Chemical engineering education

http://cee.che.ufl.edu/ ( Journal Site )
MISSING IMAGE

Material Information

Title:
Chemical engineering education
Alternate Title:
CEE
Abbreviated Title:
Chem. eng. educ.
Physical Description:
v. : ill. ; 22-28 cm.
Language:
English
Creator:
American Society for Engineering Education -- Chemical Engineering Division
Publisher:
Chemical Engineering Division, American Society for Engineering Education
Publication Date:
Frequency:
quarterly[1962-]
annual[ former 1960-1961]
quarterly
regular

Subjects

Subjects / Keywords:
Chemical engineering -- Study and teaching -- Periodicals   ( lcsh )
Genre:
serial   ( sobekcm )
periodical   ( marcgt )

Notes

Citation/Reference:
Chemical abstracts
Additional Physical Form:
Also issued online.
Dates or Sequential Designation:
1960-June 1964 ; v. 1, no. 1 (Oct. 1965)-
Numbering Peculiarities:
Publication suspended briefly: issue designated v. 1, no. 4 (June 1966) published Nov. 1967.
General Note:
Title from cover.
General Note:
Place of publication varies: Rochester, N.Y., 1965-1967; Gainesville, Fla., 1968-

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
oclc - 01151209
lccn - 70013732
issn - 0009-2479
sobekcm - AA00000383_00165
Classification:
lcc - TP165 .C18
ddc - 660/.2/071
System ID:
AA00000383:00165

Full Text








EDITORIAL AND BUSINESS ADDRESS:
Chemical Engineering Education
Department of Chemical Engineering
University of Florida Gainesville, FL 32611
PHONE and FAX: 352-392-0861
e-mail: cee@che.ufl.edu

EDITOR
Tim Anderson

ASSOCIATE EDITOR
Phillip C. Wankat

MANAGING EDITOR
Lynn Heasley

PROBLEM EDITOR
James O. Wilkes, U. Michigan

LEARNING IN INDUSTRY EDITOR
William J. Koros, Georgia Institute ,

-PUBLICATIONS BOARD

CHAIRMAN *
E. Dendy Sloan, Jr.
Colorado School of Mines

MEMBERS
Pablo Debenedetti
Princeton University
Dianne Dorland
Rowan University
Thomas F. Edgar
University of Texas at Austin
Richard M. Felder
North Carolina State University
Bruce A. Finlayson
University of Washington
H. Scott Fogler
University of Michigan
Carol K. Hall
North Carolina State University
William J. Koros
Georgia Institute of Technology
John P. O'Connell
University of Virginia
David F. Ollis
North Carolina State University
Ronald W. Rousseau
Georgia Institute of Technology
Stanley I. Sandler
University of Delaware
Richard C. Seagrave
Iowa State University
C. Stewart Slater
Rowan University
Donald R. Woods
McMaster University


Winter 2006


Chemical Engineering Education


Volume 40


Number 1


Winter 2006


D DEPARTMENT
8 Columbia University
Carl C. Gryte, Lenora Babb, Edward F. Leonard

> EDUCATOR
2 Kirk Schulz of Mississippi State University

> CLASSROOM
14 Numerical Problem Solving Using Mathcad in Undergraduate
Reaction Engineering,
Satish J. Parulekar
66 Engineering Analysis in the Chem-E-Car Competition
Randy S. Lewis, Aliakbar Moshfeghian,
and Sundararajan V Madihally


D CLASS AND HOME PROBLEMS
60 Data Analysis Made Easy With DataFit,
James R. Brenner

> RANDOM THOUGHTS
38 The Way to Bet, Richard M. Felder

> LABORATORY
24 Experimental Air-Pressure Tank Systems for Process Control Education,
Christopher E. Long, Charles E. Holland, and Edward P. Gatzke

40 A Flexible Pilot-Scale Setup for Real-Time Studies in Process Systems
Engineering,
Chanin Panjapornpon, Nathan Fletcher, and Masoud Soroush

46 Mechanical Testing of Common-Use Polymeric Materials With an
In-House-Built Apparatus,
Cristiana Pedrosa, Joaquim Mendes, Ferndo D. I

54 A Nonlinear, Multi-Input, Multi-Output Process Control Laboratory
Experiment,
Brent R. Young, James H. van der Lee, and William Y Svrcek

> LEARNING IN INDUSTRY
32 Partnering With Industry for a Meaningful Course Project,
Rhonda Lee-Desautels, Mary Beth Hudson, Ralph S. Young



23 Call for Papers



CHEMICAL ENGINEERING EDUCATION (ISSN 0009-2479) is published quarterly by the Chemical Engineering
Division, Amercan Society for Engineering Education, and is edited at the University of Florida. Correspondence
regarding editorial matter, circulation, and changes of address should be sentto CEE, Chemical Engineering Department,
University of Florida, Gainesville, FL 32611-6005. Copyright 2006 by the Chemical Engineering Division, American
Societyfor Engineering Education. The statements and opinions expressed in this perodical are those of the writers and not
necessarily those of the ChE Division, ASEE, which body assumes no responsibility for them. Defective copies replaced if
notified within 120 days of publication. Write for information on subscription costs and for back copy costs and availability.
POSTMASTER: Sendaddress changes to ChemicalEngineering Education, ChemicalEngineeing Department, University
of Florida, Gainesville, FL 32611-6005. Periodicals Postage Paid at Gainesville, Florida, and additional post offices.










] M educator


Mississippi State s





Kirk Schulz


JULIE M. LEMONS
Mississippi State University Mississippi State, MS 39762-9595 :,
balance. It's one thing that all of us strive for daily, but it seems to
come quite naturally for Kirk Schulz, dean of the James
Worth Bagley College of Engineering at Mississippi State Uni-
versity. If you know him or have ever worked with him, then you know
what I'm talking about. He seems to get more accomplished in one day
than most of us do in a week. And although his dedication to his work is
obvious, his commitment to his family is even more evident.
Kirk has been called approachable, accessible, and a great listener, and
he has also been characterized as quick-witted, decisive, and driven-
quite a combination for a 42-year-old dean. One MSU engineering alum-
nus and advisory board member recently spoke of Kirk saying, "He is .
open to new ideas and new ways of doing things, but more than that he
follows through and implements those ideas." Kirk is not one to sit around
and wait for things to happen; he is a man of action and integrity with a
strong desire to move his engineering program into the Top 50.
Faculty excellence is at the top of every dean's list of priorities. Deans
want to see their professors excel as mentors and leaders, knowing that
their efforts are molding the young minds of our world's future leaders.
Kirk knows that faculty excellence is the foundation of a top-tier institu-
tion and such excellence is crucial to the success of all its programs. Rec-
ognizing faculty for their accomplishments and successes in teaching,
research, and service is important to him and to the continued excellence
of the engineering program at Mississippi State. Simply put, Kirk is an
administrator who enables faculty to do what they do best-teach others -
to become contributing leaders in their fields.
MSU Provost Dr. Peter Rabideau remarks, "Kirk is one of the most
enthusiastic and energetic administrators with whom I have been associ-
ated. He supports quality as the number one issue, and he wants to work
with faculty, staff, and students to develop a vision that will move MSU
engineering to the next level of excellence."
Kirk updates alumni and friends of the
As the new dean of the Bagley College of Engineering, Kirk has estab- Bagley College during Engineering Day on
lished the Academy of Faculty Fellows to recognize those who have risen campus this past fall.

Copyright ChE Division ofASEE 2006
2 Chemical Engineering Education





































to the rank of Fellow in their respective engineering profes-
sional societies. This past year, 15 faculty members were in-
ducted into the academy. Another way that the college is re-
cruiting new faculty and celebrating current faculty mem-
bers' excellence is by honoring them through endowed posi-
tions. Currently, the Bagley College has 23 endowed posi-
tions, and in Kirk's inaugural year as dean he has made it a
priority to fill these important faculty positions in the col-
lege. He has already filled nine endowed chairs or professor-
ships as well as several other key leadership positions around
the college including two associate dean positions, two de-
partment head positions, and a center director. With a stel-
lar leadership team now in place, Kirk hopes to help Mis-
sissippi State's Bagley College of Engineering gain the
recognition it deserves as one of the top research univer-
sities in the country.

HIS BACKGROUND
Born into a family of educators in Portsmouth, Va., during
the summer of 1963, Kirk was destined to become a teacher.
He was raised in Norfolk, Va., by parents who were both uni-
versity employees-his father a mathematics professor at Old
Dominion University, his mother an associate registrar and
director of compliance.
In speaking of his parents, Kirk characterizes his father as
an "outstanding teacher who does a great job engaging his
students," and his mother as "very active as a researcher,
doing very creative and innovative research in historical
geography."


"Kirk is one of the most enthusiastic and
energetic administrators with whom I have
been associated. He supports quality as the
number one issue, and he wants to work
with faculty, staff, and students to develop
a vision that will move MSU engineering to
the next level of excellence."
MSU Provost Dr. Peter Rabideau

As a student at Norfolk Christian High School, Kirk first
realized that he had a natural bent toward mathematics and
chemistry. He excelled as a leader throughout his high school
years, and as a graduating senior he was awarded both the
leadership and science awards. It was during his high school
years that Kirk first gained hands-on engineering experi-
ence while participating in the Soapbox Derby. He learned
how to machine different steels, use a lathe, and paint. Build-
ing a car and steering it down a long track seemed to suit
him. He won the local Tidewater Virginia race, going on
to represent his home state in the World Championships
in Akron, Ohio, in 1977.
After graduating from Norfolk Christian in 1981 with 44
of his peers, Kirk actually wanted to study medicine and was
advised to work as a volunteer in the emergency room to see
if he would really like being a doctor. He recounts, "After
about one year of volunteering at Norfolk General Hospital,
I decided that I really wanted to go into engineering and not
medicine." Kirk's father had always spoken highly of the
engineering profession, which initially planted the thought
in his head. After a family friend took it upon himself to give
Kirk a tour of the chemical plant where he worked, the deci-
sion to pursue engineering was solidified. In 1984, with three
years of undergraduate work under his belt, Kirk decided to
transfer from Old Dominion to Virginia Tech to pursue his
chemical engineering degree, receiving his bachelor's in 1986.
At Virginia Tech, Kirk was actively involved in the Baptist
Student Union, serving as president and statewide vice presi-
dent. Kirk realized as an undergraduate that he wanted to
continue his education at the graduate level and become an
educator himself. While in graduate school, Kirk remained
active in various organizations, even helping start a Scout
troop at his church in Blacksburg, Va. The troop is still
active today.
Kirk was the first Ph.D. student that Dr. David F. Cox ever
advised at Virginia Tech, so needless to say they both learned
much during their time together. Kirk was conducting research
in metal oxide surface chemistry. Dr. Cox recounts his time
with Kirk: "He helped set up my laboratory, did all the inter-
facing of computers and experimental apparatus, and wrote
all the code for data collection and analysis. When he wasn't


Winter 2006





















C .0


"It


Kirk as a Ph.D. student under Dr. David F Cox at Virginia Tec
working in the lab, he was founding a new graduate student
association in our department, working on committees with
the dean's office, and generally helping out with any depart-
mental, college, or university task with which he was ap-
proached. It drove me crazy.
"I was a young assistant professor, and I kept thinking he
would be able to accomplish so much more if only he would
focus more of his effort on his research. In the end, we pub-
lished 12 papers from his graduate research. Fifteen years
and many students later, I feel extremely lucky when I have a
graduate student that manages to produce one-third of what
Kirk did. With each passing year my appreciation increases
for Kirk's skill, dedication, and technical abilities. Unfortu-
nately for him, I did not realize how good I had it when he
was working in my lab."
Dr. Cox explains, "When Kirk was still a Ph.D. student, I
asked him about his career goals in the hope that I could of-
fer him some sound career advice. He told me even then (I
kid you not) that he wanted to become a dean. I suggested to
him that such a career path would be a waste of a good scien-
tific career. Thank goodness he ignored my advice.
"Throughout his academic career he has continued to turn
out excellent scientific work even as he became more and
more involved in administration. Kirk is the most well-
rounded academician I know. I continue to be amazed by his
ability to perform so well in so many different arenas. These
days, I go to Kirk for career advice rather than the other way
around. Whenever I am asked, I take credit for all of Kirk's
success, but the bottom line is I have learned more from him
than he ever learned from me."
Kirk feels blessed to have had a large number of people
play an integral part in the success of his career. "My father
and my research adviser, Dr. David Cox, both stressed the
need to work hard and to finish the things I started. Tom
4


Owens, my first department chair at the University of North
Dakota, really stressed the need to communicate with people
clearly and to set high standards. Ed Fisher, my department
head at Michigan Tech, and Wayne Bennett, former dean at
MSU, both stressed the need to aggressively seek external
support and private gifts for big, visionary ideas."
After receiving his Ph.D. from Virginia Tech in 1991, Kirk
accepted a position as an assistant professor of chemical engi-
neering at the University of North Dakota in Grand Forks, N.D.
His wife, Noel, an associate professor in electrical engi-
neering and TVA Professor at MSU, recalls, "For two years
during my Ph.D. work, I lived 325 miles away from our son
Timothy and him. He was a single parent during the week so I
could get my Ph.D. It was a challenging time, but we made it
through because of Kirk's commitment to my advancement."
The two met during Noel's freshman year at Virginia Tech
and connected during a mission trip to Kentucky that Kirk led.
"She has been my number one fan and has been willing to
pick up her research program and move it when an opportu-
nity came up for me," says Kirk of his wife of 18 years.
She echoes the sentiment saying, "Kirk has always been an
extremely supportive spouse. Since he is several years ahead
of me in his professional career, he has been a mentor all
along-sometimes making the mistakes, then warning me
about them."


Chemical Engineering Education


/X_2
,.;E












i4" ';I.


Above, as director of the Swalm School
of Chemical Engineering, Kirk taught
the unit ops labs. Here, he is assisting
students as they review their results.
Right, all active in Scouting, Kirk and
sons Tim (now 15) and Andrew (now 11)
have long enjoyed spending time
together outdoors.


Kirk spent four years at the University of North Dakota
before moving to Michigan Tech University as assistant pro-
fessor in 1995. His leadership abilities were quickly recog-
nized, and he was promoted to associate professor in 1998.
That same year, he assumed the chairmanship of the Depart-
ment of Chemical Engineering.
After several years heading chemical engineering at Michi-
gan Tech, Kirk accepted a position at Mississippi State Uni-
versity in 2001 as director of the Dave C. Swalm School of
Chemical Engineering and holder of the Earnest W.
Deavenport, Jr., Chair.
Dr. Wayne Bennett, dean emeritus of the Bagley College
of Engineering, recalls, "I recognized his leadership skills
when he interviewed for the director's position of the Swalm
School of Chemical Engineering. Under his leadership, the
school progressed on every front. The undergraduate programs
flourished, graduate enrollments increased, and the research
set new records."
Success doesn't come without ample opportunities. This
idea is something that Kirk recognizes, and he acknowledges
that there have been several people throughout his career who
have given him an opportunity when conventional wisdom
would have said otherwise. One such individual was Bob
Warrington, dean of engineering at Michigan Tech, who was
willing to give a 35-year-old associate professor a chance to
be a department chair very early in his career. Another ad-
ministrator willing to take a chance was MSU Provost Dr.
Peter Rabideau, who agreed to let a 41-year-old lead the
university's flagship college. "Both of these individuals gave
me an opportunity in my career that many people never get,"
comments Kirk.
Winter 2006


HIS MOTIVATION
His motivation is simple: to make a difference in the lives
of others, especially the students. Working with students and
faculty is what Kirk enjoys the most about his job.
"Teaching at the university level in my mind is the real
chance to make a difference in someone's life. If you ask an
engineer who they had for chemical engineering reactor de-
sign, most can give you a name-even 20 years later. When
you talk with alumni, you realize just how big an influ-
ence we have on a person during their formative years,"
says Kirk.
Kirk receives great joy and satisfaction from seeing his
former doctoral students become successful in their careers.
Dr. Alan Nelson, one of Kirk's first doctoral students, is now
an associate professor and associate chair in the University
of Alberta's Department of Chemical and Materials Engineer-
ing. Regarding Kirk's abilities as a professor, Dr. Nelson says,
"I have a great deal of respect for Kirk, not only because he
is a scholar and educator, but because he has been and con-
tinues to be a benevolent mentor to so many individuals. I
would certainly not be where I am today without the research
supervision and professional guidance Kirk has provided to
me over the years."
Dr. Nelson goes on to say, "Kirk's ability to maintain per-
sonal and professional balance should be a model for all new
chemical engineering faculty. He is a case study of how to be
an effective and efficient educator, while not sacrificing his
personal goals or time with his family."
Professors work hard to impart some vast wisdom or knowl-
edge to the students they teach. The students will, of course,
remember some of a teacher's meager efforts in the lab and
5










classroom and how he or she graded them, but in the end it is
one's character that students take note of the most. It is how
they are treated as students and how faculty respond to life
and all its cliches that demonstrate what true "balance" in
life is all about.

HIS FAMILY
Family is the most important part of Kirk's life. One favor-
ite family activity at the Schulz house is Scouting, which prob-
ably has a little to do with having an Eagle Scout for a dad.
Currently, Kirk is the assistant Scoutmaster of Troop 14 in
Starkville. Son Timothy (15) is a Life Scout, while Andrew
(11) is in Webelos. The Schulz family also en-
joys spending time outdoors as well as traveling
to a wide variety of destinations, such as Disney "Wh e
World, London, and San Diego-not to men- am as
tion following the MSU Bulldogs to the SEC take c
Basketball Tournament in Atlanta. Il ,


Kirk shares a good relationship with Noel's
parents, and they have always been supportive
of his professional career. While at Virginia Tech,
Kirk was a student worker for Dr. Charles
Nunnally, the assistant dean of engineering at Vir-
ginia Tech-and Noel's father.
Being a dean and having a wife who is an as-
sociate professor oftentimes means that the week-
nights are booked with college or university
events. It is not unusual for Timothy and An-
drew to be present at some of these functions,
and they are often a crowd favorite at such
events.


to say, Kirk took it in stride and Timothy and Andrew
were greatly amused.
HIS CAREER
Throughout his career Kirk's talent for administration and
leadership have been recognized and have afforded him some
wonderful opportunities. Of course, no one begins their ca-
reer as a dean of a college; there are dues to pay and work to
be done along the way. Kirk has traveled that path, working
his way from a summer school chemical engineering instructor
at Virginia Tech all the way to dean of a major state university.
Each engineering program that Kirk has been involved in


never I
.ked, I
:redit for
Vr;r'-


ll UJJ At "f C
success, but the
bottom line is I
have learned
more from him
than he ever
learned from
me."

-Dr. David Cox,
Kirk's Ph.D. advisor
at Virginia Tech


has benefited from his leadership. As an admin-
istrator, much of his time and energy have been
devoted to improving alumni relations, grow-
ing graduate programs, increasing the diversity
of faculty and students, recruiting new faculty,
and increasing external funding. With a strong
desire to see faculty collaborating across the col-
lege at Michigan Tech, he assisted in the initia-
tion of the Carbon T:,. I i 1i .- \ Center. This is a
multidisciplinary research center involving fac-
ulty from chemical engineering, mechanical en-
gineering, civil engineering, and chemical en-
gineering tc'. iii. l_. in research focused on
polymer composites. At Mississippi State, Kirk
saw a need to make pursuing a Ph.D. in chemi-
cal engineering more enticing. So as director
of the Swalm School of Chemical Engineering,
he led the efforts to establish the first direct-
admit doctoral program, increasing the number
of Ph.D. students from three to 15 in three years.


Kirk made it a point when he first became dean of the
Bagley College to let his staff and faculty know that the dean's
office was a family-friendly environment. "We all have fami-
lies and from time to time there are family situations that
come before work. I want my staff and faculty to know that I
support them personally and am here if they ever need any-
thing," said Kirk.
Throughout the academic year Mississippi State will often
host MathCount competitions and Science Fairs, and you will
see the Schulz boys in and out of the dean's office visiting
Dad. One of the most enjoyable visits to campus for his sons
was during Engineering Week at MSU this past year. It was
toward the end of the week and just so happened to be the
same day as Timothy's MathCount competition. All week,
engineering students had purchased tickets to "pie" a faculty
member or fellow student as part of a fund-raiser. The activ-
ity on the Drill Field drew a crowd of curious onlookers, as
Kirk and two other department heads sat bravely in an-
ticipation of the firing squad of students and faculty that
stood before them. It is not often that students, or faculty
for that matter, get to throw a pie at their dean. Needless


Kirk has shared his knowledge and research findings
through 42 journal articles that he has authored or co-writ-
ten; he has presented numerous conference papers and given
100 presentations. He has had over $1.8 million in funded
research projects and has provided guidance to six doctoral
students and 14 master's students.
Kirk is currently co-chairing the Chemical Engineering
Division of ASEE's 2007 Summer School for Chemical En-
gineering Faculty, which will be hosted on the campus of Wash-
ington State University. His service to engineering professional
organizations does not end here. He is very involved with ABET
and has served as a program evaluator, a member of the AIChE
Education and Accreditation Committee, and now a member
of the Engineering Accreditation Commission (EAC).
He has served as the division chair and program vice chair
for ASEE's New Engineering Educators; he has held offices
such as secretary/treasurer, program chair, and director for
the Chemical Engineering Division; and he is a senior mem-
ber of AIChE. Since 2003, he has been a member of the Ad-
visory Board for the University of Tennessee's Department
of Chemical Engineering.


Chemical Engineering Education

















Kirk, center,
along with
Drs. Tony Vizzini
(head of aerospace
engineering) and
Glenn Steele (head of
mechanical engineer-
ing), getting pie-
faced to the delight
of students, faculty,
and especially Kirk's
sons Timothy and
Andrew during last
year's E-week
activities.


Along the way, Kirk has been recognized in numerous
capacities for his work. Early in his career, he was named
Outstanding Professor of the Year at the University of North
Dakota. In 1997, while at Michigan Tech, Kirk was recog-
nized by the Chemical Engineering Division of ASEE with
the Raymond W. Fahien Award for his outstanding accom-
plishments and commitment to his profession. He has also
received an NSF CAREER Development Award (1995-1999),
the Dow-ASEE Outstanding Young Faculty Award (1995-
1996), and ASEE's Outstanding Teaching Award for the North
Midwest Section (1999). His alma mater, Virginia Tech,
named him as the Outstanding Young Alumnus in 2001. The
next year, Kirk was recognized by Mississippi State Univer-
sity as Outstanding Professor in Chemical Engineering.

