ChE class and home problems

The object of this column is to enhance our readers’ collections of interesting and novel problems in chemical
engineering. We request problems that can be used to motivate student learning by presenting a particular principle
in a new light, can be assigned as novel home problems, are suited for a collaborative learning environment, or dem-
onstrate a cutting-edge application or principle. Manuscripts should not exceed 14 double-spaced pages and should
be accompanied by the originals of any figures or photographs. Please submit them to Dr. Daina Briedis (e-mail:
briedis@egr.msu.edu), Department of Chemical Engineering and Materials Science, Michigan State University, East
Lansing, MI 48824-1226.

Optimization Problems
Brian J. Anderson, Robin S. Hissam, Joseph A. Shaeiwitz, and Richard Turton

O
West Virginia University • Morgantown, WV 26506-6102
ptimization is often considered to be an advanced, economic optimum pipe diameter[6] and reflux ratio[7] are
highly mathematical, and sometimes a somewhat also available. Other examples of optimization problems
obscure discipline. While it is true that many ad- are available, but these do not involve an economic objec-
vanced optimization techniques exist, optimization problems tive function.[8-10] The problems presented here all involve
can be developed that are suitable for undergraduates at all an economic objective function.
levels. Two of these problems will be described in this paper,
and many others are available on the web.[1] A pedagogy is Types of Problems
described that requires students to identify the trends of the Three types of optimization problems are available, and
components of the objective function and to understand how they are summarized in Table 1. The ones highlighted in ital-
trade-offs between these components lead to the existence ics are discussed in this paper, and the others are available on
of the optimum. the web.[1] The numbers in parenthesis indicate the number
of different versions available for each problem. All of these
The ability to solve “routine” optimization problems has
have been used successfully in a freshman class designed to
been simplified by advances in computing power over the
develop computing skills appropriate for an undergraduate
last generation. Earlier editions of current design textbooks[2]
chemical engineering student. Most of these problems would
presented a sequence of optimization techniques aimed at
also be suitable for assignments or projects in unit operations
minimizing the number of cases that had to be considered
to close in on the optimum. Now, it is possible to perform Table 1
optimization calculations involving numerous cases with a Available Optimization Problems
few clicks of a mouse, and an entire chemical process can be Single Multi-variable Projects
simulated and results exported to a spreadsheet in a matter Variable
of minutes. Pipe diam- Absorber Generic chemical
Several optimization examples are routinely discussed eter (2) process (2)
in undergraduate textbooks; however, the objective func- Reactor/ Batch reactor/pre- Geothermal energy
tion does not usually involve economics. These examples preheater heater (2)
(2)
include optimum interstage compressor pressure,[3] optimum
insulation thickness,[4] and identifying conditions for the Reflux ratio Staged compressors Fuel production
from biomass (4)
optimum selectivity.[5] Qualitative representations of the
Brian J. Anderson is the Verl Purdy Faculty Fellow and an assistant professor in Joseph A. Shaeiwitz received his B.S. degree from the University of Delaware
the Department of Chemical Engineering at West Virginia University. His research and his M.S. and Ph.D. degrees from Carnegie Mellon University. His profes-
experience includes sustainable energy and development, economic modeling sional interests are in design, design education, and outcomes assessment. Joe
of energy systems, and geothermal energy development as well as molecular is a co-author of the text Analysis, Synthesis, and Design of Chemical Processes
and reservoir modeling. (3rd Ed.), published by Prentice Hall in 2009.
Robin S. Hissam received her B.S. and M.S. degrees in materials science and Richard Turton, P.E., has taught the senior design course at West Virginia
engineering from Virginia Tech and her Ph.D. in materials science and engineer- University for the past 24 years. Prior to this, he spent five years in the design
ing from the University of Delaware. After a post-doctoral fellowship in chemical and construction industry. His main interests are in design education, particulate
engineering and applied chemistry at the University of Toronto, Robin joined the processing, and modeling of advanced energy processes. Richard is a co-author
Chemical Engineering Department at West Virginia University. Her research is of the text Analysis, Synthesis, and Design of Chemical Processes (3rd Ed.),
in production of protein polymers for application in tissue engineering, biomin- published by Prentice Hall in 2009.
eralization, and biosensors. © Copyright ChE Division of ASEE 2011

