ¼ 342 m2
Catalyst: e ¼ 0.5, q ¼ 1600 kg=m3, dp ¼ 3 mm, should be replaced yearly
Tube mater.: stainless steel, shell mater.: carbon steel
Pumps For all pumps 60% hydraulic efficiency was considered, construction
material: carbon steel
a
20% over design.
b
Engineering Data Book (1998).
c
Collected from Turton et al. (1998).
d
Collected from Branan (1995).
e
Fair method presented in Sieder et al. (2010).

and then it is heated and converted into saturated vapor. Thereafter, it enters the fired
heater H-801 and is heated to the reaction temperature, 350 C. The above-mentioned
reactions occur in the fixed bed catalytic reactor R-801. Both reactions are highly
exothermic. Thus, high-pressure boiling feed water is converted into high-pressure
Economic Optimization of Cumene Production 1379

superheated steam in the shell side of the reactor in order to remove the produced
heat. Product, byproducts, and unreacted components leave the reactor and are
cooled down to 90 C. This cooled stream is fed to flash drum V-802 operating at
1.75 bar. The overhead, which has the mass flow rate of 1160 kg=h, is purged
(9 wt.% of total mass flow of feeds entering the process). Since propane is an inert
component in this process, this purge is employed here to avoid accumulation. This
stream has fuel value and can be used in the fuel cycle of the process. The liquid leav-
ing the flash drum enters the benzene fractionator column T-801 where benzene and
other light components like propylene and propane are separated from cumene and
DIPB. The overhead product is recycled back to the process and the bottom product
is sent to the purification column T-802. In this column, cumene is produced in the
overhead stream with mole purity of 0.995. The bottom stream of this column is
mostly DIPB (1.67% of total mass flow of feeds entering the process) and has no
market, hence the fuel value. Since the boiling points of propylene, benzene, cumene,
Downloaded by [University of Windsor] at 03:31 29 July 2013

and DIPB vary considerably, the separations of the components in distillation


columns are carried out efficiently and easily. Other necessary information about
the equipment and operating conditions of the process are listed in Table II (Branan,
1995; Engineering Data Book, 1998).

Economic Model
As the main goal of this study is the economic optimization of the cumene
production process, the consideration of a comprehensive economic model that takes
account of all important factors is essential. The full economic model requires
information about the fixed capital investment, the yearly utility requirements, the
raw material cost, the operating labor, and the revenue of the process. The methods
of evaluating all these factors will be introduced in this section.

Capital Investment
There are some methods for estimating the capital cost of the process at the prelimi-
nary design stage such as the Lang factor, the six-tenths rule, and bare module cost
(Turton et al., 1998). The six-tenths rule scales the cost of the process from known
capacity to the new capacity by the power of 0.6. Although the implementation of
this method is relatively easy and accurate enough for this type of estimation, it is
not applicable here. The cost determined by the Lang factor method is the cost
required for major expansion of an existing chemical plant. The total cost is the
sum of the purchased cost for the major equipment in the unit multiplied by the
Lang factor, which is 4.74 for fluid processing plants. This method is not sensitive
to changes in construction materials and operating pressure of equipment, which
may vary the cost of the equipment considerably. The bare module cost is a more
detailed cost estimation method that considers the construction materials, the oper-
ating pressure, and the special configuration of equipment in cost estimation. This
technique relates all costs to the purchased cost of equipment evaluated for some
base conditions. The bare module cost of equipment is determined by the following
relation:

0 0
CBM ¼ Cp FBM ð3Þ
1380 H. R. Norouzi and S. Fatemi

0 0
where CBM is the bare module equipment cost, FBM is the bare module cost factor,
and Cp is the purchased cost for the base condition (Turton et al., 1998). To estimate
the bare module cost of all major equipment in the cumene process, we need to know
the size, the construction material, and the configuration and the operating pressure
of all equipment. Detailed information about the construction materials and the
operating condition of the major equipment in the process are listed in Table II.
The bare module correlations for estimating the equipment costs belong to past
data on purchased equipment, and they need to be updated according to changes in
economic conditions (inflation). The Chemical Engineering Plant Cost Index
(CEPCI) is used to account for the effect of inflation.
 