HIS VISION
Over the past 20 years, Kirk has observed a dramatic change
taking place in the field of engineering education, noting the
increased importance now placed on the quality of under-
graduate teaching and faculty development at research uni-
versities. He has also seen the communication skills of engi-
neering graduates improve dramatically from what they were
20 years ago.
In the Bagley College of Engineering a strong emphasis is
placed on both of these areas. Through the newly established
Center for Engineering Student Excellence, programs such
as the Shackouls Technical Communication Program, Six
Sigma, Study Abroad, congressional internships, and the En-
trepreneurship Program emphasize the need to take technical
degrees a step further. Students are encouraged to comple-
ment their technical degrees with programs such as these,
providing them with better global awareness as well as im-
Winter 2006


proved communication and business skills.
Kirk's vision for the Bagley College is to see it recognized
as one of the top research institutions in the United States.
He knows that this must be done by investing resources in
carefully selected areas where MSU can be internationally
renowned. He strongly believes that MSU will be one of the
leading institutions in providing a diverse engineering work-
force, and he is committed to the education of African-Ameri-
can engineering students at all degree levels. The Bagley
College of Engineering, in fact, already ranks among the top
15 schools in graduating African-American engineers, and
Kirk wants his college's ranking to go even higher.
Ultimately, Kirk wants to see the MSU Bagley College of
Engineering thrive and become one of the Top 50 engineer-
ing colleges in the country. The core of his strategic plan fo-
cuses on providing first-rate education while continuing to
recognize faculty for their research endeavors and teaching
excellence. He believes that MSU has an obligation to the
state of Mississippi and the nation, and to support growth
and economic development with the expertise and knowledge
found in the faculty of the Bagley College of Engineering.

HIS MOTTO
The advice he extends to each of the engineering students
at Mississippi State is the same adage he chooses to live by:
"Seek out challenging opportunities during your career-look
for something that people say can't be done-and then go
out and do it."
Now 42 and almost a year into his deanship, is there any-
thing that Kirk would have done differently along the way?
Not a chance. 7









] l department


Chemical Engineering at



Columbia University

CARL C. GRYTE,
with contributions from
LENORA BABB AND
EDWARD F. LEONARD
Columbia University
New York, NY 13699-5705
C olumbia Chemical Engineering is
quintessentially New York, a cen-
tral part of a university whose le-
gal name is Columbia University in the
City of New York. It is a university united
to its city perhaps more than any other ur-
ban university in the United States. In 2005
Columbia Chemical Engineering cel- 1
ebrated its 100th anniversary, but it is
rooted even farther back into the chemi-
cal, financial, and public works history of
its home city; the special characteristics of -_ ,
contemporary chemical engineering at Co- ._
lumbia trace far into the department's, the ...--
city's, and the country's past.
A STORIED HISTORY F ., .. '


Columbia's engineering school was -. '
founded in 1864, initially named the f
School of Mines. It originated out of sci-
ence departments that had participated in
the 19th-century struggle in much of the
Western World to reconcile philosophical
and practical views of science.
In 1896, separate schools of engineering,
chemistry, and architecture were set off from CE
the School of Mines, resulting, finally, in
Columbia's first curriculum in chemical en-
gineering being offered by the School of Chemistry in the fall
of 1905 (having been approved the preceding February).
A towering figure at Columbia and in New York at the
time was Charles Frederick Chandler, a Bostonian whose
pivotal education in chemistry was, notwithstanding his ori-
gins, in Germany. There he met some of Europe's leading
8


LEBRATING 100 YEARS-

scientists, including Wohler, Liebig, von Humboldt, and Pas-
teur-all of whom had progressed from a purely philosophi-
cal to a decidedly practical bent. On his return with a doctor-
ate from G6ttingen, Chandler joined Columbia, where he
taught for 46 years.
Copyright ChE Division ofASEE 2006
Chemical Engineering Education










Professor Chandler actually campaigned for 14 years be-
fore 1905 for a program that would produce what we could
now only call chemical engineers. He was strongly resisted
by professors who saw chemistry as pure science, beautiful
in its own right and with deep philosophical meaning. Such
battles raged in many universities and account for the earli-
est realization of chemical engineering as a distinct discipline
being born in technological institutions such as MIT.

INFLUENCES FAR AND NEAR
New York in 1905 was one of the high-thinking, intellec-
tual centers of the young United States. It was an interna-
tional business center where agents of European-especially
German-chemical firms issued and oversaw limited licenses
to operate processes developed in Europe. In reaction, strong
incentives existed for establishing an American capacity
to develop and improve chemical processes, even before
this became a desperate priority when the first World War
broke out.
Professor Chandler was close friends with Nicholas
Murray Butler-Columbia's president from 1902 until 1949.
Both men consulted and were on good terms with the New
York business community centered on Wall Street. Columbia's
role in international business and politics was then, and re-
mains today, preeminent, affecting every department of the
university. These connections, and Professor Chandler's popu-
larity as a teacher deeply involved in his subject, impelled
the Columbia program to flourish.


In 2005 Columbia Chemical Engineering
celebrated its 100th anniversary, but it is
rooted even farther back into the chemi-
cal, financial, and public works history of
its home city.


New York City after the Civil War was also a dirty, over-
crowded, unhealthy, and unsafe place. In addition to his ex-
tensive involvement with the chemical industry, Professor
Chandler played a central role in the public health of New
Yorkers, dealing officially for the city with "the adulteration
of milk, kerosene accidents, gas-factory nuisances, and gen-
eral sanitation," as well as an issue that persists today-lead
in drinking water. Professor Chandler was also very concerned
with the chemical education of physicians and pharmacists
and presented lectures to those professions regularly. We at
Columbia like to think of him as our first biomedically ori-
ented chemical engineer.

BUILDING THE PROGRAM
Records show the first chemical engineering curriculum at
Columbia laid out four solid years of unremitting "chemis-
Winter 2006


An early view of the large electrical generators at
Columbia ChE's now-closed Heat Transfer Research
Facility. For more than 50 years the laboratory tested
electrically heated models of nuclear fuel-rod assem-
blies. Practically every configuration used in the
Western World's boiling-water nuclear reactors was
tested at this facility. Tests were run late at night to
reduce dimming of lights in Manhattan.

try, engineering, metallurgy, mathematics, mechanics, phys-
ics, and mineralology," having presumed prior preparation
in .Ily.chi geometry, plane trigonometry, chemistry, physics,
freehand drawing, English literature, composition and gram-
mar, American and English history, French, and German."
Professor A.W. Hixson, who joined the faculty in 1922
and later became the preeminent department historian of Co-
lumbia, put forth the claim that despite the preexistence of
other programs entitled chemical engineering, Columbia's
was "the first well-balanced and completely integrated cur-
riculum in chemical engineering to be established in
America." In what is arguably a less disputable first, only
five years after its 1905 founding Columbia admitted stu-
dents to study for the degree "doctor of philosophy in chemi-
cal engineering."
Professor Chandler's handpicked colleague and later suc-
cessor, Milton C. Whitaker, also became a leading figure in
chemical engineering education. He was recognized with two
honorary degrees and was an early president ofAIChE.
Engendered by these early innovators, Columbia Chemi-
cal Engineering's current specializations all have origins and
histories that reach back to the department's founding.
Polymer Surfaces
In the earliest years, polymers were mostly natural and the
coursework was concerned with materials such as cellulose,
gutta percha, and rubber. The esters of cellulose were already
in wide use, however, and as the department was being
founded Leo Baekeland was inventing the phenol-formal-
dehyde resin that was to bear his name. Indeed, Baekeland
was an honorary professor in the department, an advisor to
Columbia's President Butler, and an overseer of the chemi-
cal engineering program nearly until his death in 1944.




































With the nation's drive for independence from European
tc,, ii 1. 4. \, major emphasis was placed on process and plant
design. A steady stream of doctoral theses based on process
and plant design flowed out of the department from 1915
through the beginning of the second World War. While these
dissertations covered a wide range of processes, many were
concerned with raw materials for synthetic substances.
In 1939, James M. Church arrived at Columbia and for
more than 20 years ran an undergraduate unit processes labo-
ratory in which students conducted carefully scaled-down
versions of industrial, mostly organic, processes. The real
resurgence of interest in polymers began in the mid-'60s,
however, with the hiring of George Odian in 1966 and Harry
Gregor in 1967. They were joined by Carl Gryte in 1972
and Christopher Durning in 1983. The trend continued when
Ben O'Shaughnessy, a condensed-matter physicist, joined
the faculty in 1988, followed by Rasti Levicky in 1998, and
Jeffrey Koberstein in 2000. The lasting theme of this resur-
gence has been an interest in polymer surfaces in an excep-
tionally wide range of applications.
Electrochemical Engineering
Electrochemistry and electrochemical engineering have had
a similarly long run through the department's history. For a
long time the ability to generate electricity from the potential
energy of water far outweighed the ability to transmit elec-
tricity over long distances. In that era, a major center of elec-
trochemical manufacturing evolved at Niagara Falls, N.Y.
The first professor of electrochemical engineering,
Samuel A. Tucker, was appointed in 1910 and rapidly
built up what historian Hixson has called the most com-
10


Left, Columbia's Unit Ops Lab, circa
1929, with students dressed "properly"
for lab work in those days. Columbia's
7. department was one of the earliest
. proponents of the unit operations
" -, concept and such laboratories evolved
continuously along these lines through
the first half of the 20th century. Above,
Professor Elmer Gaden and family. Gaden was named
"Father of Biochemical Engineering" by Chemical
Engineering News in 1971; this photo appeared on the
magazine's cover.
plete electrochemical laboratory in the country. The
strength of a great university was brought to bear on this
enterprise through the influence of Columbia's Depart-
ment of Physics, with its interests in electricity.
In 1922, Colin G. Fink joined the department to begin a
long and distinguished career in electrochemical engineer-
ing. Professor Fink was a 1903 Columbia graduate who sub-
sequently received his Ph.D. in chemistry (from the Univer-
sity of Leipzig) in 1907. Fink's personal research accomplish-
ments were extraordinary, including-during earlier employ-
ment with General Electric-the process for drawing tung-
sten wire that was essential to light-bulb manufacture, as well
as the development of chromium plating. He became the ex-
ecutive secretary of the Electrochemical Society, revitalized
it, and negotiated a home for it on the Columbia campus.
Fink (who retired in 1950) was joined by Henry B. Linford
in 1942. Linford, too, served as executive secretary of the
Electrochemical Society, retiring in 1976. Joining late in
Linford's tenure was Huk Y. Cheh (who retired in 2001 to
become director of research for the Duracell Company). Cheh
was honored in 1982 with the Ruben-Viele chair named in
honor of Samuel Ruben-a protog6 of physics professor
Michael Pupin-who made important contributions to the
electrochemistry of metals. Cheh was joined in 1991 by Alan
West, a specialist in electroplating. West was joined in 2000
by Scott Calabrese Barton, specializing in fuel cells.
Chemical Engineering Education











Bioengineering
Bioprocessing, biochemical engineering, and biomedical
engineering have also long figured in the department's his-
tory. Professor Chandler's influential involvement with the
healthcare community and public health has already been
mentioned. Following in his footsteps was D.D. Jackson,
who lead the department for 23 years, from 1917. Jackson
was trained in chemistry, engi-
neering, and biology, and had
a major interest in biochemical
processes (second only to his
interest in processes for the
production of heavy chemi-
cals). Professor Jackson was
succeeded by the aforemen-
tioned Professor Hixson-who
had a major interest in yeast
chemistry. Such chemistry was
fundamental to much early
bioprocessing.
The real prominence of Co-
lumbia in the area of biopro-
cessing, however, came with
the rise of Elmer L. Gaden in
the years immediately follow-
ing World War II. The discov-
ery of penicillin and its manu- *
facture by fermentation, com-
bined with the extensive de-
mand for it during the war, had
enormously accelerated inter-
est in bioprocessing. Professor
Gaden, an eminently practical
man but also an ideologue,
quickly grasped the signifi-
cance of oxygen delivery in Professor Carl Gryte (on
fermentations and developed, left) and other doctoral s
over many years, methods for radiation source used in
measuring and increasing it.
His students were soon continuing his efforts, both in other
schools and in industry. Juan Asenjo and Alex Seressiotis
followed Gaden, who left Columbia in 1974 to ultimately
become a professor of chemical engineering at the Univer-
sity of Virginia.
Such was the influence of his work that, on the cover of its
May 31, 1971, issue, Chemical and Engineering News de-
clared Professor Gaden, "the father of biochemical engineer-
ing." But beyond Gaden's contribution to biochemical engi-
neering was his early recognition of the development of bio-
medical engineering. Largely through his efforts, by 1965
Columbia had graduate and undergraduate programs in
"bioengineering" with a decidedly medical orientation. The
graduate program was run by an interdisciplinary commit-
Winter 2006


stai
tude
poly


tee, but the undergraduate program remained within chemi-
cal engineering until 1997, when a separate department of
biomedical engineering was established.
Many faculty members contributed to the bioengineering
program, which was seen as a broad effort to focus the tools
and methods of chemical engineering on biological and medi-
cal problems. These influential individuals included Jordan
Spencer, Harry Gregor,
Frank Castellana, Mary
Anne Farrell-Epstein, Huk
SCheh, and Rakesh Jain. No
faculty member was more
involved in this effort than
Edward Leonard, how-
ever, who has worked on
problems related to artificial
organs since 1956-two
years before he joined the
Columbia faculty.
In more recent times, the
department has initiated a
program in genomic engi-
-neering, the first of its kind
in the country. Professor
Jingyue Ju is the director of
sequencing in the Columbia
Genome Center, while Pro-
fessors Ju, Levicky, Banta,
and Leonard are all involved
in research that relates to the
modeling and manipulation
of gene expression.

SHAPED BY
WORLD EVENTS
rcase) with Isao Noda (top
nts installing the Cobalt-60 Thus, the three areas of
mer research (about 1971). current concentration in the
department have extensive
histories. The full story, however, is necessarily a bit more com-
plicated. Two great wars stamped the department indelibly.
World War I matured chemical engineering throughout the
country. Europe, most notably Germany, no longer served as
the fountainhead of chemical engineering-professors were no
longer I IIii,1 cl I" in European universities-and the American
chemical industry moved rapidly toward reliance on chemical
engineers wholly formed in the United States. This shift lead
inexorably to the dominance that American chemical engi-
neering now possesses.
World War II had more specific effects. Columbia was the
home of the Manhattan Project. While the project later moved
to other universities and to the national laboratories, its be-
ginnings were at Columbia, and no other university was as
1]











much affected. Chemical engineers participated, es-
pecially in the early conceptualization of the gaseous
diffusion process for the separation of uranium iso-
topes. While the detailed story remains to be told, Pro-
fessor Thomas Drew was pivotal in these efforts. He
remained at Columbia until 1965.
Another legacy of the Manhattan Project was
Columbia's Heat Transfer Research Laboratory. This
laboratory, founded in 1951, served as the major re-
search and testing facility for thermal-hydraulic de-
sign of nuclear reactors until its closure in 2003. In
major tests it could consume 13 mW of electrical en-
ergy, which had to be accessed out of peak usage times
yet cou 11 dim lights on the west side of Manhat-
tan during tests! The first director was Professor
Charles F. Bonilla. Later directors included a num-
ber of chemical engineering professors, notably Huk
Y. Cheh in the laboratory's later years.

THE BIG PICTURE
Throughout the history of chemical engineering at
Columbia there has been a steady concern with the
"core" of chemical engineering. Notwithstanding the
historical specialties emphasized above, Columbia
Chemical Engineering has always been a broad en-
deavor, not a boutique dedicated to select applications.
The more than 50 individuals who have held profes-


The current faculty. Seated: Scott Banta, Rasti Levicky, Nina
Shapley, Jingyue Ju, Jeff Koberstein. Standing: Edward
Leonard, Alan West (chair), Ben O'Shaughnessy, Scott
Calabrese Barton. Not pictured: Carl Gryte.

sorial positions-too many to mention here-have represented al-
most every area of research: process design and development; en-
ergy conversion; particular unit operations such as distillation, heat
transfer, fluid mechanics, solids separations, extraction, and most
of the rest, as well as kinetics and reactor design; process control
and optimization; and oil and gas recovery.
Columbia Chemical Engineering today has 10 faculty, currently
chaired by Alan West. Table 1 lists their interests. -


TABLE 1
Eminent Faculty in Columbia's Three Principal Areas (current faculty in bold)
Electrochemical Engineering


1911 1922 1942 1970 1984 1991 2000
Samuel Colin Fink Henry B. Huk Yuk Ulrich Alan West Scott Calabrese
Tucker Linford Cheh Stimming Barton


Biomedical Engineering


1866 1946 1958 1977 1983 1988 2000 2001 2004
Charles Elmer Gaden, Edward Rakesh Juan Alex Jingyue Ju Nina Scott
Chandler "Father of Leonard Jain Asenjo Seressiotis Shapley Banta
Biochemical
Engineering"


Polymer Engineering


1914 1917 1939 1966 1967 1972 1983 1988 1998 2000
Leo Baekeland, D.D. Jackson James George Harry P. Carl Chris Ben Rasti Jeff
inventor of Church Odian Gregor Gryte Durning O'Shaughnessy Levicky Koberstein
Bakelite, the first
important thermosetting resin

12 Chemical Engineering Education


















Introduction to

Chemical Engineering Computing
Bruce A. Finlayson
0-471-74062-4 Paper December 2005 360 pages
$54.95 US / $70.99 CAN / 31.50 /-45.80

As chemical engineering technology advances, so does the complexity of
the problems that arise. The problems that chemical engineers and
chemical engineering students face today can no longer be answered with
programs written on a case-by-case basis. Introduction to Chemical
Engineering Computing teaches professionals and students the kinds
of problems they will have to solve, the types of computer programs
needed to solve these problems, and how to ensure that the problems
have been solved correctly.

Each chapter in Introduction to Chemical Engineering Computing
contains a description of the physical problem in general terms and in a
mathematical context, li.h ,, I,-.. step-by-step instructions, numerous
examples, and comprehensive explanations for each problem and program.
This indispensable text features Excel, MATLAB, Aspen PlusT, and
FEMLAB programs and acquaints readers with the advantages of each.

Perfect for students and professionals, Introduction to Chemical
Engineering Computing gives readers the professional tools they need
to solve real-world problems involving:
Equations of state
Vapor-liquid and chemical reaction equilibria
Mass balances with recycle streams
Mass transfer equipment
Process simulation
Chemical reactors
Transfer processes in 1D
Fluid flow in 2D and 3D
Convective diffusion equations in 2D and 3D



@WILEY


intodluctinf to
Chemic Engineeooring Computing


Order a Copy loday!

North, Central & South America
Tel: 877.762.2974
Fax: 800.597.3299
Email: custserv@wiley.com
Internet: www.wiley.com

Europe, Middle East, Africa & Asia
Tel: +44 (0) 1243 843 294
Fax: +44 (0) 1243 843 296
Email: cs-books@wiley.co.uk
Internet: www.wileyeurope.com

Germany, Switzerland, & Austria
Tel: +49 (0) 6201 606 152
Fax: +49 (0) 6201 606 184
Email: service@wiley-vch.de
Internet: www.wiley-vch.de


Winter 2006


AnInnovative Introdluction to










classroom


NUMERICAL PROBLEM SOLVING

USING MATHCAD

in Undergraduate Reaction Engineering



SATISH J. PARULEKAR
Illinois Institute of Te. *1., \' Chicago, IL 60616


With the development and availability of fast,
efficient computers, the role of computing in
analysis and solution of engineering problems and
graphical communication of results has increased dramati-
cally-leading to greater need for computer-application skills
in the curricula and practice of various engineering disci-
plines.'1 Efficient solution of problems is essential for en-
hanced understanding of chemical engineering principles at
all course levels.'11 Commercially available computational
packages, such as Maple, Mathcad, Mathematica, and Matlab,
have considerably reduced the time and effort required for
engineering calculations. Such programs allow engineers with
limited or no formal training in programming to solve rela-
tively complex problems.[2-4]
One of these packages, Mathcad, combines some of the
best features of spreadsheets and symbolic math programs,
allows efficient manipulation of large data arrays, and pro-
vides a good graphical user interface.2,4, 5] Ability to perform
calculations with units is an important feature of Mathcad
for engineering students.[2] While students need to understand
the problem they are trying to solve, they may know little or
nothing about numerical analysis; Mathcad allows them to
work on problems even if they know very little of the
program's syntax.[4] Some of the advanced and special capa-
bilities of Mathcad, such as solution of stiff differential equa-
tions, statistical methods for nonlinear parameter estima-
tion, and programming, have been used in undergraduate
courses.[3, 5-8]


Experience in using Mathcad in the undergraduate chemi-
cal reaction engineering course at the Illinois Institute of Tech-
nology (IIT) is discussed here. Pertinent illustrations are pro-
vided to demonstrate the ease with which problems with vary-
ing complexity can be solved using Mathcad. Example prob-
lems considered for illustration deal with simultaneous solu-
tion of: linear algebraic equations (i.e., kinetic parameter es-
timation); nonlinear algebraic equations (i.e., equilibrium cal-
culations for multiple reactions and steady-state behavior of
isothermal/nonisothermal CSTR with single/multiple reac-
tions); integral equations (i.e., design of steady-state plug flow
reactor, or PFR); integral-algebraic equations; and nonlinear
ordinary differential equations (i.e., solution of conservation
equations for steady-state PFR and unsteady state CSTR).
Based on these illustrations, the benefits of this user-friendly
software in accelerating learning and strengthening the fun-
damental knowledge base should be evident. With hand cal-
culations being replaced by computation, it is more impor-


Copyright ChE Division ofASEE 2006


Chemical Engineering Education


Satish J. Parulekar is a professor of chemical
engineering at Illinois Institute of Technology.
His research interests are in biochemical engi-
neering and chemical reaction engineering. His
research publications include five book chap-
ters and the book Batch Fermentation: Model-
ing, Monitoring, and Control. He has held visit-
ing appointments at the University of Minne-
sota and the National Chemical Laboratory, In-
dia, and faculty associate and faculty research
participation appointments atArgonne National
Laboratory.










tant than ever to consistently validate and verify the results.J91
This is done, where appropriate, in the illustrations that fol-
low. The Mathcad worksheet for each illustration is provided
in a table and contains problem input, solution algorithm,
and presentation of results in appropriate (numerical and/or
graphical) format.

NUMERICAL ILLUSTRATIONS
Illustration 1
This illustration pertains to estimation of kinetic param-
eters using linear regression, which requires solution of sev-
eral simultaneous equations that are linear in unknown
parameters.
Consider the following relation among variables y and x
(j = 1, 2, ... ,m) that is linear in terms of the unknown
parameters 0j ( = 1, 2,..., m).

y=yp+e, yp = xle,+x202 +...+xmne (1)
Information on y and x (j = 1, 2, . m) is available in
the form of n samples (n > m). The parameter estimation


TABLE 1
Worksheet for Illustration 1


'0.42' (0.1 (0.1 1'
0.96 0.2 0.2 1
0.18 0.05 0.05 1
r:= 0.78 PA:= 0.3 PB:= 0.01 Xi:= 1
1.2 0.4 0.02 1
0.28 0.05 0.4 1
2.88 0.5 0.5 1,

X2:1=n(pA) X3:=ln(PB) Y:=ln(r) Q:=stack(X1T, XT, X}T

X:=QT 0:=( XT.X) -I.XT.

k:= exp(01) a:=02 P:=03 rp:= exp(0 Xi +02 X2 +03 X3)
k= 6.652 a 0.997 p= 0.205

4.871x10-3
-1.481x10-3
-9.417x10-3
1.073x10-3
(r-rp)
2.959x10-3
6.006x10-3
-4.098x10-3


Winter 2006


problem then involves finding the parameter set

S(e0=ejj=1, 2, .., m)

for which I2 e is minimized. After some algebra, the nec-
essary and sufficient condition for this can be deduced to be

Ae=b, A=XTX, b=XTY => =( X ) XTy (2)

with


Y1 Xl X12 ... Xim 01
Y2 X21 X22 .. X2m 02
Y x X- x -
Y= X X= nm

Yn LXnl Xn2 x..... XL m


Each column in array X represents the collection of values of
a particular variable x (j = 1, 2, ... m) for different samples.
The goodness of fit of the least squares can be examined by
calculating the relative error for each data point or sample
(ei) defined as ei = (y y y )/y, i = 1,2,..., n.