144 Chemical Engineering Education

classes or as problem assignments for the portion of a design
class where optimization is taught.
Problem 1: Bioreactor
Background
A liquid-phase, biological reaction is used to produce an
intermediate chemical for use in the pharmaceutical indus-
try. The reaction occurs in a large, well-stirred, isothermal
bioreactor, such that the reactor temperature is identical to
the inlet temperature. Because this chemical is temperature
Figure 1. Process flow diagram of the feed preheater
sensitive, the maximum operating temperature in the reactor
and bioreactor.
is limited to 65 ˚C by using a heating medium available at this
maximum temperature. The feed material is fed to the reactor The design equation for the heat exchanger is given by:
through a heat exchanger that can increase the temperature of
the reactants (contents of the reactor), which in turn increases Q = M c C p,c (Tc ,2 − Tc ,1 ) = M h C p,h (Th ,1 − Th ,2 ) = UAF∆Tlm ( 4)
the rate of the reaction. This is illustrated in Figure 1. The
where
time spent in the bioreactor (known as the space time) must
be adjusted to obtain the desired conversion of reactant. As the
∆Tlm =
(T
h ,2
− Tc ,1 ) − (Th ,1 − Tc ,2 )
temperature in the reactor increases so does the reaction rate, (5)
thereby decreasing the size (and cost) of the reactor required ln
(T h ,2
− Tc ,1 )
to give the desired conversion. The problem to be solved is (T h ,1
− Tc ,2 )
to determine the optimal value for the single independent
variable; namely, the temperature (Tc,2) at which to maintain and
the reactor (preheat the feed). The costs to be considered are F = log-mean temperature correction factor = 0.8 (assume
the purchase costs of the reactor and heat exchanger and the that this is constant for all cases)
operating cost for the energy to heat the feed. U = overall heat transfer coefficient = 400 W/m2K
Problem Statement The optimum reactor inlet temperature is the one that
It is desired to optimize the preheat temperature for a re- minimizes the equivalent annual operating cost (EAOC). The
actant feed flow of 5,000 gal/h. The feed has the properties EAOC is given by
of water ( ρ = 1,000 kg/m3, Cp = 4.18 kJ/kg ˚C) and enters 2

the heat exchanger at a temperature of 20 ˚C. The reactor EAOC $ / y = ∑ PCi $ (A / P, i, n) 1 / y + UC $ / y (6)
feed is to be heated with a heating medium that is available i=1

at a temperature of 65 ˚C and must leave the heat exchanger where PCi are the purchase equipment costs for the heat
at 30 ˚C. Therefore, the desired reactor inlet temperature is exchanger and reactor, UC is the operating (utility) cost for
adjusted by changing the flowrate of the heating medium. the heating medium, and (A/P, i, n) is the capital recovery
The physical properties of the heating medium are ρ = 920 factor given by
kg/m3, Cp = 2.2 kJ/kg ˚C.
i(1 + i)
n

The reaction rate for this reaction, –rA, is given in terms of the (A / P, i, n) = (7 )
concentration of reactant A (CA) by the following equation: (1+ i)
n
−1
−rA = kCA (1)
For this problem, use i = 7% and n = 12 years.
where The purchase cost of the reactor is given by:
 
3, 500  PCreactor = $17, 000 V 0.85 (8)
k s−1  = 2.5 exp −  ( 2)
   T  K  
where V is the volume of the reactor in m3. The cost of the
The design equation for the reactor is given by: heat exchanger is:

{ }
0.57

V=
vo XA
(3) PCexchanger = $12, 000 A  m 2  (9)
 
k (1− X A )
where A is the area of the heat exchanger in m2. The cost of
where V is the reactor volume (m ), vo is the volumetric flow-
3
the heating medium is:
rate of fluid into the reactor (m3/s), and XA is the conversion
(assumed to be 80% or 0.8 for this reaction). UC $ / h  = $5×106 Q  kJ / h  (10)

Vol. 45, No. 2, Spring 2011 145

 The results should be presented as two plots. The first should by the following equation:
show how each term in Eq. (6) changes with Tc,2, and the −rA = kCA (11)
second plot should show the EAOC (y-axis) as a function of
Tc,2 (x-axis). The report should contain a physical explanation where
of the reason for the trends on these plots.
 