O O CEPCI2
ðCBM Þ2 ¼ ðCBM Þ1 ð4Þ
CEPCI1
Downloaded by [University of Windsor] at 03:31 29 July 2013

where subscripts 1 and 2 refer to the base time and the present time of cost estimation,
respectively. The correlations used for estimating equipment costs belong to 1996
with the CEPCI of 382. The CEPCI of the year 2010 is 550 and all costs are adjusted
according to this value (Chemical Engineering, 2010).

Utility Cost
The fuel price fluctuates more than the equipment costs, and, in fact, it does not
follow a general trend like the CEPCI. The cost of utilities in the plant is greatly
influenced by the price of the fuel and it constitutes the main portion of utility costs.
In general, the utility cost is linked to two separate variable costs: inflation (CEPCI)
and fuel cost. To reflect this dual dependence, the following two-factor equation is
used (Ulrich and Vasudevan, 2006):

Cu ¼ aðCEPCIÞ þ bðCf Þ ð5Þ

where Cu stands for the cost of utilities and a and b are utility cost coefficients. Cf is
the price of the fuel in $=GJ. The values of a and b for different utilities are presented
in the work of Ulrich and Vasudevan (2006).
The price of fuel is based on the average price for natural gas in 2010 in the U.S.,
with data from the U.S. Department of Energy (www.eia.doe.gov). The industrial
price for natural gas is 5.63$=1000 ft3. Assuming 38.1 MJ=std  m3 heating value for
natural gas (Perry’s Chemical Engineers’ Handbook, 1999), then the price of natural
gas is estimated as 5.0$=GJ. According to the price of natural gas as the reference
fuel, the costs of different utilities are calculated and listed in Table III. This table
also includes the values of a and b for different utilities.

Raw Material and Product Costs


The prices of raw materials (benzene and propylene) and products were obtained
from the internet site ‘‘Chemical prices and chemical industry trends ICIS pricing’’
(www.icispricing.com). The prices of benzene, propylene, and cumene are considered
as 830, 940, and 990 $=MT for the year 2010 respectively. High-pressure steam
generated in the reactor costs 9.3$=GJ, assuming 6.5$=kg for boiling feed water.
Economic Optimization of Cumene Production 1381

Table III. Utility cost estimation based on Ulrich and Vasudevan


(2006) for the year 2010

Utility type $=GJ a b


4
Electricity 34.7 1.30  10 0.01
HP steam (42 bar g) 13.1 3.47  106 0.004
MP steam (10 bar g) 10.4 3.47  106 0.004
CW (30 to 45 C) 1.3 7.00  104 0.003

The heat value of fuels produced in streams 11 (purge gas) and 14 (DIPB) are
compared with the price of natural gas (5$=GJ). These fuels are burned by air with
10% excess air and the exhaust flue temperature is adjusted to 200 C.
Downloaded by [University of Windsor] at 03:31 29 July 2013

Other Costs
Other costs of production are operating labor, maintenance cost, and fixed operating
cost. The techniques used to estimate these costs are described by Turton et al. (1998).
The fixed operating costs are independent of changes in the production rate, and they
include local taxes, insurance, and depreciation that are charged at a constant rate
even when the plant is not in operation. In the calculation of the fixed operating costs,
an average tax of 35% is assumed and the straight line depreciation method for a
period of 10 years is used.

Economic Evaluation of the Base Process


The base process described above is evaluated according to the economic model. The
gross root capital investment of the process is estimated to be 6.01 MM$ for the
year 2010. The utility requirement based on the energy balances on each piece of
equipment is $3.33 million=yr. More information about each type of utility required
in this process is presented in Table IV. Almost 79% of utility cost belongs to the
steam. The electricity has the lowest portion of the utility cost (2.1%).
By assuming a stream factor of 0.95, the yearly revenue of the process is estimated
to be $98.78 million=yr. The revenue of the process consists of three parts: cumene
stream as the main product, high-pressure (HP) steam generated in the reactor,
and streams 9 and 14 as fuel. Figure 2 illustrates the contribution of each part in
detail. Although the mass flow rate of these streams is 10.67% of the total raw
material entering the process, the total income of fuel is about 2.3%. This shows that
the reduction in the mass flow rate of streams 9 (fuel gas) and 14 (DIPB) can greatly
enhance the economic condition of the cumene process. Stream 9 consists of 233 kg=h
propane and 851 kg=h unreacted propylene. Since the raw materials of this process
are valuable, it will be concluded that if the reaction condition of the process is chan-
ged, the economic condition of process can be enhanced greatly. For instance, it is
desirable to seek a reaction condition in which the mass flow rates of unreacted com-
ponents and DIPB are minimized and the total profit of the process is maximized.
Therefore, we focused on the reaction condition and parameters that can influence
the conversion and the selectivity of reactions. Some other modifications can be done
to reduce the utility requirement and the capital investment of the process, but these
1382 H. R. Norouzi and S. Fatemi