The specific example considered here pertains to Prob-
lem 5-13 of Foglero101 and involves a three-dimensional
linear fit (m = 3). The dependence of the rate r of a solid-
catalyzed association reaction between A and B on partial
pressures of A and B, PA and ps, respectively, is expressed
as r = kppB with k, a, and P being the kinetic parameters
to be estimated. The units for k, PA' PB, and r are mmol/{ g
cat.h.(atm) a + )}, atm, atm, and mmol/{ g cat.h }, respec-
tively. Upon linear transformation of the expression, a re-
lation linear in terms of three unknown parameters can be
obtained as in Eq. (1), with x, = 1, x2 = In(PA), X3 = ln(ps), y
= In(r), 01 = In(k), 02 = a, and 03 = p. The data for pA'
pB, and r are listed in Table 1, where the Mathcad work-
sheet for this problem is also shown.
Mathcad allows input only of column vectors. Two-di-
mensional arrays can be constructed from column vectors
already introduced using the "stack" feature. The predicted
reaction rates, rP, are compared with the reaction rates avail-
able from measurements, r, and provide a very close fit
(Table 1). The reason for presenting the relevant equa-
tions in this and other illustrations, where necessary, is to
enable the reader to see how the equations to be solved
and the Mathcad syntax are almost identical.l41 In the illus-
trations that follow, the subscript 0 denotes variable values
at the start of a batch reactor or in the feed for a flow reactor.
Illustration 2
This illustration pertains to an autocatalytic reaction and
involves comparison of space times ( ) required for
steady-state isothermal operations of a CSTR and a PFR.
The reaction A+ B -> 2B occurs as per the kinetics r = kCAC.B
15










The feed contains Aand B in the ratio 100:1. For the feed com-
position under consideration, the reaction rate is expressed as

r=kCAf(X), f(X)=(1-X)(0.01+X) (4)
A comparison of the required space times for a CSTR and a
PFR is equivalent to the comparison of the corresponding
Damkohler numbers, Da (= kCAO ) which can be obtained
explicitly in terms of the exit conversion X.. The Mathcad
worksheet for this problem is shown in Table 2. Rather than
calculating Da for one value of X at a time, the Da's for
CSTR and PFR are expressed as a function of X a floating
variable (Table 2). The Da's for a particular Xe are then readily
obtained by plugging the value of Xe into the symbolic solu-
tions. Keeping Xe floating also enables the student to represent
the results graphically over a specified range of Xe (0 < Xe < 1
in Table 2). This beneficial feature in Mathcad is also used in
Illustrations 7 and 8. For minimizing the required space time,
a CSTR is the reactor of choice up to a critical conversion,
Xc, and a PFR beyond this conversion (Table 2). Identifying
X requires solution of an integral-algebraic equation in Xc-
the numerical solution of which is certainly challenging
for an undergraduate student. Using Mathcad, the solu-
tion is obtained rather easily and its accuracy is demon-
strated in Table 2.


Illustration 3
The gas phase reaction, SO2 (A) + L02 (B) -> S03 (C),
occurs as per the kinetics r = kCACB. For the feed composi-
tion under consideration, the reaction rate is expressed as101

r= kC2f(X) f(X)- (1-X)(0.54-0.5X)
(1-0.14X)2

with k being the kinetic coefficient, CAO the feed concentra-
tion of A, and X the fractional conversion of A. The reaction
is carried out in three CSTRs of equal volume in series with
the exit conversion being specified. Computation of the inter-
mediate fractional conversions and the required total space time,
or of the corresponding Da (= kCAot), requires simultaneous
solution of design equations for the three reactors, viz.,

Xi-X(il)--f(Xi)=0, i=1, 2, 3 (6)
In the above, X refers to fractional conversion of A in reactor i
(X0 = 0) and Da corresponds to the total space time for the
three-reactor battery. The Mathcad worksheet for this problem
is shown in Table 3. The solution proceeds by providing initial
guesses for Da, X1, and X2. The validity of the solution is veri-
fied by substituting Da, X1, and X2 generated by the solution
into Eq. (6).


Chemical Engineering Education


TABLE 2
Worksheet for Illustration 2



X1
DapFR(X),:= (xe jo--irx)

DaCSTR(Xe):= (-Xe).(0.01+Xe)
(ilXe)x(O.01+X) 20
20 |


DaCSTR(Xe)
10
DaPFR(Xe)

------------------------------------------
0
0 0.5 1
When is DaCSTR=DapFR? Xe
Xe:= 0.7

Given
Xe 1 X
e dX Xe -0
o (i-X)-(0.01+X) (1-Xe)-(0.01+Xe)
Xc:=Find(Xe) Xc=0.841 DaCSTR(Xc)= 6.222
Verify

Xc 1 X
f ( X).-+ dX- (_ =-4.187x10-8
0 (l-X)-(0.01+X) (1-Xc)-(0.01+Xc)










Illustration 4
This illustration pertains to estimation of the equilibrium composition of a reaction mixture and is adopted from Problem 4.14
of Cutlip and Shacham.E11 The reactions

0.5
H2S (A)>=H2 (C)+0.5S2 (D) Kp= YCYD p.5
YA

2H2S (A)+SO2 (B)t 1.5S2 (D)+2H20 (E) Kp,= Y P05 (7)
YAYB
occur in a gas phase batch reactor. In the above, Kpl and KP2 denote the equilibrium coefficients, P the total pressure, and y, the
mole fraction of species J. The initial pressure Po is 1.2 atm and the initial composition is (I = inerts): YAO = 0.45, YB = 0.25, and
y0o = 0.3. For Kp, = 0.45 atm05 and K2 = 28.5 atm0o5, obtain the composition of the reaction mixture at equilibrium in constant
volume and constant pressure operations of the reactor. Let n0o denote the initial number of moles of species J, while 1 and 2
equal the extents of reactions 1 and 2, respectively, and n, equals the number of moles of J after certain extents of the two
reactions. The expressions for yj's in terms of 1 and 2 then, are: yj= nj/nt, nt= :j n, nto = njo, and J= A, B, C, D, E, I, with
nA= (nAO 1- 2 2), nB= (nBo- 2), nc= (nco+ 1), n D = (nDO+0.5 1+1.5 2), nE= (nEO+2 2), ni= n10, and n,= (nto+0.5 1+0.5 2). The
mole fractions in the equilibrium relations in Eq. (7) and the reactor pressure for constant volume operation are then expressed as

(YAO- P- 22) (YBO P2) P (0.5p1+1.5p2)
YA YB Yc= YD=
Y= -N -' yB= N' N' N= v '

yE=2 W \ =(1+0.5pi+0.5p2), pl= i1, p2=
V "to "to
P=PoW (constant volume), P=Po (constant pressure) (8)


TABLE 3
Worksheet for Illustration 3

X3:= 0.9 Da:= 2 X1:= 0.3 X2:= 0.6
Given

I Da (i-Xi)-(0.54-0.5Xi) =0 X Da (1-X2)-(0.54-0.5X2) 0
3 (1-0.14.-X)2 3 (1-0.14-X2)2


Da (1-X3)-(0.54-0.5X3) 0 ,
rX-X2 XO(ix 4 .) X1 :=Find(Da,Xi,X2)
3 (1-0.14.X3)2

Da= 19.502 X = 0.635 X2 = 0.823


Verify X Da (1-X1)(0.54-0.5X1) 9.212 10)-
3 (1-0.14-XI)2

X1 Da (1-X2)-(0.54-0.5X2) .984 9
X, -X =-3.984x10-9
3 (1-0.14X2)2

Da (l-X3)-(0.54-0.5X3)
Xz-X, -=
3 (1-0.14.X3)2


Winter 2006 17










The Mathcad worksheet for constant volume operation is
shown in Table 4. The equilibrium relations are nonlinear
coupled equations in the dimensionless extents, p, and p2,
initial guesses for which need to be supplied (Table 4). The
extents calculated are substituted into equilibrium relations
to verify that these indeed are satisfied. Computations for
constant pressure operation, not shown here, proceed in a simi-
lar fashion. Illustration 4 reveals to the students the unique-
ness of the physically realizable equilibrium composition for
a given initial composition.
Illustration 5
Illustration 5 pertains to multiplicity of steady states in an
isothermal CSTR. The reaction under consideration, catalytic
hydrogenation of olefins, obeys the kinetics r = CA/(+CA)2,
with r being expressed per unit reactor volume. The operating
conditions for the reactor are: CAO=13 mol/L, V=10 L, vo=0.2


L/s.[12] The Mathcad worksheet for this illustration is shown
in Table 5. The symbolic solution of the steady-state mass
balance for A, viz., (CA -CA )/ =r(CA), reveals that the re-
actor can operate at three steady states. The students observe
that the steady-state mass balance is a cubic equation in the
unknown, CA, and therefore has three solutions, not all of
which may be real. The verification of solutions of the steady-
state mass balance, generated as a vector, follows as usual
and is done at once for all three solutions. The start-up condi-
tions are important in determining which steady state is even-
tually reached. This requires solution of the mass balance for
the transient operation, viz.,


dCA(CA-CA) r(CA), CA(0)=CAi
dt T


The results of computations pertaining to two CAl are shown


TABLE 4
Worksheet for Illustration 4


Po:= 1.2
YBO:=0.25


Kp:= 0.45 Kp2:= 28.5 yA:= 0.45


yTo:= 1-YA -YB


pl:= 0.085


STo = 0.3


p2:=0.132


Given


P1.(0.5.-p +1.5-2)0.5 0.5
(YA P- 2P2) 0.45=0
(yAO-Pl- 2-p2)


(0.5-p1+1.5.p2 )15 2.p 2 P .5
(YAO -pl- 2p2)2 '(YB0 -P2)


P := Find(p, P2)
P2j


28.5=0


p= 0.06 p2=0.157


Verify

P1 -(0.5-pl+1.5p2 )0.5 005
l -p,- +"'P2) 0.45= 1.805 x10-
(YAO -Pl- 2p2)


(0.5.-p1+1.5-p2 1.5 (2.-p)2 05

(YAO -P- 2'2)2 (YBO -P2)


28.5= -4.707 10-7


(YAO -P- 2.P2)
YA: (1+0.5.p +0.5.p2)

(0.5-p1+1.5-p2)
YD: (1+0.5-p+0.5-p2)


(YBO-P2)
YB: (1+0.5. p +0.5'p2)

2-p2
YE: (1+0.5.p +0.5.p2)


SP1
Y: (1+0.5 -p +0.5p2)

YIo
I:= (1+0.5 p +0.5'P2)


YA= 0.068 YB = 0.083 y =0.054 yD =0.24


YE = 0.284 y = 0.271


YA+YB+YC+YD+YE+YI=1


y:=1+0.5-p1+0.5.p2


Equilibrium
Composition


Y=1.109


Chemical Engineering Education










in Table 5. In this illustration and Illustration 6, integration
of appropriate differential equations has been accomplished
using the Runge-Kutta method with adaptive step size
(Rkadapt). Let the steady state concentrations of A be denoted
as CAsl CAs2, and CAs3, with CAs< CAs2< CAs3. The reactor opera-
tion started from C,, (very close to but less than CAs) leads to
the lowest concentration steady state (CAfR -Cas1, Table 5),
while that started from CA2 (very close to but greater than CA2)
leads to the highest concentration steady state (C Af2 )C As3,
Table 5). The steady state corresponding to CA2 is therefore
unstable. Working with other values of C,, the students deduce
that for 0 < CA < CA2, CA converges to CAs1 at large times and
for CA > CAs2 CA converges to CAs3 at large times (additional
computations not shown).

Illustration 6
This illustration pertains to a membrane reactor employed
to obtain higher conversions for reversible reactions, and is
adapted from Example 4-10 of Fogler.1101 A gas phase disso-
ciation reaction A <= B+C is carried out in a steady-state plug
flow reactor, the wall of which consists of a membrane which
allows transport exclusively of B. The feed to the membrane
reactor contains only A, with FAO = 10 mol/min. The reactor


and the feed are kept at 8.2 atm and 500 K. Since A and C
remain in the reaction phase throughout the reactor, it fol-
lows from the reaction stoichiometry that F, = FAO FA. As
there are two independent unit operations (reaction and mem-
brane separation), the two independent mass balances are
those for A and B, viz.,
dF dF
S = r rB (10)
dV dV
The expressions for the volume-specific rates of reaction, r,
and removal of B, rB, areo10]


r=k (CA


CBCC kFA (FAO -FA)FB
Kc I FT F2 '


rB=kBCB=kB1F, FT =(FAO+FB),
FT

kl=kCTO, CI TkBl=kBCTO CTO
Kc RT


with k = 0.7 min-', Kc = 0.05 mol/L, and k = 1 min-'. It is
desired to obtain profiles of FA and FB in a 300 L reactor via
numerical integration of Eq. (10). The Mathcad worksheet


TABLE 5
Worksheet for Illustration 5


v0:=0.2 CAO:=13


V:=10


r(CA CA
(1+CA)


f(CA) (ACA) (A)
r(A


.75153576775218804425
CA:= f(CA) solve, CA 2.1309325629587234809
8.1175316692890884749


Unsteady state CSTR-Which steady state do we reach?


ti:= 0.0


tf:= 10000.0


D(t,C):=(CA0 C1) C-
T (1+C1)2

CAf:= SolNpts,2
Sol:= Rkadapt(CAi2,ti,tf,Npts,D)


Npts:= 200


C =CA


CAil:= 2.1305


Sol:= Rkadapt(CAil,ti,tf,Npts,D)


C Af = 0.752


CAi2:= 2.1315


CAf2 := S lNpts,2


CAf2= 8.118


V
:= 0
V0


S
f1O


Winter 2006










for this illustration is shown in Table 6. The students observe
that the profile of FB exhibits a maximum (Table 6), since B
is not supplied in the feed and is subject to two serial pro-
cesses, namely generation by reaction and removal by mem-
brane. If B is not removed by membrane separation (reac-
tion-only operation, FB = Fc= FAO FA per reaction stoichiom-
etry), working with the driving force for the reaction, the low-
est FA (= FA, corresponding to reaction equilibrium) is calcu-
lated via symbolic manipulations to be 5.528 mol/L (Table
6), which corresponds to 45% conversion of A. From the pro-
file of FA in Table 6, the students observe that for the effluent
from the membrane-wall reactor, FA is much lower than F A
and therefore the conversion of A is much higher. The last
two illustrations deal with multiple reactions.


Illustration 7
This illustration, adopted from Example 6-7 of Fogler,[10
pertains to the series-parallel reactions

M+H-->X+Me, r1=klV-HCM

X+H-> T+Me, r2=k2 CHCx (12)
with M, H, X, Me, and T being abbreviations for mesitylene,
hydrogen, m-xylene, methane, and toluene, respectively. The
reactions are carried out in a CSTR. The feed contains only
M and H. In view of the stoichiometry of these mole-con-
serving reactions, it can be deduced that the concentra-
tions of species influencing the kinetics are related to one


TABLE 6
Worksheet for Illustration 6


P
P:= 8.2 R:=0.082 T:=500 C T:=- C T =0.2
R-T
k:= 0.7 Kc:=0.05 kB:=1.0 FAO:=10 FBO:= 0

kl:= kCTo kBl:=kB CT Cl:= CT IC:= FAO
Kc FBO
Npts:=100 Vi:=0.0 V,:=300.0 F= FA & F2=FB


D(V,F)::


k F C (FAO-F1)
FAO +2 (FAo+F2)2


k F2
FA0 +F2


-CF2
(F


Sol:= Rkadapt(IC,Vi,Vf,Npts,D)


FAO- ) k F
A0 +F2)2 (FAO +F2)

V:=Sol(1) FA:=Sol(2) FB:=Sol(3)


FB


Equilibrium for reaction-only operation

1C1P2)
G(F:= CF (FA F)
FA0 +F2 (FAo +F2


100 200 300


G(F1) sustF =F F (10-Fi)2
G(FI) substitute,2FAO- F--- 4.0000000000000000000 ( 2-
(20 -F) (20- F1)

F 4(10-F)2 10+2.5
(2_4- F)2 solve,F-> FAe:=10-2 FAe = 5.528
(20-F) (20-F2 10-2

9 Chemical Engineering Education










another as


CM,=(a+CH -C), a 2CMO+CXO-CHO (13)
Since there are two independent reactions, one needs to solve
only two mass balances, e.g., those for hydrogen and m-xy-
lene, in conjunction with the stoichiometric relation in Eq.
(13). The Mathcad worksheet for solution of the design equa-
tions is shown in Table 7. The kinetic and operating param-
eter values are10': k, = 55.2 (ft3/lb mol)05/h, k2= 30.2 (ft3/lb
mol)05/h, CHO = 0.021 lb mol/ft3, and CM = 0.0105 lb mol/ft3.
The profiles of CH and Cx are shown in Table 7, with the space
time for CSTR- Tc-being in hours. The profiles reveal that
the concentration of m-xylene, an intermediate, exhibits a maxi-


mum as expected, since it is not supplied in the feed.
For each Te, one has to provide initial guesses for CH and
Cx, and solve the mass balances iteratively. The same initial
guesses may work for certain range of T. This happens to be
the case in this illustration. The solutions of mass balances
are therefore obtained using Tz as a floating variable (Table
7). To verify the solution, the normalized residues associated
with the mass balances for H and X-reH and reX, respec-
tively-are calculated by substituting CH and Cx generated
by the solution into the mass balances. From the definitions
of reH and reX and magnitudes of these displayed in the pro-
files in Table 7, it is evident that the profiles of CH and Cx in
Table 7 are indeed solutions of the mass balances.


TABLE 7
Worksheet for Illustration 7


k:= 55.2 k2:= 30.2 CHo:= 0.021 CMo:= 0.0105 Cxo:=0.0
a:= 2CM +Co -CHo a=0
CH:=0.0089 Cx:=0.00312
Given


(CH Ho kl.(CH)05 .(a+CH-Cx)-0.5-
tc


k2.(CH)05.CX


CX k .(CH )05 .(a+CH -CX).0.5-k2(CH)0.5.C
cS
Soll(T):= Find(CH,Cx) CH():=Soll(^)1 Cx(Te):=Soll(T)2


reH(T):


T[kl { (CH ))05 .(a+CH( T)) x -(0.5 k2 ( H(T) .CX(

CH(T)-CHo


t .Lkl (CH ) .(a+CH(t) Cx(T)) 0.5-k2(CH(t)) CX
reX(): (
cx(z)


0.04 I I


CH(Tc)
- 0.02
10 CX c)0


0


reH(r )

reXrc 10-6
... 1 I-10


0 0.2 0.4 0.6
0 0.2 0.4 0.6


-2-106
-2.10


0 0.2 0.4 0.6


Winter 2006 21


1.10-6


-/-
K ""'''--...










Illustration 8


This illustration, adopted from Example 8-12 of Fogler,n101 pertains
to elementary liquid phase reactions A-kB-- C taking place in
an adiabatic steady state CSTR. The expressions for CA and CB ob-
tained from solution of mass balances for A and B are

C CA C kC A (14)
A (1+Tkl)' B (1+k2)

The energy balance has the form (specific heats of all species being
considered equal, CA = CpB = CpC = Cp)

CAOCP(T- To)+(kiCAAH +k2CBAH2)T=0,

ki= kjoexR-R i=1, 2 (15)

Upon substituting Eq. (14) into the above, the master equation for
the adiabatic reactor is obtained as

G(T)=Rm(T), Rm(T)=Cp(T-To),

G(T Tk,(T) Ak2(T) 1
G(T)= -+kT] AHI+AH,2+ (16)
[l+Tk(T)] (l+Tk2,(T))


with the reactor temperature T being the only un-
known. The values of various parameters are:
Cp=300 J/{mol.K}, AH, = -55,000 J/mol, AH2 =
-71,500 J/mol, CAO=0.3 mol/L, T0=300 K, T= 0.01
min, E, = 9,900 cal/mol, E2 = 27,000 cal/mol, k, =
3.03 min-1 at 300 K, k2=4.58 min-1 at 500 K. With the
exception of C all other parameter values have been
taken from Fogler,[1o] where C has been considered
to be 200 J/{mol.K} and the reactor operation has
been considered to be nonadiabatic. The Mathcad
worksheet for this illustration is shown in Table
8. By plotting G(T) and R(T) versus T, the stu-
dents observe that the two profiles intersect at
five T's for T > T0, implying existence of five
steady states. The temperature at each steady
state can be calculated via iterative solution of
Eq. (16). Alternately, the relative error associ-
ated with Eq. (16), denoted as dif(T) in Table 8,
can be calculated at various temperatures to di-
rectly zoom in on the steady state temperature.
For the parameters under consideration, the
steady state T values are 309.59, 354.33, 473.85,
540.29, and 719.58 K.