3, 500 
Problem 2: Batch Bioreactor k s−1  = 2.5 exp −  (12)
   T  K  
Background
A liquid-phase, biological reaction is used to produce an The design equation for the reactor is given by:
intermediate chemical for use in the biotech industry. The 1 1
reaction occurs in a large, well-stirred, isothermal bioreac- t s = ln (13)
 −1 
k s  1 − XA
tor, such that the reactor temperature is identical to the inlet  
temperature. Because this chemical is temperature sensitive, where t is the time spent in the reactor and XA is the fractional
the maximum operating temperature in the reactor is set to conversion of reactants to products. The amount of product
55 ˚C. The feed material is fed to the reactor through a heat formed in time t is given as NXA, where N is the number of
exchanger that increases the temperature of the reactants moles of reactant fed to the reactor.
(contents of the reactor), which in turn increases the rate of The energy balance equation for the heat exchanger is
the reaction. This is illustrated in Figure 2. given by:
The reactor runs as a batch operation in which the contents Q = M c C p,c (Tc ,2 − Tc ,1 ) = M h C p,h (Th ,1 − Th ,2 ) (14)
remain in the equipment for a given period of time. The time
spent in the bioreactor must be adjusted to obtain the optimal where
conversion of reactant. Because of the fear of contamination M is the mass of fluid to be heated or cooled (kg)
by pathogens and parasitic fungi, the reactor must be cleaned Cp is the specific heat capacity of the fluid (kJ/kg ˚C)
thoroughly between batch operations. The cleaning time
per batch (tclean) and the cost of cleaning both vary based
on the size of the reactor used.
As the time spent in the reactor increases, the amount
of product also increases but at a decreasing rate. The
problem to be solved is to determine the optimum values
of the two independent variables; namely, the time for the
products to spend in the reactor, or the batch time, and the
reactor size.
For this problem, it is assumed that only standard size
vessels are available (1,000, 5,000, or 10,000 gallons), and Figure 2. Process flow diagram of feed preheater and bioreactor.
that the costs of the feed are fixed. Therefore, the
costs that vary are the revenues from sales, the
reactor cost, and the cost for cleaning.

Problem Statement
It is desired to optimize the production of
product from the reactor. The feed has the
properties of water ( ρ = 1,000 kg/m3, Cp =
4.18 kJ/kg ˚C) and enters the heat exchanger at
a temperature of 20 ˚C. The reactor feed is to
be heated with a heating medium that is avail-
able at a temperature of 65 ˚C and must leave
the heat exchanger at 30 ˚C. The desired reactor
inlet temperature is fixed at 55 ˚C. The physical
properties of the heating medium are ρ = 920
kg/m3, Cp = 2.2 kJ/kg ˚C.
The reaction rate for this reaction, -rA, is given
in terms of the concentration of reactant A (CA) Figure 3. Optimization plot for Example 1.