Table IV. Details of capital investment and utility requirements of base and
optimized processes

Base Optimized
process process
Capital investment (MM$) Total 6.010 10.2
Heat exchangers 1.980 2.35
Towers 1.400 1.69
Heater 0.750 3.76
Reactor 0.625 1.10
Vessels 0.350 0.46
Pumps 0.825 0.84
Utility requirements (MM$=yr) Total 3.330 3.83
Steam 2.540 2.83
Downloaded by [University of Windsor] at 03:31 29 July 2013

Cooling water 0.292 0.354


Electricity 0.066 0.073
Fuel 0.320 0.420
Catalysta 0.112 0.154
Revenue (MM$=yr) Total 98.78 103.82
Cumene 96.50 101.53
Fuel production 2.305 1.165
Steam generation 0.991 1.130
NPV (MM$) 0.19 4.79
Present value ratio 1.03 1.42
Payback time (year) 7.9 4.5
a
Catalyst price was chosen to be 15 $=kg.

Figure 2. Revenue contributions of cumene process. (Figure provided in color online.)


Economic Optimization of Cumene Production 1383

modifications are less influential than the cost of raw materials and product costs of
the process.

Sensitivity Analysis of the Process


Manipulating Variables
The main limitation of this process is the cumene purity and yield. To enhance the
selectivity of the reaction, two approaches are possible. The first approach is to
increase the ratio of benzene=propylene in the reactor by enhancing the recycle of
benzene and consequently minimizing DIPB production. The increase in benzene
recycle directly reduces the residence time of reactants, thus decreasing the conversion
per pass in the reactor. This leads to an increase in utility requirements and size of the
equipment.
Reduction of reactants’ residence time as a result of benzene recycle enhance-
Downloaded by [University of Windsor] at 03:31 29 July 2013

ment can be compensated for by increasing the reactor volume. On the other hand,
a larger reactor needs much more catalyst, which must be replaced yearly, and there-
fore higher operating cost will be required.
The second approach is to maintain the temperature of the reactor low enough
to limit DIPB production as discussed above. This will increase the selectivity of
cumene over DIPB, whereas the conversion per pass of the reactor will decrease,
which will result in less purity of cumene in the product.
Since the reaction is taking place in the vapor phase, an increase in pressure will
raise the concentration of all components and will increase the rate of both desirable
and undesirable reactions. High-pressure systems need more capital investment for
the equipment used in the process.
These contradictory effects of the four above-mentioned variables on the overall
yield, capital investment, and operating cost of the process make it hard to decide
which factor to adjust and how much it must be changed. The decision is made
according to the economic model discussed before. Finding the operational con-
ditions that maximize the net profit of the process is the goal. The NPV of projects
is a cumulatively discounted cash position at the end of the project and is greatly
influenced by the level of fixed capital investment, revenue of the process, cost of
manufacturing, plant life, tax rate, and inflation.
Before performing sensitivity analysis of the influencing factors on the NPV,
there must be some constraints in the process to accomplish the aim of this process,
which is manufacturing cumene with 99.5 mol% purity. The following constraints are
imposed on the process:

. First constraint: Since water is available at 35 C for cooling purposes, the top
temperature of the benzene fractionator column should not fall under 50 C.
. Second constraint: The concentration of benzene must not exceed 0.005 in the
bottom product of the benzene fractionator column that enters the cumene
purification column in order to guarantee that pure cumene can be obtained in
the second column.
. Third constraint: The molar purity of cumene must not fall under 0.995 in the
second column.
. Fourth constraint: The temperature of the flash drum is adjusted in a way that the
first and the second constraints do not fail.
1384 H. R. Norouzi and S. Fatemi