TABLE 8
Worksheet for Illustration 8


Cp:=300 AHi:=-55000 AH2:=-71500 CAo:=0.3 To:=300 T:=0.01
El:=9900 E2:=27000


ko:= 3.3.exp l 1.
k : 1.987.ex 300)

kj(T):=kjO.exp fE


E, 1
k2o:=4.58-exp( 2. 1
k2:= .e 1.987 500

k2(T):=k2O .exp E2


-T.k,(T) T- k2(T)
(T) 1+(T-k(T)) (1+T.k2(T))


R(T):=Cp.(T- T)

1.5-105



G(T) 1105 -

R(Ts)
5 104



0


G(T)
dif(T):= (T) 1.0
R (T 1


dif(309.59)= 6.249x 10-5

dif(354.33)= 2.788x10-5

dif(473.85)= -4.697 x10

dif(540.29)= -5.812x10-

dif(719.58)=-1.027x10-


400 600 800


2 Chemical Engineering Education


6

6

6










DISCUSSION
The students also use Matlab in parallel to Mathcad. Both
packages are available on computers across the IIT campus
and in the chemical engineering computer laboratory. The
purpose of exposing students to different packages is to pro-
vide them with a broad spectrum of skills needed for solving
engineering problems and to demonstrate the differences in
the packages' capabilities for solving different engineering
problems.'11 The students recognize that some of the prob-
lems can be formulated, but not solved, by hand. They can
quickly develop worksheets for these problems and solve
them, the emphasis thus being on understanding the funda-
mentals of the problems.
Care must be taken to ensure that use of computational
software enhances students' understanding and enriches their
logic and problem-solving skills, rather than simply allow-
ing them to solve problems with only a superficial under-
standing of the problems.[13] With this in mind, the under-
graduate chemical reaction engineering course using this soft-
ware at IIT includes handouts and tutorials providing an intro-
duction to the software and to different numerical methods.
Further, the author has integrated computational software
throughout the course, with the use of software always follow-
ing solution of related simpler problems by hand.[131

CONCLUSION
The use of computational packages enhances teaching and
learning, allowing the teacher to cover more material.[2,14] In
the process, the students learn more and faster and appreciate
the course even more, while developing the skills and flex-
ibility necessary for ready adoption of different software pack-
ages for professional activities in industry.[1,4,10,14] The graph-
ics capabilities of Mathcad help in quick visualization of re-
sults as well as in reinforcing expected results and under-


standing not-so-expected results. The capabilities of Mathcad
in symbolic manipulations are of considerable use in develop-
ing analytical skills of students in solving complex problems.
The time spent outside the courses on gaining further familiar-
ity with different computational software and their applications
will allow students to reap the benefits of these programs.[14]

REFERENCES
1. Al-Dahhan, M.H., "Computing in the Undergraduate ChE Curricu-
lum," Chem. Eng. Ed., 29(3), 198 (1995)
2. Abbas, A., and N. Al-Bastaki, "The Use of Software Tools for ChE
Education: Students' Evaluations,"( 36(3), 236 (2002)
3. Davis, R.A., and O.C. Sandall, "A Simple Analysis for Gas Separa-
tion Membrane Experiments," ( Ed., 37(1), 74 (2003)
4. Sandler, S.I., "Spreadsheets forThermodynamics Instruction: Another
Point of View," Chem. Eng. Ed., 31(1), 18 (1997)
5. Aluko, M.E., and K.N. Ekechukwu, "Introducing Process Control Con-
cepts to Senior Students Using Numerical Simulation," Chem. Eng.
Ed., 33(4), 310 (1999)
6. Chen, W.-I., "Rate Measurement with a Laboratory-Scale Tubular Re-
actor," Chem. Eng. Ed., 33(3), 238 (1999)
7. Dickson, J.L., J.A. Hart IV, and W.-I. Chen, "Construction and Visual-
ization of VLE Envelopes in Mathcad," Chem. Eng. Ed., 37(1), 20
(2003)
8. Smith, W.R., M.Lisal, and R.W. Missen, "The Pitzer-Lee-Kesler-Teja
(PLKT) Strategy and its Implementation by Meta-Computing Soft-
ware,"( Ed., 35(1), 68 (2001)
9. Brauner, N., M. Shacham, and M.B. Cutlip, "Computational Results
How Reliable Are They? A Systematic Approach to Model Valida-
tion," Chem. Eng. Ed., 30(1), 20 (1996)
10. Fogler, H.S., Elements of Chemical Reaction Engineering, 3rd Ed.,
Prentice Hall PTR, Upper Saddle River, N.J. (1999)
11. Cutlip, M.B., and M. Shacham, Problem Solving in Chemical Engi-
neering with Numerical Methods, Prentice Hall PTR, Upper Saddle
River, NJ (1999)
12. Froment, G.F, and K.B. Bischoff, Chemical ReactorAnalysis andDe-
s ...I Ed., Wiley, New York (1990)
13. Dahm, K.D., R.P. Hesketh, and M.J. Savelski, "Is Process Simulation
Used Effectively in ChE Courses?" ( 36(3), 192 (2002)
14. Mackenzie, J.G., and M. Allen, "Mathematical Power Tools: Maple,
Mathematica, Matlab, and Excel,"( 2. .' 2, 156(1998)


CALL


* FOR


* PAPERS


for the
Fall 2006 Graduate Education Issue of

Chemical Engineering Education

We invite articles on graduate education and research for our
Fall 2006 issue. If you are interested in contributing, please send us your name,
the subject of the contribution, and the tentative date of submission.

Deadline for manuscript submission is April 1. 2006.

Respond to: cee@che.ufl.edu

Winter 2006 2











^]W laboratory


EXPERIMENTAL AIR-PRESSURE

TANK SYSTEMS

for Process Control Education


CHRISTOPHER E. LONG, CHARLES E. HOLLAND, AND
University of South Carolina Columbia, SC 29208
Process control education is a significant aspect of the
chemical engineering curriculum, as it provides a fun-
damental basis for modern chemical process opera-
tion. The subject is highly applied yet rooted deeply in theory.
Bridging the gap between the theory and application is often
a difficult task, particularly in the classroom setting. Experi-
mental laboratories have been shown to be useful in motivat-
ing students and reinforcing the information taught in the
classroom, '-4] often with the additional benefit of small-group
learning.5'61 The use of hands-on experimental laboratories
that are closely tied to the traditional process control lecture
course allows students to actually link the theoretical content
of the courses to its use on real-world systems. For this rea-
son, process control experiments have been developed across
the country.[7-19
The development of useful, dynamic, process control ex-
periments requires a number of considerations. Safety is the
primary consideration because an environmentally friendly
system that can be operated with minimal risk to both the
equipment and the user is necessary. The ideal system would
also be a cost-effective means to demonstrate the pertinent
material with some industrial relevance. It should be of mod-
erate complexity, as simple systems may be too trivial to
motivate students while a full-scale industrial process may
be too overwhelming. Giving it flexible configuration op-
tions will allow for its use in a variety of contexts. Reason-
able process time constants are also essential so that the sys-


EDWARD P. GATZKE

tem dynamics are slow enough to demonstrate that process
changes are not instantaneous, while also reacting quickly
enough to limit student boredom when examining dynamic
process transitions.
Undergraduate students typically have very limited expe-
rience with dynamic systems since many undergraduate
courses work under assumptions of steady-state operation. The
use of the dynamic experiments) provides this experience and
demonstrates all aspects of the textbook theory.[1-171
There are a number of well-designed, low-cost experiments
available commercially, from vendors such as Lego, for use


Christopher E. Long is currently a Ph.D. candidate in the Department of
Chemical Engineering at the University of South Carolina. His research
interests lie in the field of process systems engineering, focusing specifi-
cally on the applications of nonconvex optimization to process control and
identification. He holds a B.S. (2001) in chemical engineering from Clemson
University.
Charles E. Holland is the staff engineer for the Department of Chemical
Engineering at the University of South Carolina. He earned both his B.S.
(1997) and M.S. (2003) in chemical engineering at the University of South
Carolina. He designed and built the experimental systems described in
this article.
Edward P. Gatzke is currently an assistant professor in the Department
of Chemical Engineering at the University of South Carolina. His research
examines a variety of topics in process systems engineering, including
process identification andprocess control. He holds a B. ChE. (1995) from
the Georgia Institute of Technology and a Ph.D. (2000) from the Univer-
sity of Delaware.


Copyright ChE Division of ASEE 2006


Chemical Engineering Education










in process control education.E181 These systems, however, fail
to offer the flexibility to be utilized in many different con-
texts. Furthermore, they often fail to provide any semblance
of being industrially relevant.
At the University of South Carolina, both a simple, dy-
namic, nonlinear, two-tank, air-pressure system and a more
complex, multivariable, four-tank, air-pressure system have
been developed. These pressure-tank systems prove quite
useful in process control education, as they address the ob-
jectives for an ideal process control experiment. Inspired by
experimental liquid-level systems, '- ihc, c experiments are
exceptional instructional tools for chemical engineers. As op-
posed to liquid-level systems, in these systems pressure dif-
ferences drive the flow. This variation removes the limita-
tions in system flexibility typically associated with gravity-
driven liquid systems. The two-tank system is quite portable,
thus lending itself well to classroom and outreach demon-
strations. A variety of undergraduate topics including open-
loop modeling and traditional single-input, single-output
(SISO) closed-loop control strategies can be readily demon-
strated on the two-tank system. The more complex, multi-
variable, four-tank system can be used in a small group
setting to illustrate more advanced topics such as multi-
input, multi-output (MIMO) modeling, interacting sys-
tems, and multivariable decoupling, to name a few. This
paper presents a detailed description of both systems and
summarizes their current and future uses for both educa-
tional and research purposes.

THE TWO-TANK SYSTEM
A compact, experimental, air-pressure tank system involv-
ing a pair of tanks in series has been developed ( www.che.sc.edu/faculty/gatzke/software.htm>). A schematic
and photograph of the system are provided in Figures 1 and
2. This section describes the system itself as well as pre-
senting its uses in the context of undergraduate process
control education.
System Description
The two-tank pressure system is comprised of two con-
stant-volume aluminum tanks assembled in series supported
by aluminum framework (22 inches long X 24 inches high
X 17 inches wide). The two cylindrical tanks are each a foot
in length. Their diameters are two inches and one inch, re-
spectively. Supply air enters the system through a single one-
half-inch, air-actuated, BadgerMeter control valve.[241 The air
flows through quarter-inch tubing into the two tanks in series
and exits to the atmosphere. A small muffler is utilized at the
exit to reduce system noise. The tanks are separated by
Swagelok[25] metering valves with repeatable vernier handles.
This provides a means to accurately transform the system
between various system configurations. Note that completely
opening a valve between the two tanks effectively "joins"
the tanks, resulting in one large tank of uniform pressure, as
Winter 2006


opposed to two tanks in series. Pressure measurements are
available from each of the two pressure tanks. Gauges are
installed on each tank to provide visual indications of the
pressures while pressure transducers are used to more accu-
rately measure and transmit pressure readings to a computer.
The larger tank is also fitted with a small release valve that
vents to the atmosphere. This valve can be used to create a
disturbance on the system that might simulate a leak in the
given tank, providing the opportunity to examine disturbance
rejection as a possible control objective in addition to refer-
. .


Figure 1. Two-tank schematic.

C:;---- ------ ^ -----


Figure 2 Photograph of the twotank system
Figure 2. Photograph of the two-tank system.


Muffler





V2


Vd




















































Figure 3. Schematic of the four-tank pressure system.


High Pressure
Air Feed


\ Ia M_7_ ij '


SFigure 4. Pho h of te fr-tk p e s m.
Figure 4. Photograph of the four-tank pressure system.


Chemical Engineering Education


ence tracking. The apparatus is equipped with a National In-
struments Data Acquisition system which can be interfaced
to both Matlab/Simulink 261 and LabVIEW.[27] A complete ma-
terials listing can be obtained by contacting the authors.
It should be noted that initially the control valve exhibited
substantial hysteresis, making accurate modeling impossible.
A valve positioner was required in order to generate repro-
ducible open-loops results on the system. This also helps in-
troduce students to cascade control and the complexity of
real industrial systems.
In the lab environment, the feed air pressure can be sup-
plied in a more permanent manner from a compressor. On
the other hand, small compressed-gas cylinders or lecture
bottles can be used so that the system can be taken into the
classroom for demonstrations. Similarly, a dedicated desk-
top computer can be used in the labs, while a laptop can be
conveniently carried to the classroom.
Educational Uses
This new experimental system is quite valuable for educa-
tional purposes. In the classroom setting, it lends itself well
for demonstration to larger audiences. Alternatively, smaller
groups can experiment with the system in a laboratory set-
ting and reap the benefits of learning in a hands-on environ-
ment. The typical undergraduate class can be broken into small
groups that can be rotated between the actual pressure-tank
system and nearby computer labs. In the computer labs, stu-
dents can use a high-fidelity model of the system to carry out
simulation work that closely parallels what is to be done ex-
perimentally. This way, those entering the computer labs first
can prepare for the actual experiment, while those that see
the actual system first can later reaffirm what has been done
experimentally. These advantages are supported by the rapid
dynamics of the system. Note that the open-loop time con-
stant is on the order of 30 seconds. In an extended class pe-
riod, it is possible that numerous groups could get a substan-
tial amount of time working with the apparatus.










Using this system, many topics from the undergraduate pro-
cess control curriculum can be illustrated. Open-loop model-
ing can be performed to identify both first- and second-order
SISO models of the two tanks, depending on the configura-
tion. Both frequency- and time-domain models can be con-
sidered, including input/output descriptions such as
Autoregressive Moving Average (ARMA) models. Linear-
ization of an available nonlinear first-principles model can
also be carried out. Traditional closed-loop control method-
ologies such as Proportional-Integral-Derivative (PID) and
Internal Model Control (IMC) can be implemented. Ad-
ditionally, related topics such as closed-loop stability can
be demonstrated.

THE FOUR-TANK SYSTEM

This section describes the four-tank system in comparison
to the two-tank apparatus. A schematic and photograph of
the system are provided in Figures 3 and 4. This system's
uses for undergraduate, intermediate, and advanced process
control education are presented along with its utility in pro-
cess systems engineering research.


Figure 5. Flow diagram for alternative configurations of
the four-tank system.
Winter 2006


System Description
The MIMO experimental system consists of four intercon-
nected air tanks arranged in two parallel trains of two tanks,
in series, built upon a steel framework. Each tank is 35 inches
in length with diameters of 4 inches and 2.5 inches for the
upstream and downstream tanks, respectively. Supply air
flows into the system through two air-actuated BadgerMeter
control valves which serve as the manipulated variables for
the system. The air flows through copper tubing and the tanks
before exiting to the atmosphere. Again, mufflers have been
installed at the system exit to reduce the noise level. Specifi-
cally, the air flowing through control valve 1 (CV,) proceeds
into tank 1 and subsequently into tank 2 downstream before
exiting the system. Additionally, a portion of the flow from
the control valve can be routed into the downstream tank of
the adjacent train (tank 4). In a similar manner, control valve
2 (CV,) affects the pressure in tanks 3 and 4, with cross-flow
effects on tank 2. Valves V14 and V32 are directly responsible
for the cross-train flow. In some cases, the interacting nature
of the system as a result of the cross-train flow leads to the
presence of an adjustable, multivariable, right-half plane zero
and inverse response. Physically, this is a result of the fast
and direct response of the downstream tank pressures to
cross-train flow, in contrast to the slow indirect effects of
the flow from the large upstream tanks into the smaller
downstream tanks.
The flow of air through the system is driven by pressure
gradients. Check valves are not used, therefore air could flow
back upstream provided that the pressure gradient is in the
appropriate direction. (Similar liquid levels have limitations
in these regards as the flow path is dictated by gravity.) The
result is a more flexible, dynamic experiment. As with the
two-tank system, the various tanks are separated by a num-
ber of Swagelok metering valves; their placement allows the
system to be configured in a variety of ways. By opening or
closing select valves between the tanks, the system can be
quickly transformed from one configuration to another. The
possible configurations include: a single tank of numerous
possible sizes (depending on the number of tanks utilized),
two to four tanks in series, a pair of tanks in parallel, and
other setups that would have tanks in both series and paral-
lel. For example, V14, V22, and CV2 can be completely closed,
resulting in an SISO fourth-order system with air flowing
through all tanks in series (see Figure 5b). Note that in the
interest of saving laboratory space, the system is "f. 1l IcL I so
that the smaller tanks are placed above the larger ones, leav-
ing a system with total dimensions of 72 inches long, 22 inches
high, and 22 inches wide.
Educational Uses
Although not portable enough to be taken to the classroom,
this system is well suited for use in the laboratory environ-
ment. This apparatus can again be used for large group dem-
onstrations or in a more personal setting for individual-to-
27











small-group work (see Figure 6).
The multivariable, four-tank, pressure system can be con-
figured in such a manner that it closely mimics the operation
of the simple two-tank system, thus allowing one to demon-
strate similar concepts. The additional complexity and flex-
ibility of the four-tank system, however, also allow for its
use in a wider variety of contexts, particularly with regard to
its multivariable nature. The system can be configured such
that one control valve acts as a measured disturbance into the
downstream tank-thus allowing for feedforward control.
This configuration is shown in Figure 5a. Input/output mod-
eling of multiple tanks in series can be carried out given the
appropriate configuration, but MIMO modeling techniques
such as continuous and discrete-time, linear-time-invariant
(LTI), state-space approaches can also be applied. Interact-
ing systems can be demonstrated as well as dynamic decou-
pling. The simulink interface showing PI control of the four-
tank system is shown in Figure 7. In this feedback arrange-
ment, the two downstream tank pressures are being controlled
by manipulating the two control valves at the inlet. The dis-


Figure 6. Students performing lab on the tank system.


Figure 7.
Simulink inter-
face showing
closed-loop
control of the
four-tank system.


turbance rejection capabilities of this control scheme can be
shown by simulating a leak in either of the upstream tanks or
by changing the supply air pressure.
In addition to aiding in the presentation and reinforcement
of the undergraduate material, more advanced undergraduate
and graduate topics can be covered using this system. Linear
and nonlinear state and parameter estimation routines can be
developed for the system. Advanced control schemes can be
used including multivariable IMC, Hc, and linear Model Pre-
dictive Control (MPC). With some tank configurations, the
system can exhibit a multivariable right-half plane zero thus
inverse response-motivating the examination of input di-
rectionality and control performance limitations.[16
Student Assignments
For illustrative purposes, two relevant assignments typi-
cally given to students in the undergraduate and advanced
(intermediate and graduate-level) courses are provided.
Undergraduate Assignment

Configure thefour-tank system into an SISO arrange-
ment that involves two tanks in series. Develop a
transfer function representation relationship
between the control valve and the pressure
downstream tank. Using this model, implement an
Internal Model Control scheme on the system in
MatlablSimulink and test the closed-loop performance
system by introducing both setpoint changes
and disturbances.

Advanced Assignment

Configure the four-tank system into a 2-by-2 MIMO
arrangement that involves two parallel trains of two
tanks in series with cross flow. Consider the two
downstream tanks as process outputs and the two
control valves as the manipulated variables. Use


File Edit View Smulaton Format Tools Help
D | H a ; > .- i | 1 ua -Y %




YI+_p2 c


Int egraor 1 a Dlipla




1,7ub 50Em
1 1 aul2



Ready 100% ode45

Chemical Engineering Education










subspace identification methods in Matlab to develop
a linear state-space representation system.
Using this model, implement a traditional Model
Predictive Controller on the system and test the
closed-loop performance system by introducing
both setpoint changes and disturbances. Test the
impact various tuning parameters on the
stability and performance controller.
These assignments exemplify those used in the different
control courses. They provide students
with the opportunity to explore the mod-
eling and control the experimental pres-
sure-tank system. Again, note that in the
interest of time, some students can develop
their control en i'll il1. dll .. using a high-
fidelity process model as the system to be
controlled before implementing their work
on the actual system.
Related Research
In addition to its utility in the instruc- Figure 8. Sch
tion of process control theory, this four- relationship
tank system has potential for use in re- volved in calcul
across a val
search in the field of systems engineer- across a val
ing. To date, this particular system has Ref
been the focus of a number of research
endeavors.
For instance, system modeling is an important precursor to
many advanced model-based control schemes. In limited re-
gions of operation a simple linear model could suffice. Pro-
cess nonlinearities, however, often require more complex
model forms. The nature of this system is such that the pro-
cess can exhibit hybrid dynamic behavior as the flow of air
through the valves of the system can discretely switch be-
tween distinct, multiple, continuous regimes of operation. Un-
der low pressure-drop conditions, the air flowrate across a
given valve is dependent on both the up- and downstream
pressures. In high pressure-drop conditions, however, a sonic,
or choke, flow regime is encountered in which the flowrate
across a valve becomes solely dependent on the upstream
pressure. The respective valve manufacturers, Sw .,dc "k'1
and BadgerMeter,[241 provide "hybrid" flow expressions based
on first principles to capture these dynamics. For the
BadgerMeter control valves the flow can be described by:

q=NC, -P if Pb)0.5Pa
G gTa

or

(3/2)P2
q=NCv(3 if Pb<0.5Pa
S Ta

while for the Swagelok needle valve the flows can be de-


scribed by:


(2AP ____. ibaAP
q= NC Pl- PaG if Pb) 0.5Pa
3Va PaG STg,


emat
of th
ating
ve. A
renc


1
q=0.471NCvPa if Pb<0.5Pa (2)
GgTa

where q is a volumetric air flowrate
across the valve at standard conditions,
N is a numerical constant for units, C
is the valve coefficient, P is the up-
stream pressure, G is the specific grav-
ity of the fluid, and T is the tempera-
ture of the system. Temperature mea-
surements are not available at the vari-
ous points in the system. For conve-
nience it is assumed that the tempera-
ic showing the ture of the air in the system is approxi-
e pressures in- mately constant throughout. The first
Sthe gas flowrate flow expression defines the low pres-
dapted from sure drop regime where the flow across
e 25. the valve is a function of both the up-
stream and downstream pressures. The
second flow expression defines the
choked flow regime where the downstream pressure has no
influence on the flowrate. Under ideal conditions, these flow
expressions can be used in conjunction with the ideal gas law
to develop discrete-time models of the pressure in each tank.
To model the rate of change of pressure in a given tank (Pi),
the ideal gas law is assumed as the system is operated at both
a reasonable temperature and pressure.

i i=RT (3)
Vi

where V1 is the volume of the tank, hi is the molar rate of
change of air in the tank, R is the gas constant, and T is the
temperature inside the tank.
Provided that flow expressions define a volumetric flow
across a valve at standard conditions, the ideal gas law
can be utilized a second time to convert to a molar flow
across a valve.

=( RTP std
6= q (4)

where Pa,, is the standard (atmospheric) pressure, Ts is the
standard temperature, and again q is a volumetric flowrate.
Thus


PViatm T


Winter 2006


q










Based on this general expression, a discrete-time model of
the system can be developed. Using the switching conditions
prescribed by the valve manufacturers, a least squares regres-
sion can be performed to identify model coefficients that rep-
resent parameters such as the valve coefficients, temperature
influences, etc. For the simple case of modeling the pressure
within a single tank, the results are presented in Figure 9. It
can be seen that the hybrid model that considers both low-
pressure drop and choke flow regimes is better able to cap-
ture the system dynamics than a model based solely on low-
pressure drop flow.

Alternatively, mixed integer methods128-301 can be used to de-
velop strictly empirical hybrid descriptions of the process. Propo-
sitional logic can be used to formulate Mixed Integer Linear
Programs (MILP) whose solution yields optimal coefficients
and switching conditions for a variety of model forms includ-
ing hybrid Volterra, autoregressive moving average (ARMA),
and more general nonlinear state-space representations.

I _


Figure 9.
Comparison of a
fundamental
low-pressure-drop
flow model and a
hybrid dynamic
model in their
ability to describe
the pressure in
a downstream
tank.


On a similar note, six process states can be considered in
the modeling of the dynamics of the system. The pressures in
each of the four tanks can act as states in the model, as well
as two states that are not so obvious. The placement of the
two supplemental valves leading into the two larger tanks
causes some resistance to air flow, regardless of their posi-
tion. This, in effect, makes the small sections of entrance tub-
ing between the control valves and the supplemental valves
act as two additional but very small tanks. The pressure in
these two regions will act as the remaining process states. No
pressure measurements are available in the areas, yet the size
of these "tanks" and the nature of the system imply that the
associated dynamics are extremely fast. A set of ordinary dif-
ferential equations (ODEs) can be developed for the tank
system to describe each respective state. Under the assump-
tions that these two extra tanks exhibit fast dynamics in com-
parison to the rest of the system, however, an approximation
can be made that reduces the respective ODEs to algebraic
relationships as the derivative term can be approximated as


- 0.. . model
20 data
00
_j
15
2000 2500 3000 3500 4000


30
-c
0 u) 25


- 20 .. model
,o data

15
2000 2500 3000 3500 4000


80


>G(
S0L40
4-(
0 20

0
2000 2500 3000 3500 4000
Time (sec)

Chemical Engineering Education











zero. This leads to the use of a system of differential alge-
braic equations (DAE) to describe the system, as well as
motivating studies in the area.