146 Chemical Engineering Education

 T is the temperature (˚C) The ability to solve “routine”
1 and 2 refer to inlet and outlet conditions, respectively.
h and c refer to the hot and cold stream, respectively. optimization problems
The optimal reactor configuration is the one that minimizes the equivalent has been simplified by
annual operating cost (EAOC). The EAOC is given by:
2 3
advances in computing power
EAOC $ / y = ∑ PCi $ (A / P, i, n) 1 / y + ∑ UCi $ / y − R $ / y
i=1 i=1
(15)
over the last generation.
where PCi are the purchase equipment costs for the heat exchanger and reactor;
UCi are the operating (utility) costs for the heating medium, the cost of the The final results should be presented as two
feed stream, and the cost of cleaning; and R is the revenue from sales of the plots. The first plot should show how each
product. For this problem, use i = 0.07 and n = 12 years. term in Eq. (15) changes with the batch time,
t, and the second plot should show the EAOC
The purchase cost of the reactor is given by
(y- axis) as a function of t (x-axis). The report
PCreactor = $17, 000 V 0.85 (16) should contain a physical explanation of the
reason for the trends on these plots.
where V is the volume of the reactor in m3. The cost of the heat exchanger may
be taken to be equal to 20% of the cost of the reactor from Eq. (16).
Optimization Problems
The cost of the heating medium is given by:
In Problem 1, the optimum reactor feed
UC heating $ / h  = 5×10−6 Q  kJ / h  (17 ) temperature is to be determined. There is
a trade-off, which is necessary to obtain
where Q is the heat duty obtained from Eq. (14).
an absolute maximum or minimum in the
The price of the feed is $2/mol, the value of the product is $10/mol, and the objective function (EAOC) as the decision
molar density (concentration) of both feed and product is 100 mol/m3. The variable (reactor feed temperature) varies.
cost of cleaning the reactor is given by In this case, at higher temperatures, it costs

 Vreactor  gal  more to heat the reactor feed, but, since the
UCclean $ / cleaning  = 1, 000 $ / cleaning  1 + 0.5  (18) reaction rate increases with temperature, the
 1, 000  gal 
 reactor cost is lower because a smaller reac-
and the time to clean a reactor is tor is needed. Additionally, at higher reactor
feed temperatures, a larger heat exchanger is
 Vreactor  gal 
 needed. Students can develop a spreadsheet
t clean  h  = 4  h  1 + 0.5  (19)
 1, 000  gal  that varies the reactor inlet temperature and

plot the EAOC vs. the reactor inlet tem-
perature. This plot is illustrated in
Figure 3. They can also plot EAOC
vs. reactor cost, heating medium
cost, and heat exchanger cost to
see the trends. This is illustrated
in Figure 4. The trend for the heat
exchanger clearly illustrates how the
heat exchanger cost goes to infinity
as the reactor feed temperature ap-
proaches the heating medium inlet
temperature, causing the log-mean
temperature driving force to go to
zero and the heat exchanger area to
become infinite. This is an example
of why it is important for students
to analyze a series of data and un-
derstand the trends. It is possible to
solve this entire problem on Excel
using the Solver tool; however,
Figure 4. Component optimization trends in Problem 1. much of the understanding/synthesis