Sensitivity Analysis
Figure 3 shows the sensitivity of the NPV to the inlet temperature of the reactor.
Benzene recycle, number of tubes in the reactor, and pressure are 106 kmol=h, 234,
and 30.5 bar respectively. As can be seen, the NPV approaches its maximum value
at 360 C due to the increase in overall conversion and reduction of the mass flow rate
of stream 9. However, further increase in the inlet temperature of the reactor increases
the rate of the DIPB reaction and more cumene and propylene are converted into
DIPB. In addition, increase in the reactor inlet temperature requires a bigger fired
heater and cooler (higher capital investment) and much more utilities. These factors
are reflected in the NPV, and the analysis shows that the best inlet temperature is
360 C. Therefore, the next analysis will be done at 360 C for benzene recycle.
Figure 3(b) illustrates the sensitivity of NPV to the benzene recycle at 360 C, with
234 tubes in reactor and an operating pressure of 30.5 bar. When benzene recycle is
increased in the process, more capital investment and utilities are required (negative
Downloaded by [University of Windsor] at 03:31 29 July 2013

effects on the NPV), while the selectivity of cumene over DIPB is increased. Thus,
increasing benzene recycle from 70 to 150 improves the NPV due to much more

Figure 3. Sensitivity analysis of the NPV to (a) the inlet temperature of the reactor, (b)
benzene recycle, (c) number of tubes in the reactor, and (d) operating pressure of the reactor.
(Figure provided in color online.)
Economic Optimization of Cumene Production 1385

cumene produced in the process. Further increase in benzene recycle causes much
more reduction of conversion, and therefore the NPV of the process decreases at
higher rates of benzene recycle.
The effect of volume on the NPV of the process is illustrated in Figure 3(c). The
reactor temperature and benzene recycle are fixed at 360 C and 150 kmol=h for study-
ing the effect of reactor volume. The residence time of reactants and the overall con-
version of reactions increase with increasing reactor volume. Hence, less unreacted
components leave the process and a better economic condition can be obtained, which
encourages us to use a larger reactor in the process. On the other hand, a larger
reactor needs a higher level of capital investment and more catalyst, which should
be replaced yearly. Therefore, the final decision is made according to the NPV that
suggests 700 tubes in the reactor when other factors are fixed.
Pressure is the fourth effective factor that is analyzed, and its effect on the NPV of
the process is illustrated in Figure 3(d). The net effect of pressure rise in the reactor is
Downloaded by [University of Windsor] at 03:31 29 July 2013

an increase in the reaction rate and the residence time of reactants. As it can be seen in
this figure, at very low pressure, the NPV is very low and it rises with increasing the
pressure up to 30 bar, whereas further increase in the pressure reduces the NPV. This
is mainly due to the fact that by increasing the pressure beyond 30 bar, the second
reaction dominates and less cumene is produced.

Design of Experiments for Economic Optimization


According to the above discussion, it can be concluded that the inlet temperature of
the reactor, benzene recycle, and volume and pressure of the reactor greatly influence
the process economics. All these analyses were performed with one variable at a time.
Since the importance of the level of each variable and their probable interaction are
not clear, it is necessary to apply an appropriate statistical design to study the effects
of parameters simultaneously. In the first step of our statistical analysis, full factorial
design was applied to determine the impact of effective parameters on NVP at two
levels. All statistical calculations were done by Minitab 15 statistical software.

Full Factorial Design


According to full factorial design, 24 evaluations were conducted to determine the
significant effects of the main factors and their binary interactions on NVP. A central
point was added to evaluations in order to investigate the curvature behavior of the
variables in the model. A total number of 17 evaluations were done according to the
operating conditions tabulated in Table V. The NPVs were calculated and tabulated
in this table as the response of each evaluation.
The analysis of variance (ANOVA) was carried out on the NPV values to deter-
mine the significance of single-factor and two-way interactions. The significance level
of 5% was considered for determining the significant variables. The results are pre-
sented in Table VI. The test values of Fischer and the p-values were considered as
decision criteria. Those factors with 5% and lower p-values would be considered to
have significant effect on the NPV and the variables with p-values greater than 5%
are reported as nonsignificant parameters. As can be seen in Table VI, all individual
variables exhibit significant effect on the NPV. Among the binary interactions,
Temp  Recycle and Recycle  Press have shown significant effect on the response
of the process.
1386 H. R. Norouzi and S. Fatemi