Additionally, the system has been utilized as a testbed for
the development of advanced control strategies. In one case,
the prioritized objective inferential control of unmeasured
process states is considered. The system is operated in a 2-
by-2 fashion with measurements of the downstream tank pres-
sures available. The two upstream tank pressures are consid-
ered as the unmeasured process states to be controlled. Tra-
ditional MPC methods are often limited to the control of
measured outputs and typically rely on a heuristic tuning to
address the trade-off between satisfying different control ob-
jectives. A state-space modeling approach can be utilized to
explicitly describe unmeasured process states. Using infor-
mation from this state-explicit model and using propositional
logic, a mixed-integer MPC algorithm[311 can be developed
that relies on the online solution of an MILP or MIQP for the
optimal control move. Such a formulation can allow for a
more intuitive tuning in which control objectives, possibly
involving unmeasured states, are met in order of their as-
signed priority.

CONCLUSIONS
Chemical process control education is often limited by the
availability of practical hands-on educational tools. Few in-
dustrially relevant systems are available that offer both rea-
sonable size and cost while providing interesting dynamics
with the flexibility to be used in numerous contexts. This paper
describes two such systems that provide students with the
opportunity to actually apply and demonstrate experimen-
tally many of the theoretical concepts that are fundamental
to the subject. A small, experimental, two-tank system has
been developed for use as a tool in process control educa-
tion. The size and simplicity of the system lend themselves
well to particular use in the undergraduate classroom. A simi-
lar yet more complex multivariable four-tank has also been
developed. Its flexibility enables its use in a variety of appli-
cations. Many aspects of both the undergraduate and gradu-
ate-level process control curriculum can be presented. Addi-
tionally, the system is the focus of a variety of interesting
research problems. Among these are studies on the hybrid
dynamic nature of the flow through the system, and the sys-
tems' use as a testbed for advanced control schemes such as
prioritized objective MPC.

ACKNOWLEDGMENT
The authors would like to acknowledge financial support
from the National Science Foundation Early Career Devel-
opment grant CTS-0238663.

REFERENCES
1. Doyle, F.J., III, E.P. Gatzke, and R.S. Parker, Process Control Mod-
ules-A Software Laboratoryfor Control Design, Prentice Hall (1999)
2. Doyle, EJ., III, E.P. Gatzke, and R.S. Parker, "Practical Case Studies
Winter 2006


for Undergraduate Process Dynamics and Control Using the Process
Control Modules, Comp. App. in Eng. Edu., 6(3),181 (1998)
3. Jung, J.H., M. Lee, J. Lee, and C. Han, "A Development of Experi-
mental Education Program: Computer Control of Multi-Stage Level
Control System," Comp. ( 24(2), 1497 (2000)
4. Marlin, T.E., "The Software Laboratory for Undergraduate Process
Control Education," Comp. Chem. Eng., 20, S1371 (2000)
5. Millis, B.J., and PG. Cottel, Cooperative Learningfor Higher Educa-
tion Faculty, Oryx Press, Phoenix (1998)
6. Johnson, D.W., R.T. Johnson, and K.A. Smith, Active Learning: Co-
operation in the Classroom, Interaction Book Co., Edina, MN,
(1998)
7. Gatzke, E.P, R. Vadigepalli, E.S. Meadows, and F.J. Doyle III, "Ex-
periences with an Experimental Project in a Graduate Control Course,"
Ed., 33(4), 270 (1999)
8. Joseph, B., C. Ying, and D. Srinivasagupta, "A Laboratory to Supple-
ment Courses in Process Control," Chem. Eng. Ed., 36(1), 20 (2002)
9. Ang, S., and R.D. Braatz, I i ......... I !'r. .i ,cts forthe Process Con-
trol Laboratory," ( Ed., 36(3),182 (2002)
10. Ogunnaike, B.A., and W.H. Ray, Process Dynamics, Modeling, and
Control, Oxford University Press (1994)
11. Riggs, J.B.,ChemicalProcess Control, FerretPt .I.- I.... .'... I.I I.' "' I
12. Seborg, D.E., T.F Edgar, and D.A. Mellichamp, Process Dynamics
and Control, John Wiley and Sons (1989)
13. Astrom, K.J., and B. Wittenmark, Computer-Controlled Systems:
and Design, Prentice Hall, Inc., 3rd Ed. (1997)
14. Bequette, B.W., Process Control: Modeling, Design, and Simulation,
Prentice Hall (2003)
15. Marlin, T.E., Process Control: Designing Processes and Control Sys-
tems for Dynamic Performance, McGraw Hill, 2nd Ed. (2000)
16. Skogestad, S., and I. Postlewaite, Multivariable Feedback Control
Analysis andDesign, John Wiley and Sons, New York, 1st Ed. (1996)
17. Stephanopoulus, G., Chemical Process Control: An Introduction to
and Practice, Prentice Hall (1984)
18. Moor, S., P Piergiovanni, and D. Keyser, "Design-Build-Test: Flex-
ible Process Control Kits for the Classroom," in Proceedings of the
ASEE Annual Conference, Nashville, TN (2003)
19. Johansson, K.H., and J.L.R. Nunes, "A Multivariable Laboratory Pro-
cess with an Adjustable Zero," Proc. American Control Conf 2045
2049, Philadelphia (1998)
20. Johansson, K.H., and A. Rantzer, "Multi-Loop Control of Minimum
Phase Systems," Proc. American Control Conf., 3385-3389, Albuquer-
que, NM (1997)
21. Vadigepalli, R., E.P Gatzke, and F.J. Doyle III, "Robust H-infinity
Control of an Experimental 4-Tank System," Ind. Eng. Chem. Res.,
40(8), 1916(2001)
22. Dai, L., and K.J. Astrom, "Dynamic Matrix Control of a Quadruple
Tank Process," Proceedings ofthe 14thIFAC, 295-300, Beijing (1999)
23. Gatzke, E.P, and F.J. Doyle III, "Use of Multiple Models and Qualita-
tive Constraints for Online Moving-Horizon Disturbance Estimation
and Fault Diagnosis," J. Proc. Cont., 12(2), 339 (2002)
24. BadgerMeter, Inc.: Industrial Division, Tulsa, OK, Research Control
Valves: Installation, Operation, and Maintenance Procedures.
25. Swagelok Company, Swagelok: Valve : .,. (2000)
26. The MathWorks, Matlab 6.5, Prentice Hall (2002)
27. LabVIEW 6.1, National Instruments Corporation (2003)
28. Roll, J., A. Bemporad, and L. Ljung, "Identification of Piecewise Af-
fine Systems via Mixed-Integer Programming," Automatica, 40, 37
(2004)
29. Bemporad, A., and M. Morari, "Verification of Hybrid Systems via
Mathematical Programming," Lecture Notes in Computer Science,
1569,31 (1999)
30. Frerrari-Trecate, G., M. Muselli, D. Liberati, and M. Morari, "A Clus-
tering Technique for the Identification of Piecewise Affine Systems,"
Automatica, 39, 205 (2003)
31. Long, C.E., and E.P Gatzke, "A Mixed Integer Model Predictive Con-
trol Algorithm for the Prioritized Objective Inferential Control of Un-
measured States," Ind. and Eng. Chem. Res., 44(10), 3575 (2005) 1
31











[fj learning in industry


This column provides examples of cases in which students have gained knowledge, insight, and
experience in the practice of chemical engineering while in an industrial setting. Summer internships
and co-op assignments typify such experiences; however, reports of more unusual cases are also
welcome. Description of the analytical tools used and the skills developed during the project should
be emphasized. These examples should stimulate innovative approaches to bring real-world tools
and experiences back to campus for integration into the curriculum. Please submit manuscripts to
Professor W.J. Koros, Chemical Engineering Department, University of Texas, Austin, TX 78712.




PARTNERING WITH INDUSTRY

for a Meaningful Course Project


RHONDA LEE-DESAUTELS
University of Kentucky at Paducah Paducah, KY
MARY BETH HUDSON
Wacker Specialties Calvert City, KY
RALPH S. YOUNG
Air Products and Chemicals, Inc. Calvert City, KY
Engineering students can gain valuable benefits from
an industry-sponsored project. Not only do students
gain exposure to a full-scale chemical process, they
also work closely with process engineers to collect and evalu-
ate data. Students may even be allowed to collect data them-
selves by running product-testing equipment on-site. Once
the data are analyzed, students can present their findings in a
formal environment in front of industry personnel.
Many chemical engineering programs provide opportuni-
ties for students to tour regional industries, thus exposing them
to the complexities of a full-scale chemical process. Rarely,
however, are students given the chance to do coursework on
a real problem with an actual state-of-the-art industry pro-
cess. Yet such experience is especially valuable to students
who do not receive a co-op or internship opportunity.
The University of Kentucky at Paducah has an advanta-
geous location in close proximity to many industries. Calvert
City, 17 miles east of Paducah, is home to 16 multinational
industrial plants including Arkema Chemicals (formerly
Atofina Chemicals), ISP Chemicals, Degussa Corporation,
Celanese Chemicals, Westlake Vinyl Corporation, Wacker
Polymer Systems, and Air Products and Chemicals. Many of
these industries were involved in establishing the UK-Paducah
engineering program, and now participate on an Industrial
Advisory Board (IAB) that provides input into course content.


Through the IAB, contact was made with one member in-
terested in collaborating on a course project. Wacker Poly-
mer Systems, whose manufacturing site is on the Air Prod-
ucts plant site, provided the opportunity for an industry project
applicable to Introduction to Particle T h. lii. d. N\, a course
offered biannually to upper-level undergraduates. Air Prod-
ucts is a minority partner in a joint venture with Wacker Poly-
mer Systems on the operation of a spray-dryer system. The
system manufactures a powder used in dry-mix mortars and
other construction-related products.

Rhonda Lee-Desautels is an assistant pro-
fessor of chemical and materials engineering
at the University of Kentucky at Paducah. She
received her Ph.D. in 1994 from The Ohio State
SUniversity, under the direction of L.-S. Fan.
Before taking a position in academia, she was
employed by International Paper for seven
years. Her research areas include particle-par-
ticle interactions, gas-solid fluidization, and
advanced battery materials.

Mary Beth Hudson is the site manager of
Wacker Polymer Systems in Calvert City, Ky.
She received a B.S. in chemical engineering
from the University of Kentucky in 1989. She
began her career as a process engineer forAir
Products and Chemicals in 1989 and joined
Wacker Polymer Systems in her present role
in 1998.



at the Air Products and Chemicals plant in
Calvert City, Ky. He received a B.S. in chemi-
cal engineering from Cornell University in 1971
and an M.B.A. from State University of New
York (SUNY) at Buffalo in 1981. In 1991 he
received a master's in environmental tech-
nology from Murray State University in
Murray, Ky.

Chemical Ergireering Educatior


Copyright ChE Division ofASEE 2006










Three projects were identified that: were of interest to
Wacker; involved the spray-dryer system; and applied to the
course content. One important project-selection criterion was
that students would have the opportunity to perform particle-
sizing analyses using the company's Beckman Coulter Counter
laser diffraction analyzer. Therefore, each student would be in-
volved in data collection on a real project, and would gain ex-
perience running a particle-sizing instrument.
This industry project, taking the place of the usual term
paper assignment, counted as 20% of the final grade. The
requirements of the industry project were: to tour the process
site; obtain all available data from sponsors; collect additional
data; compile and analyze the data; formulate conclusions
and recommendations; write the report; and present to spon-
sors. One of the first steps was separating the 10 students
enrolled in the course-all undergraduate seniors-into
one of the three projects.


Figure 1. The Wacker spray-dryer system
in Calvert City, Ky.


The industry projects were introduced during the fourth
week of class, after students had been exposed to particle-
size analysis, mixing and segregation of particles, and sepa-
ration of particles from a gas-subjects related to the three
chosen projects. Given a form containing a short description
of the projects, the students were asked to rank their interest
in each. All students were then assigned to their first or second
project choice. One project group had four students and the
other two groups each had three.
The industry tour of the spray-dryer process site (See Fig-
ure 1) took place during the fifth week of the course. The
regular class meeting time was at 2 p.m. on Tuesdays/Thurs-
days for 75 minutes each. Arrangements were made to carpool
on a Thursday to the Air Products plant site, leaving at the
beginning of regular class time, and returning before 5 p.m.
(one student had a 5 p.m. class). This three-hour time span
allowed for 20 minutes travel to plant site, 30 minutes for
introductions and a safety/orientation video, a one-hour plant
tour, a 30-minute break-out session with engineers to discuss
specific projects, and 20 minutes return travel. On the day of
the tour, students were instructed to wear long pants, no open-
toe shoes, and no sleeveless shirts. Our industry contacts pro-
vided flame-retardant smocks, hard hats, and safety glasses
for the students at the plant site. After the tour, groups were
responsible for making arrangements with a Wacker engi-
neer for any experiments or analyses required by the projects.

THE INDUSTRY PROJECTS
Figure 2 shows a schematic of the Wacker spray-dryer pro-
cess indicating the locations of the three projects.M11 In this
process, the facility produces vinyl acetate-ethylene copoly-
mer redispersible powders.[21 The conglomerated polymer
powder that forms during the process is redispersed when
contacted with water. These powders are used to improve







eBaune Figure 2.
Spray-dryer
process
flow
[ I70* diagram.


Winter 2006


PROACT 41


PROJECT -3










adhesion, impact resistance, flexible strength, water and
freeze-thaw resistance, and abrasion resistance properties of
Portland cement and other architectural coatings. In the pro-
cess, polyvinyl alcohol (PVOH) is mixed with emulsions and
fed to the spray dryer.
High-pressure air and the solution are supplied to the top
of the tower through spray nozzles. In the tower, water is
driven from the mix leaving a dry powder at the bottom of
the tower. The dried powder is pneumatically transported from
the spray dryer to the main baghouse, where particles are
separated from the gas before being transported to the prod-
uct baghouse; there particles are screened and then stored in
a silo. From the silo, the product powder is packaged and
warehoused until delivery to the end user.
Project 1. Nozzle Configuration versus Particle-Size
Distribution (PSD) of Spray Dryer Product
In the spray-dryer tower, polymer is supplied to the top of
the tower through a high-pressure ring of spray nozzles. The
high pressure forces the liquid droplets through a small orifice,
causing them to atomize into a fine spray. The first project in-
vestigates the effect of the nozzle configuration-that is, the
sequence of nozzles that are operational-to the final PSD of
the product. Students measured the PSDs based on three differ-
ent spray-nozzle
S configurations us-
ing the Beckman
Coulter Counter
(See Figure 3).
Students com-
pared the PSDs
and analyzed the
results based upon
differences in tra-
jectories between
Figure 3. Student Melissa Barrett and the various con-
Professor Lee-Desautels use the figurations.
Beckman Coulter Counter at the plant. The students






S[b)"I'









Figure 4. Types of agglomeration occurring through-
out the spray dryer.


found little variation between sample distributions for the
three nozzle configurations. Wacker provided an airflow
model of the spray dryer to aid the students in their analy-
sis.[3] The airflow model showed a vortex forming in the tower,
causing much turbulence. The students attributed the small
variation in PSDs to the presence of this highly turbulent
vortex region, which formed in the tower independently of
nozzle configuration. The students connected the project to
their coursework by proposing the various forms of agglom-
eration that can occur throughout the tower (See Figure 4)
with capillary (c) and droplet (d) occurring at the top of the
tower, nearer to the atomized liquid spray, and pendular (a)
and funicular (b) agglomeration dominating toward the bot-
tom of the column, where much of the liquid has evaporated.[4]
This student group recommended a study to maximize poly-
mer feed to the tower without causing excessive agglom-
eration by controlling nozzle configuration, nozzle pres-
sure, and airflow.
Project 2. Baghouse Segregation Analysis
Once the polymer powder has exited the spray tower, it has
an average diameter of about 100 microns. It is mixed with
clay particles (average size 60 microns) and pneumatically
transported down flexible ductwork to the main baghouse. The
main baghouse serves to separate the transport gas from the
powder while controlling particulate emissions. The pneumatic
ductwork splits into six separate ducts (labeled A, B, C, D, E,
and F as shown in Figure 5) before entering the main baghouse.
The second project involved analyzing the uniformity of
particle loading on the main baghouse after the splitting of
the ductwork. Samples were collected by industry personnel
at each of the six separate ducts leading into the baghouse.
The students analyzed the samples with the Beckman Coulter
Counter and compared distributions. The students found that
the mean particle size differed widely among the ducts. Duct
A contained the largest particles at a median size of 159 mi-
crons; Duct B particles had a median size of 76 microns; Ducts
E and F averaged 60 microns; and Ducts C and D averaged


From Splay Diyei

Figure 5. Pneumatic ductwork to main baghouse.
Chemical Engineering Education


D


A\B


A B


L


V%


/













LS Particle Size Analyzer


5 Nov 2003


UKsampl.$06
UK sample 3 top
6
separation UK sample 3top
Fraunhofer rff
Dry Powder Module
15.28 5Nov2003
18
17%
301


Group ID UK sample 3 top
Operator JVB


Run length 30 seconds
Auger 34
Firmware 202 0


File name.
Sample ID.
Run number
Comments
Optical model:
LS 100Q
Start time:
, il,
Software


5-

4-


Volume Statistics (Geometnc) UK sample $06
Calculations from 0 375 pm to 948 3 pm
Volume 100%
Mean 82 70 pm SD.: 3342
Median 103.1 pmn Variance: 11 17
Mean/Median Ratio 0 802 Skewness. -1,456 Left skewed
Mode 1276 pm Kurtosis 3 102 Leptokurtic
% > 10 25 50 75 90
pm 268 171 1031 5431 1993

Figure 6. PSD of particles sampled from top of container.


BuEC LS Particle Size Analyzer
COgFER. 5 Nov 2003

File name' UK sampl.$05 Group ID: UK sample 3 bottom
Sample ID: UK sample 3 bottom
Run number 5 Operator JVB
Comments: separation UKsample 3 bottom
Optical model Fraunhofer.rff
LS 100Q Dry Powder Module
Start fme. 15.25 5 Nov 2003 Run length: 30 seconds
Vibrator 18 Auger 34
Obscuration: 6%
Software: 301 Firmware 202 0

Differential Volume
UK sampl.$05

4


3-

S 2- Note lack of
Large particle
fraction
1-'1


04 1 2 4 6 10 20 40 100 200 400 1000
Particle Diameter (pm)


Volume Statistics (Geometric) UK sampl.$05
Calculations from 0 375 pm to 948.3 rm
Volume. 100%
Mean. 68.78 pm S D- 3424
Median: 91.07 pm Variance. 11.72
Mean/Median Ratio. 0755 Skewness: -1547 Left skewed
Mode: 116.3 pm Kurtosis: 2.647 Leptokurtic
% > 10 25 50 75 90
pmr 2332 1524 9107 4540 14.46

Figure 7. PSD of particles sampled from bottom of container.

Winter 2006


COULTER.


Differential Volume
UK sampl.$06






Note fraction of
Large particles



1 2 4 6 10 20 40 100 200 400 1000
Particle Diameter (pm)


45 microns. The students realized they couldn't explain these
results in terms of inertial considerations alone, as the larger
particles would be more likely to settle out when making the
turn to Ducts A, B, E, and F-an effect that would lead to
smaller particles in those ducts. The students decided they
needed to gather more information about the ducting. On
speaking with plant personnel, they were made aware that
the lines had never been cleaned. The students also learned
that directional plates had been installed in the transport lines
to direct powder flow, but were nonfunctional due to buildup
of wet product-essentially "gluing" them in place. Students
proposed in their analysis that blockage due to material
buildup was occurring in the pneumatic lines, and proposed
it was concentrated around Ducts C and D, creating a region
of restricted flow and high pressure drop. This restriction to
flow in turn resulted in the smaller average particle sizes in
these ducts, they theorized.

In addition to regular sampling of the transport lines to
monitor particle distributions, the students recommended the
directional plates in the ductwork be made operational to con-
trol powder fed to each duct. To prevent recurring problems,
students proposed that since the majority of this buildup oc-
curred during start-up of the process, developing stricter pro-
cess start-up guidelines was recommended.

Project 3. Product Segregation During Transport

Once the powder has been sent through both sets of
baghouses, it is transported to a silo where it is bagged and
transported to consumers by truck. The third project investi-
gated the segregation of powder product during the transport
process. Some additional PVOH powder is added to the spray
dryer product before reaching the product baghouse, and the
company suspected some segregation might be occurring with
handling and transport due to the PVOH having a smaller
average particle size than the product. Having learned about
the mechanisms of particle segregation,[41 students decided the
mechanism of percolation was responsible due to the rise of
coarse particles with agitation.

To test if segregation could occur, the students used a Ro-
Tap device to agitate a sample container for a given amount
of time to simulate the transport process. The students then
took samples from the top and bottom of the shaken sample
container and measured PSDs in the Beckman Coulter. The
students also had an unshaken control sample that was mea-
sured. They found that the control had little difference in par-
ticle-size distributions between the top and bottom samples,
with mean sizes of 95 and 96 microns, respectively. The
shaken samples showed a greater percentage of large par-
ticles in the top samples than in the bottom samples, indicat-
ing the percolation and coarse particle-rise phenomena. In
one shaken sample, after shaking for 30 seconds particles
removed from the top of the container had a mean size of 83
microns, while particles from the bottom had a mean size of
69 microns (See Figures 6 and 7).













"One of the most

valuable aspects of

this assignment

from an industry

perspective was

the 'Presentation

to Plant Technical

Professionals.'

Many entry-level

engineers do not

have the communi-

cation skills to

clearly share their

ideas with techni-

cal management.

In many cases,

engineering super-

visors spend sig-

nificant amounts

of time working

with entry-level

engineers on their

presentation and

communication

skills."


-Industry
feedback


The students concluded that particle segregation is a negative effect for a product
intended to meet certain requirements and specifications for its end use. Because this
product had received no complaints, however, the students recommended no changes
to the transportation of these powders. In spite of this concession, they further recom-
mended making customers aware that this phenomenon occurs as a courtesy in case end
users might want to homogenize the powder post-transport.

PROJECT PRESENTATIONS
At the end of the semester, each team presented its project findings to industry
personnel at a seminar held in the Air Products Engineering Building conference
room. Attending the proceedings were the three industry participants plus an addi-
tional invited engineer.
All students were required to participate in the presentation, and were given an
outline on the required presentation format:
A - round (Define the System and the Problem)
A Experimental (What You Did to Collect Data)
A Results/Analysis (Present the Data and Analysis)
A Discussion (Your Interpretation Results)
A Conclusions
A Recommendations
The students in each group took turns presenting portions of the findings and were
graded on the quality of the visual aides and delivery. The conference room was
equipped with state-of-the-art audiovisual equipment including a projector and
screen. The students were told to bring their presentations on a CD, with addi-
tional copies to hand out to industry attendees. Most students had never pre-
sented in this kind of corporate environment.

INDUSTRY PERSPECTIVE
In an effort to capture the industry viewpoint on the project experience, industry
participants were asked to submit comments on the project. Their comments are sum-
marized below. The comments are valuable, not only for students, but also for faculty
to gain insight into what qualities industry values from their engineering employees.
From the responses, it is obvious that the industry participants looked at the project
more as a way to prepare students for the workforce, offering words of advice and
critique, than a means of obtaining free labor. The industry participants had a genuine
desire to provide a distinctive learning experience for our engineering students.