Vol. 45, No. 2, Spring 2011 147

Since these problems have been used components lead to the existence of the optimum. That is why
methods, such as using the Excel Solver, are not emphasized,
successfully in a freshman class for and making plots to investigate trends is emphasized. Once
the trends are understood, Excel Solver can be used to obtain
several years, we believe they can be a more exact value of the optimum.
We have used these problems as part of a freshman class
used anywhere in the curriculum. taken by students who know that they are interested in chemi-
cal engineering. Other students take a college-wide program-
of the problem is lost. We believe that optimization is more ming class. In our class, students are taught computer skills
than finding an answer. An understanding of the underlying applicable to chemical engineering, mostly using the advanced
trends is essential. features of Excel in addition to some elementary program-
ming techniques and algorithms. All assignments are based on
It is also possible to illustrate how changes in operating con- industrially relevant chemical engineering problems. Some of
ditions change the optimum. In a problem similar to Problem these problems also appear in the optimization chapter of our
1,[11] if the reaction kinetics are increased (pre-exponential textbook.[11] Since these problems have been used successfully
factor increased to 7.0 and the activation energy reduced to in a freshman class for several years, we believe they can be
3300), the optimum temperature shifts down to about 35 ˚C. used anywhere in the curriculum.
Many different versions of this and other problems can be
created by changing some parameters or by changing the Since all students in chemical engineering do not take the
economics. We use different versions of these for different class in which these problems are assigned, assessment of
groups in the same class. During oral presentations, we ask their long-term impact is difficult. The freshmen do a good
them to explain why the optima differ. job on these problems, and they seem to appreciate the actual
chemical engineering application compared to their peers in
In Problem 2, there are two decision variables (bivariate the programming class.
optimization) due to the batch processing. Therefore, this
problem in slightly more complex than Problem 1, and it il- Additional optimization problems are available on the
lustrates that there may be more than one decision variable. web.[1] It is observed that virtually an infinite source of these
One decision variable is the reactor volume, which in this problems could be obtained by manipulating some of the
case is limited to three standard sizes (an arbitrary number), values given in these problems.
and the other decision variable is the processing time. The
trade-off is that for longer processing times, more product is Conclusion
made, but fewer batches can be made per year. For a larger Two example optimization problems that are believed to be
reactor, more product can be made per batch, but fewer suitable for all levels of chemical engineering students have
batches can be made per year due to the longer cleaning been presented. These problems do not require advanced
time. Although this problem does not include it, the reactor mathematical techniques; they can be solved using typical
feed temperature could also be varied, as in Problem 1, to software used by students and practitioners, such as Excel.
create a three-variable optimization. In this problem, it turns These problems involve an economic objective function with
out that the optimum is the 10,000 L reactor with a
reaction time of 9.1 h, at about 97% conversion, as
is illustrated in Figure 5. For higher conversions, the
additional processing time is long enough to make
the annual product revenue drop. This problem also
illustrates some of the issues associated with batch
processing to students who might be very used to
continuous processes. Figure 5 also illustrates a bi-
variate optimization plot, with the x-axis containing
one decision variable with several curves indicating
the second decision variable.

Discussion
We believe that an important part of the pedagogy
of optimization is for students to understand the
trends of the components of the objective function
and to understand how trade-offs between these Figure 5. Optimization plot for Example 2.

148 Chemical Engineering Education

component capital and operating cost terms. An important Ed.), Prentice Hall PTR, Upper Saddle River, NJ, 2003, Chapter 4.3F
part of the pedagogy of these problems is an understanding 5. Fogler, H.S., Elements of Chemical Reaction Engineering (4th Ed.),
Prentice Hall PTR, Upper Saddle River, NJ, 2006, Chapter 6
of how the trends of the components terms in the objective 6. de Nevers, N., Fluid Mechanics for Chemical Engineers (3rd Ed.),
function contribute to the trade-off involved in most optimi- McGraw Hill, New York, 2005, Chapter 6
zation problems. 7. Peters, M.S., K.D. Timmerhaus, and R.E. West, Plant Design and
Economics for Chemical Engineers, (4th Ed.), McGraw Hill, New
York, 2003, Chapter 9
References 8. Barolo, M., “Batch Distillation Optimization Made Easy,” Chem. Eng.
1. <http://www.che.cemr.wvu.edu/publications/projects/index. Ed., 32(4), 280 (1998)
php#opt> 9. Smart, J., “Using the Evolutionary Method to Optimize Gas Absorber
2. Peters, M.S., and K.D. Timmerhaus, Plant Design and Economics for Operation,” Chem. Eng. Ed., 38(3), 204 (2004)
Chemical Engineers, (3rd Ed.), McGraw Hill, New York, 1980, Chapter 10. Mitsos, A., “Design Course for Micropower Generation Devices,”
10 Chem. Eng. Ed., 43(3), 201 (2009)
3. Sandler, S.I., Chemical, Biochemical, and Engineering Thermodynam- 11. Turton, R., R.C. Bailie, W.B. Whiting, and J.A. Shaeiwitz, Analysis,
ics (4th Ed.), Wiley, New York, 2006, Chapter 4, Problem 4.21b Synthesis, and Design of Chemical Processes (3rd Ed.), Prentice Hall
4. Geankoplis, C., Transport Processes and Separation Principles (4th PTR, Upper Saddle River, NJ, 2009, Chapter 14 p

Vol. 45, No. 2, Spring 2011 149