Table V. Full factorial design table and the NPV as response

Evaluation Temp Recycle Ntube Press


no. (A) (B) (C) (D) NPV
1 340 80 468 29 1.81
2 370 80 468 29 6.54
3 340 180 468 29 3.56
4 370 180 468 29 1.93
5 340 80 936 29 0.28
6 370 80 936 29 7.79
7 340 180 936 29 2.94
8 370 180 936 29 0.64
9 340 80 468 35 1.06
10 370 80 468 35 8.22
Downloaded by [University of Windsor] at 03:31 29 July 2013

11 340 180 468 35 3.43


12 370 180 468 35 1.28
13 340 80 936 35 2.87
14 370 80 936 35 10.29
15 340 180 936 35 2.10
16 370 180 936 35 0.11
17 355 130 702 32 0.99

The Pareto chart of standardized effect for significant variables is shown in


Figure 4. The dashed red lines indicate the significance limit of each term, which is
considered to be 5%. As can be seen, recycle has a negative and the most significant
effect on the NPV. Temperature, pressure, and number of tubes, in decreasing order,
have negative effects on the NPV. The binary interaction of Temp  Recycle has a
positive effect on the NPV. This confirms our previous discussion on temperature
and benzene recycle. This means that increasing the temperature and recycle leads
to higher conversion in the reactor and selectivity of cumene over DIPB respectively,

Table VI. ANOVA analysis of NPV for determining significant variables

Variable Mean sum of squares F-test p-value Significance


Temperature 96.482 1083.01 0.00 Sig.
Recycle 159.075 1785.63 0.00 Sig.
Ntube 7.967 89.42 0.00 Sig.
Press 9.875 110.85 0.00 Sig.
Temp  Recycle 32.234 361.83 0.00 Sig.
Temp  Ntube 0.032 0.35 0.578 Non-sig.
Temp  Press 0.124 1.39 0.291 Non-sig.
Recycle  Ntube 0.258 2.89 0.15 Non-sig.
Recycle  Press 3.832 43.01 0.001 Sig.
Ntube  Press 0.228 2.56 0.171 Non-sig.
Curvature 4.440 49.83 0.001 Sig.
Residual error 0.121
Economic Optimization of Cumene Production 1387
Downloaded by [University of Windsor] at 03:31 29 July 2013

Figure 4. Pareto chart of standardized effect for significant variables. Dashed lines indicate
statistically significant bond for each term.

while the enhancement of recycle reduces overall conversion. Thus, simultaneous


change in these variables can have a positive effect on the NPV. Recycle  Press is
another two-way interaction that has a positive effect on the NPV. Clearly, increasing
the pressure increases the rate of both reactions and increasing the recycle leads to
more production of cumene than DIPB.
Adding a center point to the factorial design enables us to verify the curvature
effect in the response surface. Curvature usually exists when factor settings are close
to a maximum or minimum response value. According to Table VI (p-value < 0.05),
the curvature does exist in the response value of the NPV. The apparent significant
curvature effect in factorial design requires more evaluation to derive a reliable
model with quadrature term of each factor. Therefore, further analysis for the
optimum NPV will be continued by response surface methodology to obtain a
second-order model.

Response Surface Method (RSM)


For performing RSM with four independent variables, the number of evaluations
is determined by the following equation (Vining, 1998):

Nevaluation ¼ 2k þ 2k þ n0 ð6Þ

where k is the number of independent variables and n0 is the number of evalua-


tions at the center point (here one). The first term shows the number of corner
points and the second term depicts the number of axial points. In the RSM, the
corner points exactly correspond to full factorial design with the same independent
variables and levels. Thus, in this stage 2k additional evaluations are required in
order to perform an RSM analysis (Vining, 1998). Therefore, eight additional eva-
luations were conducted on the axial points, and the results were added to Table V
for statistical analysis. Table VII is the RSM table of variables and the calculated
1388 H. R. Norouzi and S. Fatemi