THE COMMENTS
Concerning the Performance of the Students
A "From an industry perspective, Ifound the students enthusiastic and ready to
do a 'hands-on' project. I'm not sure if everyone was trying to build their
resume, but each student approached the project with an open mind and was
prepared to learn new. They quickly learned how to operate the test
equipment and collect useful data."
A "In most cases, once the 'newness' of running the Coulter Counter and other
test equipment wore the tedium of repetitive testing and analysis was
apparent. In this respect each student was exposed to real industrial experience:
10-25% new and exciting opportunities versus 75-90% less exciting work.
Every student has their own threshold of tedious, repetitive work. These types of
assignments provide the opportunity to help students decide career paths such
as process engineering in a plant environment or research assignments in lab
environments."
A "In this project, it was obvious each student had some prior presentation
Chemical Engineering Education











training and experience. Many engineering curricula
include this training in their degree requirements.
Project leaders divided the presentation so that it
flowed and used graphics to help the
audience understand the project and results."

A "The only element that was lacking in these presenta-
tions was the business case that would make or break
a decision to allocate more resources. Since this
aspect was not expected from the students, the
technical was able to question the students and
guide their *' during the presentations. When
the business case was made for a project-such as to
increase plant production yields or benefit custom-
ers many light bulbs seemed to go on in students'
minds about the importance work. The
interaction between students and industrial profes-
sionals was invaluable and one ' most important
aspects projects."

Concerning What is Valued in an Employee
A "One most valuable aspects ' assignment
from an industry perspective was the 'Presentation to
Plant TechnicalP Many entry-level
engineers do not have the communication skills to
clearly share their ideas with technical management.
In many cases, engineering supervisors spend
significant amounts of time working with entry-level
engineers on their presentation and communica-
tion skills."

A "Most new engineers get bogged down in project
details and sophisticated analysis, and cannot
summarize pros and cons to drive a management
decision."

A "Key qualities I value in employees are: problem
solving ability, creativity, communication, teamwork,
ability to accomplish goals with minimal direction,
initiative, dependability, time-management skills, and
the ability to manage multiple con-
straints. The students' analytical ability is proven by
their successful completion engineering
curriculum. This project allowed them to demonstrate
the other key qualities above as well."

A "Among the biggest constraints in industry are time
and personnel. We are expected to accomplish more
with less. i we need goal-oriented employ-
ees who can drive projects to completion. I have seen
many engineers spend too much time evaluating
options in trying to find the 'best' solution, only to
create more problems by not achieving I
was told as a young engineer that you will be seen as
more successful if you attempt to solve a problem five
times over a year and only succeed on the, try
than if you spend the whole year developing the
perfect solution for the first try."

A "We do not have clearly defined problems with one
correct answer in our work environment. Often, data
to analyze the problem are missing or incomplete.
Winter 2006


Resources such as money, personnel, and time are
limited. Engineers are to determine the
best solution to the problem based on the information
and resources at hand. There is always an economic
impact that has to be evaluated."

Concerning the Benefit to Industry
A "The results from the three projects reinforced our
knowledge and confidence in what was happening.

A "The data will be useful to support the allocation of
resources to cleaning the ducting to the main
baghouse, alleviate any concerns with nozzle
configuration final product quality, and
increase awareness of product segregation with
transport."

A "The particle-size data collected in these projects
have been used to address customer issues associated
with particle size. Examples are a recent modification
to a powder grade to decrease particle size increase
bulk density in response to a bulk issue with
one of our largest customers, and a recommendation
ofpowder grades to address an application which
will require a coarser particle size."

A "One main benefits to industry in participating
in these programs is that we get a better introduction
to the students who will be entering the job market."


STUDENT EVALUATIONS

The students were asked to evaluate the industry project in
the optional-items section of the evaluation form. Four que-
ries were made on the project. Students were also asked to
provide personal comments specifically about the industry
project. Eight of the 10 students taking the course were present
for the evaluation.
A Query 1. Rate your overall perception of the
industry project.
Response: One rated it outstanding, four rated it good,
two rated it average, and one rated it poor.

A Query 2. The industry project has allowed me to
learn more about a specific area of particle
technology.
Response: One rated it outstanding, four rated it good,
two rated it average, and one rated it poor.

A Query 3. The industry project has helped me feel
better prepared to seek employment with a
company that manufactures/uses particles.
Response: Two rated it outstanding, three rated it
good, two rated it average, one rated it poor.

A Query 4. The industry project was a valuable
component of the course.

Response: One rated it outstanding, four rated it good,
and three rated it as poor.

Continued on page 53
37











Random Thoughts...







THE WAY TO BET



The race is not always to the swift, nor the battle to the \i,, i,, but that's the way to bet.


RICHARD M. FIELDER
North Carolina State University Raleigh, NC 27695
here is no such thing as certainty in science-every-
thing we believe ultimately rests on unprovable as-
sumptions and imprecise observations. Our current
theory may seem to work beautifully, but if we really under-
stand science we know that new data can overthrow it at any
time. Nevertheless, if there's enough evidence to back it up,
we can base predictions on it and sleep peacefully without
worrying that we might be wrong. If I pick up a heavy object
and drop it, I feel comfortable predicting that it will fall down.
I can't prove Newton's theory of gravitational attraction and
I'm clueless about why gravity works the way it does (as was
Newton), but I'm confident that down is the way to bet.
As much uncertainty as there may be in science (and by
extension, engineering), there is far more in education. Stu-
dents are infinitely more complex and unpredictable than
cantilever beams and airborne projectiles and fruit flies. Even
in education, however, there are some propositions that give
you a great chance of coming out ahead if you bet on them
often enough. I've got a few like that to offer you.

STUDENTS
1 If a student who fails a test claims .'. : that
helshe really understood the material, then either
helshe really didn't understand it or the test was
S(too too tricky, . ). The first one hap-
pens far more often than most students believe and
the second far more than most professors believe.
0 Students who argue vehementlyfor additionalpoints


on every test
dents and
marriages.


have lives as both stu-
.' I also worry about their


r Students who routinely come up with bizarre but
valid ways of approaching problems may
in school but will do very well as researchers and
engineers (if they survive school).


-Damon Runyon



0 Students who drop out of engineering are on aver-
age no worse academically than students who stay
in. We like to believe that our absurdly high dropout
rates in engineering mean we are eliminating weak
students and retaining good ones, but that's not how
it goes. Lots of students who leave have fine aca-
demic records but just don't like what they see in
our classes. (Don't bet against this one-I've got
the data to back it up.)

GOOD AND BAD TEACHERS
An engineering faculty member is a good teacher (i.e., a
teacher who motivates his/her students to learn and facili-
tates their learning) or a bad teacher (i.e., a teacher who does
not motivate or facilitate learning and may even interfere with
it) if he or she:
D (good) gets all of his/her students actively involved
in class and knows of their names (or at least
most of them in large classes).
0 (bad) makes classes PowerPoint shows, or spends
most of every period deriving equations, or puts
high-level problems on exams that are qualitatively
S. anything students have seen in class
or on homework "to see if they can think for them-
selves."


Richard M. Felder is Hoechst Celanese Pro-
fessor Emeritus of chemical engineering at
North Carolina State University. He received his
\B.ChE. from City College of CUNYand his Ph.D.
from Princeton. He is coauthor of the text El-
ementary Principles of Chemical Processes
(Wiley, 2000) and codirector of the ASEE Na-
rfj m tional Effective Teaching Institute.


Copyright ChE Division ofASEE 2006


Chemical Engineering Education










D (good) always has students waiting in the dur-
ing hours and coming with questions before
and after class.
D (bad) uses words like" .' "and"
when about student-centered teaching meth-
ods (e.g., problem-based ' and .
who use those methods, and dismisses educa-
tional research as nonrigorous. I'm also betting that
individuals who do this have never read an educa-
tional research study and could not name a journal
that publishes them.
D (good) gets consistently excellent student ratings.
It's possible that the ratings are high because of
easy grading or whatever other spin colleagues
with lousy ratings put on it, but I'm betting (again
with a lot of research backup) that a highly rated
engineering instructor would also show up as a
good teacher in peer ratings and assessments of
learning outcomes.
(bad) gets consistently poor student ratings. Some-
one who is regularly shredded by most students may
claim it is because he is "rigorous" or "unwilling to
lower standards," or because she I'llic,' to be an
entertainer," and "the students don't appreciate me
now, but after they graduate they'll see how good I
was." Maybe, but if I always bet that those instruc-
tors are simply poor teachers I say I'll come out way
ahead in the long run.
D (good) has students coming back years later saying
what an outstanding teacher he or she was. I'd bet
my life savings on this one-and I'd do so even if
that individual has never gotten a grant or published
a research paper.

MISCELLANY
D Little or nothing . 'ul will be accomplished
at faculty committee meeting. The more frequently
the committee has regularly scheduled meetings, the
more I would be willing to bet on this one. Fur-
thermore, the larger the committee, the less it will
accomplish.
D New faculty members who get some formal training
Smentorship i be better teachers and more
4ul researchers after two years than their
counterparts who get the traditional amount of train-
ing and mentorship (none). More and more schools
are choosing to bet my way by giving their new hires


meaningful orientation and formal mentorship.
D Departments tha decide to give tenure and promo-
tion to, .'". ,". members who focus on teach-
ing and educational scholarship have stronger
teaching programs than they had before, and their
research productivity and quality will not
D A high school senior contemplating engineering I
get a better education by avoiding schools where
much of the administration and faculty think ABET
is the enemy.
D Textbooks with CD supplements will soon be re-
placed by interactive DVDs that may or may not
have text supplements, which lead to improved
learning. The present generation of faculty and stu-
dents may find the adjustment difficult, but the next
generation will have no trouble with it at all.
D Traditional campus-based departments willfind it
increasingly hard to compete with excellent distance
programs for good applicants. An online course that
includes user-friendly interactive tutorials, electronic
interactions between students and instructors and
among students, and individual conferencing with
the professor and tutors, provides a better educa-
tional experience than a campus-based course that
is mostly chalk and talk-and distance programs are
getting better at those things all the time.
More and more traditional engineering jobs i be
by computers, technicians, and engineers
in India and China (and Malaysia and Croatia
and ... ). Graduates of schools that continue to fo-
cus on traditional content will have a harder and
harder time finding and keeping jobs. Graduates of
schools that focus more on entrepreneurship, criti-
cal and creative thinking, multidisciplinary project
management, and global economics will do fine.
I have undoubtedly tipped over some sacred cows here.
Some of you will tell me that "Professor X dumbed his tests
down and started to get great student evaluations," or "Pro-
fessor Y's students bur her in effigy every year but as alumni
they create multimillion-dollar endowments in her name," or
"I can so name an educational research journal!!!" You don't
have to send me an angry e-mail message about it-I'll cheer-
fully concede right now that if I bet against Professor X or Y
or against you I'll lose. In Vegas the casinos lose thousands
of gambles every hour. They make many thousands of
gambles, though, and the odds are with them. In the long run,
they always win. -


All of the Random T columns are now available on the World Wide Web at
http://www.ncsu.edu/effective_teaching and at http://che.ufl.edu/~cee/

Winter 2006











MRa laboratory









A FLEXIBLE PILOT-SCALE SETUP

FOR REAL-TIME STUDIES

IN PROCESS SYSTEMS ENGINEERING







CHANIN PANJAPORNPON, NATHAN FLETCHER,* AND MASOUD SOROUSH
Drexel University Philadelphia, PA 19104
T he inclusion of process control experiments in chemi- Chanin Panjapornpon is currently a Ph.D. can-
cal engineering curriculums and the introduction of didate in the Department of Chemical and Bio-
new process control :'.plC ,enic, indicate recog- logical Engineering at Drexel University He re
ceived his B.Sc. from Chulalongkorn University,
nition of the importance of real-time experiments in process Thailand, in 1995 and his M.S. from Drexel
systems engineering. The experiments allow an instructor to University in 2002. His industrial experience
includes five years with a petrochemical com-
reinforce and demonstrate theoretical systems concepts pre- pany in Thailand, and his research interests
sented in lectures. Laboratory systems experiments in an aca- are in the areas of nonlinear model-based con-
trol, optimization, computer control, and con-
demic setting provide students with an invaluable opportu- troller-design software.
nity to familiarize themselves with important practical issues
(i.e., nonideality of industrial processes), such as process- Nathan W. Fletcher received his B.S. in
Smis i r chemical engineering from Drexel University
model mismatch, measurement noise, inadequate number of in 1999. He was with Automation Application
measurements, digital measurements, actuator saturation, Inc., in Exton, Pa., from 1999 to 2004. He
unmeasured disturbances, and process nonlinearity-issues implemented DCS, PLC, and hybrid systems
for the specialty chemical, oil and gas, pulp
often neglected in computer simulations. and paper, and food industries. In mid-2004,
he joined Fluor Life Sciences in Media, Pa.
This manuscript describes a low-maintenance, low-safety- His professional interests are in instrumenta-
risk, flexible, 0.9-m X 1.5-m X 2.4-m, pilot-scale setup that tion and control.
can be used for training students and carrying out research in
process systems engineering. It briefly states typical applica- egeegd Soroush received a.anS. (Institte
tions of the setup. Detailed specific sample applications of Technology, Iran, and two M.S. (chemical en-
the setup, together with real-time results, will be presented in gineering, 1988, and electrical engineering:
systems, 1991) and a Ph.D. (chemical engi-
forthcoming paperss. The setup was built in the Department neering, 1992) from the University of Michi-
of Chemical and Biological Engineering at Drexel Univer- gan. He is now a professor of chemical and
biochemical engineering at Drexel University,
sity and is located in the Process Systems Engineering Labo- and has worked as a visiting scientist at
ratory. The setup allows one to study a variety of process- DuPont Marshall Lab, Philadelphia. His cur-
systems engineering concepts such as design feasibility, de- based chntrhhert are innonlnrmrela-
tion, nonlinear state and parameter estimation, fault detection and iden-
tification, and fuel-cell modeling, optimization, and control.
*Current address: Fluor Enterprises, Inc., Rose Tree II, Suite 5000,
1400 N. Providence Rd., Media, PA 19603 Copyright ChE Division of ASEE 2006
40 Chemical Engineering Education










sign flexibility, control configuration selection, parameter
estimation, process and instrument fault detection and
identification, controller design and implementation, in-
strument calibration, and process modeling. Notable fea-
tures of the setup are its flexibility and low safety risk
(because it uses water only). The setup can be single-vari-
able or multivariable, mildly or strongly nonlinear, interacting
or noninteracting, and/or single- or multi-tank. It has features
of both apparatus # 4 and 10 described by Ang and BraatzM1];
it can be configured to be the same as apparatus #4 or 10,
or a combination of apparatus #4 and 10. The setup can
be used in both undergraduate and graduate process con-
trol laboratories to reinforce, through hands-on experi-
ments, the concepts taught in process control and process
analysis lectures.

PILOT-SCALE SETUP
A picture of the 0.9-m X 1.5-m X 2.4-m pilot-scale setup
is shown in Figure la, and a schematic in Figure lb. The
setup has two identical, clear-plastic, cylindrical tanks. Each
tank has an outside diameter of 0.2 m and a height of 1.0 m.
The tanks can be connected to each other (by easy-connect/
disconnect flexible hoses) in several ways, which allows one to
operate the setup as a system of a single tank, two parallel
tanks, two interacting tanks in series, or two noninteracting
tanks in series. The elevation of the second tank can be ad-
justed (via a jack) to alter the level of the interaction between
the two tanks. Inside both tanks, there are helical copper tubes
(i.e., coiled copper tube banks) that can be used for heating
or cooling, depending on the temperature of the water flow-
ing into the copper tubes. One end of each copper tube is
connected by a hose to a city water supply that is cold, hot, or
a mixture of both-allowing adjustment of the inlet tempera-
ture of the water stream flowing into the copper tubes. Ther-
mal energy can also be supplied to each tank by an electrical
heater consisting of two heat cartridges inside the tank. Each
tank has a variable-speed agitator.
The setup has eight resistance temperature detectors
(RTDs), two flowrate sensors, two level sensors, and one
control valve. The RTDs measure the temperature of the in-
let and outlet streams of the tanks and the .*,Iili % lic'.liiui
copper tubes. The level sensors measure the level of water in
the tanks. The flowrates of two inlet streams are measured
by the two online and two off-line (rotameter) flow meters.
A control valve adjusts the flowrate of a water stream flow-
ing into Tank 1.

ELECTRONIC HARDWARE
Analog Input Devices
Each of the sensors measures a process variable and gener-
ates a 4-20 mA analog signal, which is then sent to an analog
input channel of a data acquisition board. The board then
converts the analog signal to a digital signal. There are three
41


Figure 1. The pilot-scale setup, in photograph (a)
and schematic (b).


Winter 2006












types of analog instruments in this setup
3 Pulse-Output Flow Meter. The paddle flow meter
generates a positive pulse signal when its rotor
is rotated by the fluid flow. The pulse signal is then
converted to 4-20 mA signal proportional to the
flowrate (0-5 gpm).

3 RTD Temperature Sensors. The resistance tempera-
ture detectors are connected to a Wheatstone bridge


and sends one analog signal to the control valve and two bi-
nary signals to the two heaters. The data-acquisition board
can communicate with the central processing unit via inter-
face software, such as Visual Basic, Visual C++ (Microsoft
Corp.), and LabVIEW (National Instruments Corp.). The soft-
ware then analyzes the data. A graphical user interface (GUI)
is then used to present the data. In this setup, the data-acqui-
sition application is developed by using Visual Basic as the


circuit that uses a reference
resistor of 100 ohms, which
corresponds to 0 oCelsius.

3 Level Sensors. The level sensors
used in the setup are
pressure-transducer type. The
liquid static pressure in a tank
presses on the diaphragm
transducer, generating a
proportional analog signal.

Analog Output Devices

The proportional control valve is an
analog output device that receives an
analog input signal (4-20 mA) and sets
the flowrate proportionally. It is a fail-
to-close control valve. The power of the
heaters is adjusted by a solid-state re-
lay (SSR), which is connected to a
pulse-controller module (both from
Omega Engineering, Inc.). The module
allows simple conversion of the on/off
SSR to a proportional power regulator.
Therefore, the average power to the
heater is proportional to the input 4-20
mA analog signal to the module. Each
of the electrical heaters consists of
two High Watt Density Cartridge
Heaters (rated power of each: 1.5 kJ/s at
240 volts).

Data-Acquisition Board

A DAS-1701ST-DA data-acquisi-
tion board[4 5] and an EXP- 1800 exten-
sion board[6] (both from Keithley Instru-
ments, Inc.) are used. The data-acqui-
sition board has eight analog input chan-
nels. It receives and time-discretizes the
incoming analog input signals. The ex-
pansion board allows one to expand
each input channel of the data-acquisi-
tion board to eight input channels.
Therefore, the data-acquisition system
can support up to 64 analog input chan-
nels. It receives 12 analog signals from
the eight temperature sensors, the two
flow meters, and the two level sensors,

42


Level [ank i. j. -


L.


-*i" .. .' [-
ir Ir I: 1

tIF r '1r-:r


,,r, 1ip ,,w _


j.. :
Hieal Lonbol pdll bp Digldal Saysem
r .I r 'ri.i. I




4 n a l ol ) O u lp -il No C .n h o l V 'l e
rF r [ -

I... .... i... :


lank I







PI[ .1- ,,',, Ii- .
I 1- .: ri ,




, -, j i





ilnl n

I I
51 '~


S I, : I.


,II i II. i. .TE

FT_


he l z .


l I.J, I. j ., ,J I, h. t

Figure 2. Front-end interface for the level control.


Chemical Engineering Education


J. U


Iu ,


I 1. i II. i ..I .t -1.. n.iw .1. I tl 'l


-- - -- - - -- --





i-











front end and C++ as the backbone. Visual Basic can be
used to create a GUI easily, and C++ used to support Win-
dows-based input/output operations. For this setup, front-
end windows for temperature control, flow control, and
data storage are developed. Two of the windows are shown
in Figures 2 and 3. Data from the setup can be saved as
Excel files and then be imported to Matlab (Mathworks,
Inc.) easily.


TYPICAL APPLICATIONS
OF THE PILOT-SCALE SETUP
With the flexibility to operate in various configurations,
and its many sensors and actuators, the setup allows real-time
study of a variety of process-systems engineering concepts. Fig-
ure 4 shows the variables that can be measured and/or adjusted
in this setup. Below is a brief description of typical real-time
studies that one can perform using the setup.

Process Modeling
Given the online measurements, models including first-prin-
ciples, empirical (black box), or hybrid (first-principles/empiri-
cal)[7 can be developed to describe water temperature and/or
level in one or both tanks. In the case of empirical and hybrid
modeling, the students can be taught model-parameter esti-
mation as well. Hybrid model parameters include the resis-
tances of the tank exit pipes as well as the overall heat trans-
fer coefficients of the coiled copper tube banks. The model
structure is obtained from mass and energy balances in the
cases of first-principles and hybrid modeling, and from prior
process knowledge (an assumption) in the case of empiri-
cal modeling. The empirical modeling can be off-line or


online. In the latter case, one must use a model identifi-
cation method."1
Process Design Analysis
The setup can be used to analyze the following process
design aspects:
3 Feasibility. Given desired steady-state values of
temperatures) and levelss, and nominal values of
temperature and flowrate disturbance stream
(inlet stream with no control valve), students are asked
to evaluate theoretically and experimentally the
feasibility design to operate at the desired
steady state; that is, to check whether the design can
provide heater power, water flowrate, and energy
the heating cooling coils) adequate to
operate the process at the desired steady state. 8 For
example, a desired water temperature below the city
water temperature is definitely infeasible.
3 Flexibility. Flexibility is feasibility in the presence of
uncertainties such as disturbances and parameter
uncertainties variations. In this analysis, the students
are asked to evaluate the feasibility l design to
operate at a given steady state when the temperature
and flowrate disturbance stream vary within a
given range.8 Students can map theoretically the
disturbance region in which the design is feasible and
then verify the region experimentally.

Process Control
The setup can be used to carry out the following process
control studies:
3 Measurement Selection. Many control problems with
one or more objectives can be posed, and students are


SLLl _,J.



ii I I I "1 l -i


." J: .I .r J .',.J.'" ,I ... ..2.. .. IJ .,. . I i.t .., I. .?: .. I' ..


Figure 3. Front-end interface for temperature control and data storage.
Winter 2006


Figure 4. Adjustable and
measured variables of the
setup.











Laboratory systems experiments in an academic setting provide
students with an invaluable opportunity to familiarize themselves
with important practical issues (i.e., nonideality of industrial processes),
such as process-model mismatch, measurement noise, inadequate number
of measurements, digital measurements, actuator saturation,
unmeasured disturbances, and process nonlinearity-issues
often neglected in computer simulations.


then asked to list the measurements needed to achieve
the control objective. These objectives include control
of temperature andlor level in Tank 1, and or control
of temperature and/or level in Tank 2. For example,
for control of temperature in Tank 1, at least, the
temperature measurement T5 is needed.
3 Control Configuration Selection. After choosing the
necessary measurements, students can be asked to
propose a set of manipulated inputs that can be used
(adjusted) to realize the control objectivess. The
controlled outputs should be controllable from the
manipulated inputs. For example, temperatures in
Tanks 1 and 2 are controllable from heater power P1
and P2. The state and/or output of
the control configuration can be tested.
3 Input-Output Pairing. For multi-input, multi-output
(MIMO) control problems, students can be asked to
pair the inputs and outputs selected control
configuration so that completely decentralized control
can be implemented. To evaluate the level of
interactions among the process variables, students
can use tools such as the relative gain array,1o01
relative orders,t"I and/or time delays to propose
,e pairs.121
3 Controller Selection. One can select a feedback or a
S1 control system depending on
what control system is desired: completely decentral-
ized (set of single-input, single-output, or SISO,
controllers) or centralized (multivariable). For
example, one can use the flow measurement F2 and
the temperature measurement T2 (measurements of
disturbance inputs), to add loop(s) to
feedback control of temperature and/or level control
in Tank 1. Furthermore, the controller can be: (1) a
conventional controller, such as a proportional (P), a
proportional-integral (PI), or a proportional-integral-
derivative (PID) controller; or (2) an advanced
controller such as a model-based controller., '01 The
setup can be used to understand the limitations of
decentralized control and implement decouplers in
reducing the of interactions. Further, the model-
based controller can be analytical (such as an
input-output linearizing controller) or numerical
(such as a model-predictive controller). Whether
conventional or not, adaptive features can be


added to the controller.'7 o In real time, students
can observe and compare the performance of .
controllers, and evaluate the pros and cons of each.
Parameter Estimation
Given the flowrate, level, and temperature measurements,
students can estimate the heater powers and the heating/cool-
ing coil-tank overall heat-transfer coefficients. In the case of
the heater powers, since the heater powers are set by the com-
puter, the values of the heater powers are known. This allows
one to evaluate the accuracy of the estimated heater powers
by comparing them to the actual values.
A parameter estimator that can be implemented in real-time
on this setup is described by Tatiraju and Soroush.[13]
Fault Detection and Identification
The equipment can be used to demonstrate fault detection
and identification.
3 Sensor Fault Detection and Identification. The setup
can be used to learn sensor fault detection and
identification in real-time. Noise, andor bias are
added to a sensor reading, and a sensor fault
detection and identification method is then used to
detect the fault in the sensor and identify the fault type
(noise, '" and/or bias). An example of such a study
can be found in Mehranbod and Soroush,1'4 in which
sensors L1, Fl, and F2, and a PI controller to control
the liquid level in Tank 1, were considered.
3 Process Fault Detection and Identification. Partial
or complete failure of one process actuators is
an example of a process fault. A process fault
detection and identification method can be used to
detect an actuator failure and the type failure.
An intentionalfault can be introduced in any
actuators, and it can be detected and identified in real
time by using a process fault detection and identifica-
tion method.

Instrument Calibration
The setup has three actuators and 12 sensors. For each ac-
tuator, a calibration curve is obtained by finding the relation
between the raw digital signal (that the computer sends to the
data-acquisition board) and the actual value of the correspond-
ing physical variable. For example, the control-valve cali-


Chemical Engineering Education











bration curve can be obtained by measuring the flowrate with the rotame-
ter at different, constant, raw, digital signals set at the computer. For a
sensor, a calibration curve is obtained by finding the relation between the
raw digital signal that the computer receives from the data-acquisition
board and the actual value of the corresponding physical variable. For
example, an RTD is calibrated by placing it in beakers of water at
different known temperatures and recording the value of the corre-
sponding steady-state, raw, digital signal received by the computer. A
typical calibration curve is presented in Figure 5. It shows how the
flowrate of the water stream through the control valve depends on the
raw digital signal.

Calorimetric Studies
The electrical heaters can be used to simulate heat of reactions. An exo-
thermic reaction or set of exothermic reactions can be considered and
simulated on the microcomputer, and the rate of heat production by the
simulated reactions) is then sent to the heater to set the heater power to
the calculated rate of heat generation. Material and energy balances
for the tanks, considered with the temperature and flowrate measure-
ments, can then be used to estimate the power to the heaters; that is,
the rate of heat production by the simulated reactions.

CONCLUSIONS
This manuscript describes a low-maintenance, low-safety-risk, flexible,
pilot-scale setup that can be used for training students and carrying out
research in process systems engineering. It briefly states typical applica-
tions of the setup. Detailed specific sample applications of the setup to-
gether with real-time results will be presented in forthcoming paperss.
The setup allows one to study a variety of process-systems engineer-


4000


3000



"F 2000



1000
1000


4 5


0 1 2 3
flow-rate (gpm)


Figure 5. Calibration curve for the control valve.
Winter 2006


ing concepts in real time. Among these concepts
are design feasibility, design flexibility, control
configuration selection, parameter estimation,
process and instrument fault detection and iden-
tification, controller design and implementation,
instrument calibration, and process modeling.
The setup can be used to provide graduate and
undergraduate students with hands-on experi-
ence and to carry out research in process sys-
tems engineering.

ACKNOWLEDGMENTS
The authors would like to thank Srinivas
Tatiraju, Neeraj Zambare, and Roberto Pena for
their input into the project, and Dan Lau for his
essential role in assembling the setup. The au-
thors would also like to thank the Department of
Chemical and Biological Engineering at Drexel
University for supporting this project.

REFERENCES
1. Ang, S., and R.D. Braatz, "Experimental Projects for
the Process Control Laboratory," Chem. Eng. Ed.,
36(3), 182 (2002)
2. Gatzke, E.P, R. Vadigepalli, E.S. Meadows, and FJ.
Doyle III, "Experiences with an Experimental Project
in a Graduate Control Course," 33(4),
270 (1999)
3. Johansson, K.H., "The Quadruple-Tank Process: A
Multivariable Laboratory Process with Adjustable
Zero," IEEE Trans. Contr Sys. Tech., 8, 456 (2000)
4. Keithley Instruments, DAS-1700 Series User's Guide
(1996)
5. Keithley Instruments,DAS-1700 Series Function Call
Driver (1996)
6. Keithley Instruments,EXP-1800 User's Guide (1995)
7. Ogunnaike, B.A., and W.H. Ray, Process Dynamics,
Modeling, and Control, Oxford University Press, 1st
Ed. (1994)
8. Grossmann, I.E., and M. Morari, "Operability, Resil-
iency, and Process Design Objectives for a Changing
World," Proceedings of the 2nd Int. Conf on Founda-
tions ofComputer-Aided Process Design, Westerberg,
A.W., and H.H.Chien, Eds., 931-1030 (1983)
9. Chen, C.-T., Linear System and Design, Holt,
Rinehart, and Winston (1970)
10. Seborg, D.E., T.F. Edgar, and D.A. Mellichamp, Pro-
cess Dynamics and Control, 2nd Ed. (2003)
11. Daoutidis, P., and C. Kravaris, "Structural Evaluation
of Control Configurations for Multivariable Nonlin-
ear Processes," ( Sci., 47, 1091 (1992)
12. Holt, B.R., and M. Morari, "Design of Resilient Pro-
cessing Plants-V: The Effect of Deadtime on Dynamic
Resilience," ( Sci., 40, 1229 (1985)
13. Tatiraju, S., and M. Soroush, "Parameter Estimator De-
sign with Application to a Reactor," Ind. Eng. Chem.
Research, 37(2), 455 (1998)
14. Mehranbod, N., and M. Soroush, "A Method of Sen-
sor Fault Detection and Identification," J. ofProcess
Contr., 15(3), 321 (2005) 1












[e]R: laboratory


MECHANICAL TESTING OF

COMMON-USE POLYMERIC MATERIALS

WITH AN IN-HOUSE-BUILT APPARATUS


CRISTIANA PEDROSA, JOAQUIM MENDES, FERNAO D.
University ofPorto 4200-465 Porto, Portugal
he polymer production or transformation industries
employ a very significant percentage of chemical en-
gineers. This justifies the presence of a variety of poly-
mer science and engineering subjects in the chemical engi-
neering undergraduate curricula. Topics on solid polymer me-
chanics, in particular, are often quite useful for future engi-
neers. They establish the basic tools for evaluating whether a
material is appropriate for an intended use, or to tune its per-
formance by acting on the synthesis/processing conditions.
These subjects are present in general polymer textbooks (e.g.,
References 1 and 2). Because ChE student laboratories are
not traditionally equipped with the machinery used for me-
chanical testing, however, introductory courses on this sub-
ject often suffer from not having an appropriate applied com-
ponent. Therefore, students don't gain a hands-on understand-
ing of the phenomena involved.
One of the most commonly used mechanical tests in indus-
try is tensile testing, in which the stress exerted by the mate-
rial is measured at a constant strain rate. In addition to pro-
viding direct measurements of relevant properties, the stress-
strain curves constitute "fingerprints" of a material's mechani-
cal characteristics. The stress-strain curves are also often used
for quality control of either raw materials or final products.
Most mechanical testing machines can perform several stan-
dardized tests (tensile, compression, flexural, etc.) by using
appropriate accessories. Such machines are designed for
heavy loads, however, and are significantly expensive. In
addition, in order to use such a "heavy duty" machine, one
has to prepare a test specimen that will exhibit a measurable
stress-strain behavior. Furthermore, the specimens must be
cut or cast into a standard shape and dimension, which might
not be easy for many materials of interest. Standard-shape
polymeric specimens of known composition and molecular
weight are commercially available, but at a cost.


MAGALHAES


There is also equipment available for low stress/strain mea-
surements-suitable for testing specimens of smaller dimen-
sions, such as thin films-but these are also high-priced (about
$8,000 USD).
On the other hand, polymeric materials are readily avail-
able in the form of everyday-use items. Even informal obser-
vation of the behavior of these materials under mechanical
solicitation may give the attentive student a wide variety of
illustrations for important concepts in solid-polymer mechan-


Cristiana Pedrosa graduated in chemical en-
gineering from the Faculty of Engineering of
the University of Porto (FEUP), Portugal, in
2004. She is currently working as a research
assistant on permeation measurements on
porous materials.
Joaquim G. Mendes
is an assistant pro-
fessor in the me-
chanical engineering
department at FEUP,
Portugal. He gradu-
ated in mechanical engineering from FEUP in
1988 and obtained a post-graduate degree in
automation and management of industrial pro-
cesses in 1989. He received his M.Sc. in indus-
trial computing and his Ph.D. in industrial auto-
mation from FEUP and the University of Minho,
respectively. His research interests include sen-
sors, data acquisition, remote labs, and virtual instrumentation.
Fernao D. Magalhaes is an assistant profes-
sor in the chemical engineering department at
FEUP, Portugal. He graduated in chemical en-
gineering from the same faculty and received
his Ph.D. from the University of Massachusetts,
Amherst, in 1997. Among other courses, he is
currently teaching an introductory course on
polymer science and technology. His main re-
search interests involve-in addition to poly-
meric materials applied to the wood and paint
industries-mass transport and sorption in po-
rous solids and membranes.


Copyright ChE Division ofASEE 2006


Chemical Engineering Education











ics. An inspiring example was provided by J. Walker in the
"Amateur Scientist" column of '. '' American maga-
zine,[31 in which the molecular phenomena involved in the
stretching of a polyethylene film are illustrated and discussed
in a simple and captivating fashion.
Our challenge was to build, on a very tight budget, a ten-
sile-testing machine that chemical engineering undergradu-
ate students could use freely in the lab for testing polymeric
specimens gathered from common-use objects selected and
prepared by them. Some of the design criteria we adopted for
this project were:
Keep costs as low as possible, without compromising
the quality machine's measurements (i.e.,
reasonable accuracy and reproducibility).
Keep in mind that the specimens should be easy to
obtain and prepare. Using thin films for testing
seemed to be a good idea. Many materials (plastic or
otherwise) are commonly available in that form and
can be easily cut into standard shapes. This option
implies designing the machine for small loads and
strains.
Build a compact setup, so as to allow portability and
use inside temperature-controlled chambers (e.g.,
refrigerators and ovens).
E Make it fully automated, total control of its
functionalities a computer-based data-
acquisition system.
E Make it operationally robust, safe, and intuitive, since
students are supposed to operate it themselves.
This paper describes the machine developed, as well as
some of the experiments performed. This project has been
successful in giving students a hands-on perspective on some
key aspects of the mechanical behavior of polymeric materials.

SETUP OF THE TENSILE-TESTING MACHINE
The design of the testing machine comprised four key com-
ponents: (1) a set of two grips, which firmly hold both ends
of the test specimen; (2) a motor, which pulls one of the grips
at a constant pre-set speed; (3) a force sensor, which mea-
sures continuously the force exerted on the material; and (4)
a displacement sensor, which measures the distance traveled
by the moving grip during the test. Figure 1 shows the origi-
nal sketch of the machine's layout, comprising these compo-
nents. Further details are discussed below.
We intended to mostly test polymeric materials, such as
polyethylene, in the form of thin films. We looked at some of
the standardized tests used in industry to have an idea of the
sizes and shapes of the test specimens used. According to
ASTM D882-02,[41 the plastic films being tested are cut into
rectangular specimens (at least 150 mm in length and 5 to 25
mm in width). On the other hand, on ASTM D2370-
98(2002), [5 which applies to organic coatings (e.g., elasto-
Winter 2006


meric paint films), the specimens are also rectangular in shape,
but the length may be lower (at least 50 mm in length and 13
to 25 mm in width). We decided to use a short specimen length
to minimize the maximum strain involved in the tests, and
thus allow the use of a reasonably priced continuous-displace-
ment transducer-and keep the machine's size small. There-
fore, we adopted the latter standard. (Note that we actually
also tested paint films with this machine.) Some crude pre-
liminary tests done with film strips cut from supermarket
plastic bags gave us the basic information for the specifi-
cations for the load sensor, the motor, and the displace-
ment transducer.
To provide linear motion to a lead screw, we chose a per-
manent magnet-stepper motor that employs a rotor with an
internal thread. One end of the screw pulls one of the grips.
The other end is attached to the moving lead of the LVDT-
type displacement transducer. This transducer was the most
expensive component in the setup (about 35% of the total
cost). Cheaper alternatives are available, but an LVDT offers
high accuracy and reproducibility, as well as wear-free and


Figure 1. A 3-D sketch of the tensile-testing machine,
showing the major components: (1) grips; (2) motor;
(3) force sensors; (4) displacement sensor.


0-I











frictionless operation. Two force sensors, of piezoresistive
type, are attached to the end of the lead screw. These sensors
can measure only compression loads. The pulling grip is at-
tached to a transversal bar that sits on top of the force sen-
sors, thus transferring the tensile load, as is shown in Figure
1. Table 1 lists the main components and their costs. Figure 2
shows a photo of the actual unit, in its final working form.
The machine can measure loads up to 30 N. The maximum
strain rate is 300 mm/min and the maximum linear displace-
ment is approximately 140 mm, but this value can be increased
by using a longer lead screw combined with longer lateral
support bars. The machine is monitored and controlled with
a desktop computer using a data-acquisition card (National
Instruments PCI-6014). The program LabVIEW 7.1 (from
National Instruments) was used to develop the software that
fully controls the apparatus and analyzes the measured data.
Figure 3 summarizes the information flow between the com-
puter and the machine.


This machine was first implemented in an introductory
on solid polymer mechanics, which is an optional part
chemical engineering under-
graduate program. The students
perform the tests themselves, on
materials that they have gathered


Figure 2. Photograph of the
tensile-testing machine
(showing the holders used for
the three-point bending test).
48


according to the instructor's suggestions. They analyze the
results both qualitatively and quantitatively, computing dif-
ferent parameters from the measured data.
Some representative tests and results are described in the
ensuing text. For the sake of conciseness, the discussion in
this paper is kept on a qualitative level.

TENSILE AND TEAR TESTING HDPE FILMS
Students are asked to prepare test specimens from plastic
shopping bags obtained at their favorite supermarkets. These
are commonly made of high-density polyethylene (HDPE).
Students can easily identify the polymer by noting the recy-
cling symbol that is typically printed on the bags.
The specimens to be used on the stress-strain tests are cut
into 60 x 15 mm rectangles and reinforced with adhesive tape
at their extremities, covering 20 mm on each end (see Figure
4a). The grips hold the specimens by grabbing on these rein-
forced ends.
Special care must be taken in cutting these specimens. The
cut should be perfectly straight and without indentations; im-


Component Maker and Model Specifications Price
(pre-tax, USD)
Stepper motor + lead screw Mclennan L92411-P2 Max. linear force = 88 N $263
Driver board for stepper motor Eurocard $48
Displacement transducer Solartron DC50 920128 Range = 75 mm $463
(LVDT)
Force sensors (2) Honeywell FSG Max. load p/sensor = 15 N 2 X $75
(piezoresistive)
Power supplies (2) EMS B811 and Astec LPS23 12 V $60 + $53
Holding structure and grips Local workshop $250
(carbon steel + polyacetal)
Other components $33
TOTAL $1320


Computer Data Acquisition Board Step Motor Drive Tensile Machine
SPulse train
Up / down direction ..-
Half / full step I





-Load '
Displacement


Figure 3. Schematic representation of digital and analog signal flow between the
machine and the data-acquisition system.
Chemical Engineering Education


TABLE 1
Tensile-Testing Machine Components
(Not Including Data-Acquisition Board and Computer)










perfections may cause the films to tear prematurely instead
of reaching the ultimate rupture point. Sharp scissors or a
fresh razor blade should be used. Students are asked to cut
the specimens in two different ways: longitudinally and
transversally in relation to the direction of the bags' "verti-
cal" position (see Figure 5). Each specimen is labeled with a
soft-tip marker so that the information on the specimen's
cutting orientation is not lost.
Students also prepare specimens for tear-strength tests,
which consist of 40 x 40 mm squares with an initial indenta-
tion (10 mm long) at the center of one of the sides (see Figure
4b). These indentations are done so that the tearing will propa-
gate as intended: along the longitudinal or transversal direc-
tions mentioned before.
Before performing the stress-strain tests, students measure
the thickness of each film, using a digital micrometer. This
information is used to compute the initial cross-sectional area
of the specimen, on which the loading stress will be based.
Typical values are in the order of 10-2 mm.
Figure 6 shows representative results for two specimens
cut along perpendicular directions as described before. The
distinct behavior presented by the two is quite noticeable.
The specimen cut along the bag's "longitudinal" direction (L


a) 15mm b) 40 mm




E adhesive tape
EI



indentation

Figure 4. Specimen cut from plastic bag films: a) for
stress-strain tests and b) for tear-strength tests.


I-



I

longitudinal
direction


transversal
direction

"T" specimen: cut along
transversal direction
"L"specimen: cut along
longitudinal direction


Figure 5. Definition of the longitudinal and transversal
directions on a plastic bag.
Winter 2006


specimen) shows a rapid increase in stress, followed by a
short plateau and afterwards a gentler increase, up to rupture.
On the other hand, the T specimen (cut along the transversal
direction), after a similar initial raise goes through a very
well-defined maximum in stress and then stabilizes on an
essentially constant value, almost up to the final rupture. The
stress for this specimen is always significantly lower than for
the L specimen.
Another fundamental difference can be observed when,
prior to performing the tests, horizontal lines (perpendicular
to the direction of elongation) are drawn with a soft-tip marker
at different sections along the specimens. As the L specimen
is elongated, one can see that the lines increase almost iden-
tically in thickness, up to rupture; this indicates that the ma-
terial is being uniformly deformed. On the other hand, on the
T specimen some lines become noticeably thicker as others
remain almost unchanged; this indicates that the specimen is
being stretched at the expense of deforming the material in
limited regions. As elongation continues, the extent of the
undrawn regions successively decreases until the entire speci-
men becomes uniformly stretched and rupture occurs. Some
students recognize this as being an example of cold drawing-
a phenomenon discussed in previous classes. It occurs on some
semicrystalline polymers, like HDPE. The stress maximum cor-
responds to the yield point and the onset of necking.
But why is cold drawing not observed on the L specimen?


0 50 100 150 200 250
Elongation (%)


300 350 400


Figure 6. Typical stress-strain curves obtained for high-
density polyethylene films from a plastic shopping bag. The
two specimens were cut along two perpendicular directions
(see description of L and T specimens in Figure 4). The end
point on each curve corresponds to rupture of the film.
Operating temperature = 20 oC; strain rate = 200 mm/min;
original film thickness = 0.016 mm. Note that the stress
indicated is the "conventional" or "engineering" stress,
i.e., the measured load divided by the initial cross-
section of the specimen. Decreases in cross-section along
the test are disregarded.










This mechanical anisotropy is not usually expected by stu-
dents, and they are encouraged to offer explanations. The tear-
strength tests that are performed afterward supply extra ma-
terial for the discussion.
The test is performed as schematized in Figure 7. The force
exerted by the material is measured as the ends of the speci-
men are pulled at a constant rate and the tear propagates.
Typical results from this test are shown in Figure 8.
When the tearing propagates along the longitudinal direc-
tion, the force is essentially constant and relatively low; in
the end one observes that tearing originated a straight and
clean cut. On the other hand, a much higher force is neces-
sary to tear the specimen along the transversal direction; vi-


Figure 7. Schematic representation of a tear-strength test.


0 20 40 60
Displacement (mm)


80 100


Figure 8. Typical tear-strength curves obtained for polyethyl-
ene films cut from a plastic shopping bag. Tearing propagates
along perpendicular directions on the two specimens (see
description of transversal and longitudinal directions on
Figure 4); same operating conditions as in Figure 5.


usually one can see that the material is stretched and distorted
during the test and the final cut shows permanent deforma-
tion of the material at the edges.
After analyzing this second set of results, students often
suggest that this anisotropy is associated to a particular mo-
lecular orientation of the polyethylene chains in the shop-
ping bag. It becomes clear that this is the correct hypothesis
after the instructor describes the manufacture process for
HDPE bags, commonly known as ii extrusion. This
process is described in several processing handbooks.16, It
involves submitting the polymer to a sequence of transfor-
mations: melting, extrusion, blowing, drawing, cutting, and
sealing. A continuously extruded thin-polymer tube is inflated
by a jet of air blown into it. The blowup ratio (defined as the
ratio between the diameters of the expanded film bubble and
the die) controls the molecular orientation along the trans-
versal direction. This ratio is usually 2 to 4. On the other
hand, the drawdown ratio (the ratio between the speeds of
the film at the nip rolls and at the die exit) determines the
longitudinal orientation (called machine direction). A balance
between these two parameters governs the final orientation
within the film. Partial crystallization occurs as the material
is cooled, thus conserving the molecular orientation imposed.
This flow-induced crystallization is actually a bit more
complex than it would seem, due to the particular morphol-
ogy that polymers exhibit upon crystallization. At a suffi-
ciently high drawdown ratio, film-blown HDPE undergoes a
so-called row-nucleated c. ' Extended molecu-
lar chains oriented along the machine direction form fibrillar
structures that act as nuclei for the crystallization of the bulk
material, in the form of radially grown lamellae. Figure 9
schematically illustrates these structures.



fibrils
/



1- lamellae






Machine
direction

Transversal
direction

Figure 9. Row-nucleated morphology of film-blown
HDPE. The fibrils oriented along the machine direction
act as nuclei for the growth of the lamellae (chain-
folded crystalline structures).
Chemical Engineering Education


direction of tear
propagation










These long fibrils with perpendicular growths are often
appropriately described as shish kebabs. They are respon-
sible for the high tensile strength of the material along the
machine direction (corresponding to the response of the L
specimen in Figure 6).
Because, in the case of HDPE, there is no significant inter-
connection between the "kebabs" of adjacent fibrils, the ten-
sile strength in the transversal direction is significantly lower.
This transversal straining causes a noticeable yielding of the
material, associated to local fibrillar reorientation toward the
direction of the applied strain. The stress remains typically
constant along this drawing process (see the curve for the T
specimen in Figure 6). When the fibrillar rearrangement has
extended throughout the entire material, the stress often rises
slightly and rupture occurs shortly after. This cold-drawing
phenomenon is characteristic of many semicrystalline poly-
mers and is described in most polymer science textbooks. A
recent paper by Zhang, et al,[9 provides an interesting dis-


Figure 10. Schematic representation of a three-point
bending test. The two holders at the extremities are
connected to the machine's upper grip and move at a
constant rate. The holder at the center is attached to the
lower grip and remains stationary. The two holders at the
extremities are 30 mm apart. The specimen length is
about 50 mm.