Table VII. Central composite design table of variables and


NPV for each point

Evaluation Temp Recycle Ntube Press


no. (A) (B) (C) (D) NPV
1 340 80 468 29 1.81
2 370 80 468 29 6.54
3 340 180 468 29 3.56
4 370 180 468 29 1.93
5 340 80 936 29 0.28
6 370 80 936 29 7.79
7 340 180 936 29 2.94
8 370 180 936 29 0.64
9 340 80 468 35 1.06
Downloaded by [University of Windsor] at 03:31 29 July 2013

10 370 80 468 35 8.22


11 340 180 468 35 3.43
12 370 180 468 35 1.28
13 340 80 936 35 2.87
14 370 80 936 35 10.29
15 340 180 936 35 2.10
16 370 180 936 35 0.11
17 340 130 702 32 3.25
18 370 130 702 32 0.26
19 355 80 702 32 4.94
20 355 180 702 32 2.03
21 355 130 468 32 2.29
22 355 130 936 32 0.79
23 355 130 702 29 2.32
24 355 130 702 35 1.37
25 355 130 702 30 0.99

NPV for each evaluation. A full quadrature model is introduced on the base of
important terms with 5% significance level by the analysis of variances. The final
results of this analysis are listed in Table VIII, and the model Equation (7) is
derived for NPV. Adjusted coefficient of determination more than 98% suggests
that the determined model is precise. Among second-order terms of manipulated
variables, benzene recycle (B) has significant effect on the NPV. This seems to
be inconsistent with sensitivity analysis done for each variable above. This can
be explained by two facts. First, sensitivity analysis reflects the effect of change
in an individual variable while others are fixed, and it does not take into account
the effect of simultaneous changes of the variables. Second, as observed in the pre-
vious section, benzene recycle has the most significant effect on the NPV, and the
effect of the second-order terms of other parameters are negligible compared to the
benzene recycle. Variance analysis of the response (NPV) against the operational
parameters revealed that binary interactions of recycle-pressure and recycle-
temperature are significant, whereas the other binary interactions showed negligible
effects. Therefore, the following equation was derived as a quadratic model for
Economic Optimization of Cumene Production 1389

Table VIII. Results of linear regression; 95% of confidence level was


considered for calculations

Confidence interval
Coefficient t-value p-value of coefficients
Constant 139.9 16.39 < 0.001 (121.6, 158.2)
Temp 0.403 18.53 < 0.001 (0.448, 0.357)
Recycle 0.419 6.44 < 0.001 (0.556, 0.282)
Ntube 0.003 6.39 < 0.001 (0.004, 0.002)
Press 0.674 6.21 < 0.001 (0.904, 0.445)
Recycle  Recycle 0.001 13.44 < 0.001 (0.0009, 0.0011)
Temp  Recycle 0.0019 12.04 < 0.001 (0.0018, 0.0022)
Recycle  Press 0.0033 4.15 0.001 (0.002, 0.005)
Downloaded by [University of Windsor] at 03:31 29 July 2013

NVP response with coefficient determination of 99%:

NPV ðMM$Þ ¼ 139:88  0:403A  0:419B  3:036  103 C  0:674D


þ 1:89  103 AB þ 3:26  103 BD  1:11  103 B2 ð7Þ
2
R ¼ 99%; R2adj ¼ 98:6%

Figure 5 presents the contour plots of the binary interactions on NVP derived from the
model. The plots demonstrate that an optimum recycle ratio at lower levels of tempera-
ture and lower levels of pressure would be the appropriate condition to achieve the
highest NPV. Therefore, the response surface model of NPV can be employed for
the prediction of optimal operational conditions to achieve the maximum NPV. The
optimization was performed on a model equation (Equation (7)) by the Simplex

Figure 5. Contour plots of NPV vs. benzene recycle and temperature when number of tubes is
702 and pressure is 30 bar (a), and benzene recycle and pressure when temperature is 355 C
and number of tubes is 702 (b).
Downloaded by [University of Windsor] at 03:31 29 July 2013

Table IX. Stream table of optimized process

Fuel
Benzene Propylene 3 7 8 gas 10 11 12 Cumene DIPB
Temperature 25 25 39.6 340 341 81 81 49.4 175.6 174.3 234.6
( C)
Pressure (bar) 1 11.471 1 29.75 29.75 1.75 1.75 1.75 1.75 1 1
Mole flow 105 110.27 245 355.27 250.7 8.255 242.445 140 102.445 101.1 1.345

1390
(kmol=h)
Mass flow 8201.932 4650.856 19771.4 24422.2 24422.2 461.6 23960.6 11569.4 12391.1 12172.9 218.2
(kg=h)
Benzene 105 212.367 212.367 109.648 2.281 107.367 107.367 0.001 0.001 trace
Propylene 105 0.708 105.708 1.138 0.43 0.708 0.708 trace
Propane 5.27 8.994 14.264 14.264 5.27 8.994 8.994 trace
Cumene 22.93 22.93 123.798 0.274 123.525 22.93 100.595 100.595 trace
DIPB 0.001 0.001 1.852 0.001 1.851 0.001 1.85 0.505 1.345
Economic Optimization of Cumene Production 1391

method (Vining, 1998), and the optimal conditions were determined when the benzene
recycle was 141.4 kmol=h and temperature, Ntubes, and pressure were 340 C, 468, and
29 bar respectively. The maximum predicted value of NPV was calculated as $4.56 mil-
lion by the quadratic model of Equation (7), whereas $4.79 million was obtained from
the economic model at the same process conditions. Details of the capital investment,
utility requirements, and revenue of the optimized process are listed in Table IV. As can
be seen in this table, the capital investment and the utility requirement of the optimized
process increase from $6.01 to 10.2 million and from $3.33 to 3.83 million=yr respect-
ively. But due to the better conditions in reactor that favor more production of cumene,
the revenue of the process is increased greatly. The annual income from cumene is
increased by $5.03 million=yr and steam generation by $0.139 million=yr in the
optimized process, while the annual income from fuel production is decreased by
$1.11 million=yr. The increase in steam generation is mainly due to two reasons. First,
more moles of reactants are converted in the reactor via exothermic reactions, and
Downloaded by [University of Windsor] at 03:31 29 July 2013

second, the benzene recycle of this process is more than the base process. As a result,
a great deal of energy must be removed from the reactor, and consequently more steam
is generated in the reactor. The stream specifications of the optimized process are listed
in Table IX.
Recently, Luyben (2010) performed an economic optimization for the cumene
production process on SPA catalyst. The optimized condition in his study is different
from the optimized condition in this study; his optimized design uses a higher reactor
temperature and volume than those in this study. There are some differences between
the procedures followed that lead to these differences:
1. Since the emphasis of this study was on the reaction section and to adjust the
reaction conditions in a way that less DIPB and more cumene were produced with
the highest profit, two distillation columns were not optimized and Turton’s
design (overdesigned columns) was used instead. Thus, higher separation cost
was imposed on the process, which favored a smaller reactor in the process.
2. In the process optimized by Luyben (2010), some heat was recovered in the
feed-effluent heat exchanger (FEHE) and a smaller electric heater or the like
was used, while a fired heater was used in this study. This can change the operat-
ing and capital investment of the process noticeably.
3. Two different economic functions were used in Luyben’s study and this study,
although both economically optimized the same process. The economic function
of this study was based on 10 years of plant life, 10% interest, labor, maintenance,
capital investment, raw material cost, utility cost, revenue, etc., all of which were
reflected in the NPV, while Luyben’s was based on 3 years of payback time and
raw material, capital investment, revenue, and operating cost, all of which were
reflected in the TAC. Although they both show the profitability of the process,
they are not identical functions of the same variables with the same sensitivity
to the manipulated variables.

Conclusion
The process of manufacturing cumene from propylene and benzene on solid phos-
phoric acid catalyst was studied. A comprehensive economic model that considers
all influencing factors was established to assess the profitability of this process.
Among all criteria for economic assessment, the NPV that reflected the whole eco-
nomics of the process was chosen. The NPV of the base process was calculated as
1392 H. R. Norouzi and S. Fatemi

$0.19 million. The low value is mainly due to inappropriate operating condition of the
reactor. The economic optimization of this process was conducted by a statistical
method. Four major factors including benzene recycle, inlet reactor temperature,
number of tubes in the reactor (reactor volume), and reactor pressure were chosen
as manipulating variables. Full factorial design and response surface methodology
were utilized to identify the significant factors and obtain a mathematical model
for NPV, respectively. The variance analysis showed that benzene recycle and inlet
temperature of reactor have the most influence on NPV, while the two others have
the least influence. The mathematical model of NPV was obtained. and optimization
was performed on the basis of a significant quadratic model. The maximum value of
NPV was achieved when the benzene recycle, inlet temperature, reactor volume, and
reactor pressure were 141.4 kmol=h, 340 C, 468 tubes, and 29 bar respectively. The
maximum value of NPV was predicted as $4.56 million by the response surface model,
whereas $4.79 million was obtained from the process simulation at these optimized
Downloaded by [University of Windsor] at 03:31 29 July 2013

conditions. The capital investment and the utility requirement of the optimized pro-
cess were increased from $6.01 to 10.2 million and from $3.33 to 3.83 million=yr,
respectively. The enhanced economic state of the process was caused by reduction
in purge and DIPB streams ($1.11 million=yr) in the process and more production
of cumene ($5.03 million=yr).

References
Branan, C. R. (1995). Rules of Thumb for Chemical Engineers, Gulf Publishing Company,
Houston, Tex.
Branchu, S., Forbes, R. T., York, P., and Nyquvist, H. (1999). A centeral composite design to
investigate the thermal stabilization of lysozyme, Pharm. Res., 16(5), 702–708.
(2010). Chemical Engineering, Economic Indicators, www.che.com, April 2010, p. 80.
(1998). Engineering Data Book, 11th ed., Gas Processors Suppliers Association, Tulsa, Okla.
Li, C., Bai, J., Cai, Z., and Fan, O. (2002). Optimization of a cultural medium for bacteriocin
production by Lactococcuslactis using response surface, J. Biotechnol., 93, 27–34.
Llompart, M. P., Lorenzo, R. A., Cela, R., and Paré, J. R. J. (1997). Optimization of a
microwave-assisted extraction method for phenol and methylphenol isomers in soil
samples using a central composite design, Analyst, 122, 133–137.
Luyben, W. L. (2010). Design and control of the cumene process, Ind. Eng. Chem. Res., 49,
719–734.
Oughlis-Hammache, F., Hamaidi-Maouche, N., Aissani-Benissad, F., and Bourouina-Bacha,
S. (2010). Central composite design for the modeling of the phenol adsorption process in
a fixed-bed reactor, J. Chem. Eng. Data, 55, 2489–2494.
(1999). Perry’s Chemical Engineers’ Handbook, 7th ed., McGraw-Hill, New York.
Prinsloo, N. M. (2007). Preparation of a solid phosphoric acid catalyst from low-quality
kieselguhrs: Parameters controlling catalyst quality and performance, Ind. Eng. Chem.
Res., 46, 7838–7843.
Rigas, F., Dritsa, V., Marchant, R., Papadopoulou, K., Avramides, E. J., and Hatzianestis, I.
(2005). Biodegradation of lindane by Pleurotusostreatus via central composite design,
Environ. Int., 31, 191–196.
Seider, W. D., Seader, J. D., Lewin, D. R., and Widagdo, S. (2010). Product and Process
Design Principles, 3rd ed., Wiley, Hoboken, NJ.
Turton, R., Bailie, R. C., Whiting, W. B., and Shaeiwitz, J. A. (1998). Analysis, Synthesis, and
Design of Chemical Processes, Prentice-Hall PTR, Upper Saddle River, NJ.
Ulrich, G. D., and Vasudevan, P. T. (2006). How to estimate utility costs, Chem. Eng., April,
66–69.
Economic Optimization of Cumene Production 1393

Vining, G. G. (1998). Statistical Methods for Engineers, Duxbury Press, Pacific Grove,
Calif.
Wee, L. L., Annuar, M. S. M., Ibrahim, S., and Chisti, Y. (2011). Enzyme-mediated
production of sugars from sago starch: Statistical process optimization, Chem. Eng.
Commun., 198(11), 1339–1353.
Zivorad, R. L. (2004). Design of Experiments in Chemical Engineering, Wiley-VCH,
Weinheim.
Downloaded by [University of Windsor] at 03:31 29 July 2013

View publication stats

Вам также может понравиться