15




5 -T= 80 C v= 100 mm/min
-- --- T= 20 C v= 2 mm/min

0 1 2 3 4 5
Deflexion (mm)

Figure 11. Typical force vs. deflexion curves obtained for
polystyrene coffee stirrers. The labels indicate the
specimens' temperature in Celsius (T) and the deflexion
rate (v). The end point on each curve corresponds to
rupture of the material.
Winter 2006


cussion of the mechanical anisotropy and crystalline mor-
phologies of different kinds of polyethylene-blown films.
From Figure 6 we see that the tensile strength (defined as
the maximum stress measured during the test) is about 2.5
times higher when the material is strained along the direction
that coincides with the fibrillar orientation (which we named
the longitudinal direction). It makes a lot of sense that plastic
bags are assembled so that normal use implies applying the
stress along this direction.
The tear-strength test results (Figure 8) further confirm these
findings. It is clear that it will be much easier to tear the ma-
terial along the direction of the fibrils than along the trans-
versal direction.
The "cherry on top" for this set of experiments comes when
the instructor suggests that students take a piece of the plas-
tic bag and heat it above a flame lighter (holding it with pin-
cers and taking care to not actually bum the material). Imme-
diately they see that the film starts to crumple and shrink.
The temperature rise causes the gradual melting of the crys-
talline regions, loosening the mobility of the polymeric chains
(the melting temperature of polyethylene is about 140 C).
This allows the originally extended chains to rearrange to-
ward the more favorable coiled conformation. The result is
the crumpling of the polyethylene film. Further heating would
cause the total disappearance of the crystalline regions, re-
sulting in a polymer melt.

FLEXURAL TESTING OF PS BARS
The machine is limited to tensile testing of thin films of
relatively soft materials. Glassy polymers cannot be tested,
since they involve much higher tensile stresses. Nonetheless,
we have adapted the machine to perform a different kind of
test on such materials: a three-point bending test at constant
deflexion rate. We used polystyrene (PS) coffee stirrers, col-
lected from a nearby vending machine, as test specimens.
Figure 10 schematizes how the test is implemented: by using
hard-wire hooks to attach the specimen to the grips. PS is
glassy at room temperature (its glass-transition temperature,
Tg, is about 100 C).
Figure 11 shows some of the results obtained for different
temperatures and deflexion rates. Since it is faster to place
the rigid PS specimen on the support hooks than it was to
attach a film to the grips, these tests did not involve placing
the machine in a temperature-controlled chamber. Rather, the
specimens were stabilized in an oven at the desired tempera-
ture. Prior to testing, each specimen was quickly transferred
to the machine. The entire procedure (including performing
the test) took no more than 30 seconds. In the figure one can
see that the maximum deflexion is relatively low, as expected
from a glassy polymer. For a deflexion rate of 100 mm/min,
as the specimen temperature approaches the glass-transition
temperature, its softness is significantly increased. Students


-$ deflexion
----- ----- --- T










are asked to visually inspect the broken specimens. They
notice that the ones tested at higher temperature show a vis-
ibly higher extension of crazed material crazingg, i.e., the ap-
pearance of semi-opaque transversal bands in the neighbor-
hood of the break surface, is a localized molecular-orientation
phenomenon that occurs when some glassy polymers are close
to rupture).
Note that the measured values of the deflexion at break are
not reproducible and should not be considered. In the many
tests performed, some discrepancies were obtained for this
parameter. This was probably due to sample heterogeneity.
The force-deflexion curves measured at lower deflexions were
always quite reproducible.
When the test is repeated at room temperature, but at a
lower deflexion rate (2 mm/min), the material's stiffness de-
creases, coincidentally giving a curve similar to the one pre-
viously obtained at 50 C. This is a good illustration of
time/temperature equivalence. In polymeric materials, mo-
lecular response is highly time and temperature dependent.

CONCLUSION
The in-house-built tensile-testing machine proved to be an
economical tool for allowing students to test the mechanical
behavior of different polymeric materials. The results can be
analyzed both quantitatively and qualitatively. The fact that
the test specimens can be obtained from everyday-use mate-
rials is not only an economic advantage but also an added
factor of interest for students, since they can do the material
selection and preparation themselves. The pedagogical ben-
efits obtained from direct experimentation were confirmed
by the interest and motivation shown by our students. Aware-
ness that the machine was built in-house actually seemed to
raise the students' curiosity toward comparing its components
to the ones used in the commercial models.
The results shown here are representative of some impor-
tant aspects of the mechanical behavior of solid polymers,
such as the influence of processing conditions and the effects
of temperature or strain rate. Students are encouraged to
analyze their results and provide interpretations on a mo-
lecular level.
Other materials used successfully with this machine include
rubber bands and films of elastomeric wall coating (EWC).
The latter constitute a type of paint that, due its elastomeric
character, is able to protect cracked walls from rain damage,
since the film stretches to keep the gaps covered. It is an en-
riching exercise to analyze the mechanical response of EWC
films under conditions such as low temperatures, aging (UV
degradation), or water exposure (Ipl.,I i i/.il i i
The tests performed with this device can be used either as
an illustration of the concepts and phenomena previously dis-
cussed in class or, perhaps more interestingly, in a reversed
approach. Indeed, the experimental observations are quite


thought-provoking and motivate students to ponder and hy-
pothesize on the reasons for the results obtained-thus pav-
ing the way for a structured discussion led by the instructor.
It must be remarked that the tensile and tearing tests de-
scribed here for polymeric films are actually similar to some
of the standard industrial practices, both in terms of speci-
men dimensions and operation parameters. It can be noted,
as an example, that a data sheet for blown film obtained from
ExxonMobil's HDPE resin HTA 001HD110' (recommended
for shopping bags, among other uses) reports an elongation
at break of about 380% and a tensile strength (stress at break)
of 56 MPa (it does not exhibit anisotropy for the particular
processing conditions employed); this result is quite consis-
tent with the values obtained with our shopping-bag material
(of unknown origin).
In addition, a fairly good reproducibility is obtained with
our in-house-built machine. The only problems are usually
associated with discrepancies in the rupture points. For stress-
strain tests, this is due to premature breaking caused by tears
that initiated at imperfections in the specimen side cuttings,
as mentioned before. These anomalous rupture behaviors can
be easily detected visually and can be minimized by cutting
the samples carefully. In the flexural tests, rupture discrepan-
cies are probably associated to imperfections or inhomoge-
neities among samples.
Naturally, the machine presents limitations when compared
to the models used industrially, namely in terms of measure-
ment accuracies and limitations on load and strain ranges.

ACKNOWLEDGMENT
The authors would like to thank the Department of Chemi-
cal Engineering of the Faculty of Engineering of the Univer-
sity of Porto for the financial support provided for assem-
bling this machine.

REFERENCES
1. Kumar, A., and R.K. Gupta, Fundamentals of Polymers, McGraw-
Hill International Editions, New York, 376 (1998)
2. Sperling, L.H., Introduction to Physical Polymer Science, 3rd Ed.,
Wiley-Interscience, New York, 477 (2001)
3. Walker, J., "The Amateur Scientist," Scientific American, February,
86 (1990)
4. ASTM D882-02, Standard Test Methodfor Tensile Properties of Thin
Plastic %STM, Philadelphia
5. ASTM D2370-98(2002), Standard TestMethodfor Tensile Properties
of Organic Coatings, ASTM, Philadelphia
6. Rosato, D.V., Extruding Plastic A Practical Processing Handbook,
Springer-Verlag, New York, 315 (1998)
7. Crawford, R.J., Plastics Engineering, 3rd Ed., Elsevier, Amsterdam,
265 (1998)
8. Kumar, A., and R.K. Gupta, Fundamentals of Polymers, McGraw-
Hill International Editions, New York, 340 (1998)
9. Zhang, X.M., S. Elkoun, and M.A. Huneault, "Oriented Structure and
Anisotropy Properties of Polymer Blown Films: HDPE, LLDPE and
LDPE," Polymer, 45, 217, (2004)
10. O


Chemical Engineering Education











Partnering With Industry
Continued from page 37

Instructor comment: I suspect the three students who rated
this query as poor may have been reflecting on how valuable
they felt their work was to Wacker and Air Products. This
perception is expressed in the student comment #3 below.
Three students provided personal comments of the indus-
try project on the evaluation form:
A Comment #1: "I think the project would have gone
better if we were able to run the equipment and take
the samples ourselves."
A Comment #2: "It would be beneficial to our under
standing of particle technology ifwe were allowed a
more hands-on approach rather than analyzing
given data."
Instructor comment: I believe these two students were re-
ferring to collecting samples from the process, as all students
were required to run the Beckman Coulter Counter.
A Comment #3: "I it was neat to see an actual
application of particles, but I didn't feel we actually
accomplished
Comments from Industry Participants on evaluation results:
S"I thefeedback from the students was
interesting and very candid. The students that rated
the exercise as fair to poor shouldn't be viewed
negatively, but rather that their engineering interests
might lie in marketing, sales, or areas other than
manufacturing."
A "Many students saw this project as a research study
or 'make-work' study with no commercial application
or contribution to a company's profit. When we started
to connect the dots to commercial applications during
the presentations and relate to benefits for the
company, many students felt better about the project
and started to appreciate their contributions."

CONCLUSIONS AND RECOMMENDATIONS
It is obvious from the feedback that certain students were
frustrated with the amount of contact they had with the pro-
cess, and didn't perceive any benefit to the company from
the projects.
Benefits weren't discussed until the end of the projects, in
the presentation phase, which, in retrospect, was too late. In
the future, it would be better to introduce benefits earlier in
the execution of the projects. This might be best accomplished
by having the industry personnel visit the classroom and in-
troduce projects themselves, including potential benefits for
the company. The students, however, should also be made to
realize that these projects are chosen partially for the benefit
of the industry, but the main driving force is to provide the
students with a real-world learning experience.
Winter 2006


"It would be beneficial

to our understanding

of particle technology if we were

allowed a more hands-on approach

rather than analyzing given data."

Student feedback



Two of the biggest challenges of this exercise were: (1)
finding industry projects that could feasibly be completed by
the students in the project time frame, and (2) finding three
projects requiring a comparable quality of student experience.
As is obvious from these three projects, one resulted in a
better student experience than others. In Project 3, the stu-
dents had more project participation since they were able to
plan and run experiments using the Ro-Tap machine, as well
as run the particle analyzer. Projects 1 and 2, on the other
hand, were straightforward as far as obtaining samples, which
were collected by industry personnel, and the students' only
participation in data collection was running the particle ana-
lyzer on the samples. In the future, this deficiency could be
overcome by suggesting students shadow the industry par-
ticipants during procedures that they can't perform themselves
due to safety and liability issues. Also, more pre-planning by
the instructor to assure better equity of the project experi-
ence may be necessary (initiation of partnership occurred in
July, with the course beginning in August).
By the very nature of the projects being based on unanswered
questions about the process, however, it would be impossible
to predict project results and effects in this scenario.
Overall, the majority of the students felt the industry project
was beneficial to their careers and experience. The project
accomplished the main goals of (1) exposing students to a
real-life particle manufacturing process, (2) gaining hands-
on experience running a state-of-the-art particle measur-
ing device, and (3) applying the basic concepts presented
in the course.

ACKNOWLEDGMENT
We wish to acknowledge the support given by Josh Brien,
Wacker engineer, in assisting students with data collection.

REFERENCES
1. State of Kentucky Title V Permit No. V-99-057 for Wacker Polymer
Systems Spray Dryer Plant at Calvert City.
2. Wacker Polymer Systems, VINNAPAS: "Redispersible Powders and
Dispersions Product Brochure," Nr. 5838-5838 (USA) 04 (2001)
3. Wacker Polymer Systems, Air-Flow Model for Spray Dryer Process,
Burghausen, Germany (not formally published).
4. Rhodes, Martin, Introduction to Particle Technology, John Wiley and
Sons, West Sussex, England (1998) [











[fj^ laboratory


A NONLINEAR,

MULTI-INPUT, MULTI-OUTPUT

Process Control Laboratory Experiment




BRENT R. YOUNG, JAMES H. VAN DER LEE, AND WILLIAM Y. SVRCEK
University of Calgary Calgary, Alberta T2N IN4, Canada


he laboratory course in process control constitutes an
important component of an undergraduate chemical
engineer's education because it provides hands-on
training in the application of process control to real processes.
The laboratory course exposes the student to industrial process
control hardware and the impact of measurement noise and un-
measured disturbances upon the control of real processes.

Brent Young is a senior lecturer of chemical
and materials engineering at the University of
Auckland, New Zealand, and an adjunct as-
sociate professor at the University of Calgary,
Alberta, Canada. He received his B.E. (1986)
and Ph.D. (1993) degrees in chemical andpro-
cess engineering from the University of Can-
terbury, New Zealand. Dr. Young's teaching
and research interests center on process con-
trol and design. He is a registeredprofessional
engineer and is actively involved in applied
research.
James van der Lee is a software engineer
with Virtual Materials Group, Inc., Calgary,
Alberta, Canada. He received his B. Sc. de-
gree in chemical and petroleum engineer-
ing from the University of Calgary in 1999
and successfully defended his Ph.D. the-
sis in 2004. He was instrumental in the
design of the new laboratory while a
graduate student.


William Svrcekis a full professorof chemical "
and petroleum engineering at the University
of Calgary, Alberta, Canada. He received his
B.Sc. (1962) and Ph.D. (1967) degrees in
chemical engineering from the University of
Alberta, Edmonton, Canada. Dr. Svrcek's
teaching and research interests center on pro-
cess control and design. He is a registeredpro-
fessional engineer in Alberta and Ontario and
is actively involved in applied research.

Copyright ChE Division of ASEE 2006


In most university courses these laboratories are essentially
linear single-input, single-output (SISO) unit operations. Until
recently, the Department of Chemical and Petroleum Engi-
neering at the University of Calgary was no exception. Yet
such SISO control laboratories do not expose the student to
the complexities of nonlinear or multi-input, multi-output
(MIMO) processes.
A few laboratories in the hlc'i.iiic' "~ have attempted to
address these shortcomings. Rivera at Arizona State Univer-
sityM1l describes a salt-mixing laboratory that examines the
concentration dynamics at different tank levels using system
identification techniques in a first process dynamics and con-
trol course. Fisher and Shah at the University of Alberta[21
describe a complex three-tank-level plus temperature arrange-
ment that allows MIMO processes and process nonlinearity
to be studied at the senior undergraduate or first-year-gradu-
ate course level. Braatz, et al., at the University of Illinois3, 4]
describe a nonlinear but SISO pH neutralization process and
a quadruple-tank apparatus that illustrates time-varying dy-
namics for a senior undergraduate process control course.
In this paper we describe a relatively simple salt-mixing
laboratory in the undergraduate chemical engineering pro-
cess control course at the University of Calgary that allows
students to study both MIMO behavior and nonlinearity.

THE UNIVERSITY OF CALGARY'S
PROCESS CONTROL COURSE
The University of Calgary requires process dynamics and
control as part of the degree requirements for undergraduate
students in chemical engineering, in a course that pioneered
the hands-on, real-time (time domain) approach to teaching
process dynamics and control."5 Students in the class employ
dynamic process simulation using a dynamic process simu-
lator, such as HYSYS or Aspen Dynamics,[61 to model chemi-
Chemical Engineering Education










cal process plants and their control systems. The students then
create "disturbances" in the plant, which may involve changes
in feed composition, flow, system temperatures, and/or pres-
sures. The simulator demonstrates in real time what the ef-
fects of these "disturbances" would be on the plant opera-
tion, and it allows the student to evaluate the strengths and
weaknesses of a given process control scheme.
The course is accompanied by a textbook written by the
course instructors, A Real-Time Approach to Process Con-
trol.[7 The text has 10 chapters, each of which focuses on a
given aspect of process dynamics and control, whether it be
investigating the concepts of process gain, time constants,
and deadtimes, studying control schemes for distillation col-
umns, or examining plant-wide control. Associated with the
chapters are eight workshops[8l that are to be completed by
the student using a dynamic simulator. Each individual work-
shop explores the concepts explained in the associated chapter,
allowing students to assign meaning to the words.
Due to the electronic nature of the workshops, hands-on,
real-time experiments on laboratory unit operations equip-
ment were considered a necessity to further reinforce the prac-
tical approach of the textbook. As a consequence, there is a
compulsory laboratory component to the course.

LABORATORY OVERVIEW
The laboratory component of the process dynamics and
control course includes two traditional experiments: (1) a


three-tank cascade where simple process identification and
level control are the objectives, and (2) a double-pipe heat
exchanger with a variable deadtime leg which can be config-
ured to investigate feedback, cascade, and feedforward con-
trol. While these experiments offer students the chance to
experience the effects of process/measurement noise and
unmeasured disturbances, the behavior of the experiments
is essentially linear, and the control loop studied is SISO
in structure.

SALT-MIXING LAB EXPERIMENT
The salt-mixing lab experiment that incorporates
nonlinearity and MIMO behavior was designed in 2002 for
immediate introduction into the curricula.
Figure 1 is a schematic of the laboratory process experi-
ment. The following is a description of what occurs in the
process:
A concentrated salt solution is mixed and stored in a large
holding tank that was sized to give a five-hour or more run
time. This solution is pumped into the conical mixing tank,
passing a magnetic flow meter and flow-control
valve, which are used to regulate flow via a flow-control
loop. Fresh water is supplied via building utilities; the
water passes a magnetic flow meter and control
valve that are used in a flow-control loop to regulate the
fresh-water flowrate. Upon entering the mixing tank the
fresh- and saltwater streams are blended using a stirrer. The
conical section mixing tank provides a strong process


Figure 1. A
schematic of the
Salt-Mixing
Laboratory
Process.


Winter 2006


ENCH 529
SALT MIXING LAB PROCESS SCHEMATIC











nonlinearity. The level in the mixing tank is measured using
a pressure cell. The blended solution enters a
pump, is pressurized, and then moves to a pipe segment that
allows for one flow paths of larger tube diameter to
be selected. This setup allows one deadtimes to be
examined. The stream will then pass ' a conductivity
cell transmitter, which is used as the input to the master
conductivity control loop. This loop's output is a cascaded
setpoint to the slave fresh-water flow controller. Before
going to drain, the stream passes a control valve
that is manipulated in order to regulate the level in the
mixing tank. The flowrate, level, and conductivity inputs are
allied to the DCS system, as are the fresh-water, saltwater
and level-control-valve-manipulated variables for this
MIMO system. The input and manipulated variables are


Figure 2. A photograph
of the Salt-Mixing
Laboratory.


Figure 3. A
screen-shot from
theDeltaVDCS.


used within the DCS system with predefined function blocks
to create the appropriate control loops.

Figure 2 shows the salt-mixing laboratory skid. The instru-
mentation, tank pumps, and additional parts were purchased
from suppliers but the construction of the skid and commis-
sioning of the equipment was completed in-house with the
help of university support staff. This resulted in a compact
unit that has capacity for expansion and is completely por-
table, allowing for more efficient use of laboratory space.
Figure 3 is a screen shot from the Emerson DeltaV distrib-
uted-control system (DCS) that is used for process data ac-
quisition, monitoring, and control in the laboratory. The ad-
vantage of using a DCS is that they are common to modern
industrial installations; as such, undergraduate engineering
students should be taught what a DCS looks like as well as
be provided with experience in controlling processes using
such graphical interfaces.
Other laboratories in the liici.ic'' )-11] have also realized
this necessity and addressed it in different ways. Rivera, et
al.,[1 also employed an industrial DCS (Honeywell, in that
case), as did Skliar, et al., at the University of Utah[91 in a
graduate course also open to seniors (Opto 22, in the latter
work). The approach of Bequette, et al., at Rensselaer Poly-
technic Institute[101 was perhaps the more typical use of
Matlab/Simulink block diagrams as an interface to simulated
experiments. Braatz, et al.,[1] employed the Hewlett Packard
Visual Engineering Environment (HPVEE) to construct their
student-operator interfaces to have a similar look and feel to
an industrial DCS.


LW ZL
Mo.D -1 ,, I L,.L I Ma.n r I T..1,- 21 UO nrame z- l.lIr, T, T._,- I ] 2i, Pl


I I I lII







- rI .'. i '. h 11 1





,' F..,.. -. ..
r i -

S .: :. r, ,r, i.:. I n i i, ,


Ii iI n,1 U tn,





Chemical Ergineering Education












The overall mass and species balance equations that de-
scribe the dynamics of the system are included in Table 1,
and the system nonlinearities are delineated and linearized in
Table 2 so that the nonlinearities are clearer to non-control
experts who have been assigned to teach process control. Fig-
ure 4 gives a time-domain plot from the DCS showing sys-


TABLE 1
Overall Mass and Species Balance Equations


Overall Mass Balance Equation FFresh Water +FSaltwater F
(Assuming constant density and isothermal)

Salt Species Balance Equation d(d
(Assuming constant density and isothermal) FSaltwater x FProduct "Y




TABLE 2
System Nonlinearities

Nonlinearity Nonlinear Characteristic Lineal

Volume change with level V =-tan2 0.h V=t
in the conical section 3

Product flowrate change Fproduct = K rdu
with the level due to the valve productdu

Multiplicative nonlinearity
between the volume and d(Vy)
the salt concentration dt


R =l~~mB~


Di B << a (LJ)< L-,I so l l .:4 7- 1 : ,
Ench 529 salt mixing Lab
250 0 2.50 200 200
60 60
225 225
55 55 175 17 -
50 200 200 50
150- 150-
45- 45
1 75 1 75 -
404 12 40
40- 4 0 12.5 125 -
5 150 -

125 1-25 100- 1
30 -30 -
S- 1 100 2 5 7-

075 075 75
2 0 -2 0 20
0 75 0J75
5 15 50
1 50 0.50
10 10-
02 025-5- 2- 5
25 25 0
0 -05 0 5

00- 000- 00- 00
1330 1400 1430 1
30 Thu Oct 2003


CONDINDCONT1D1 SPCV Maer PID controlloop 699073 103003 408 49
CONDINDCONT1PID1 OUT CV Master PID control loop 6 Lnn 10/3003 4 08 49
SALTFICPIDI PV CV Control Module 0478776 Lhnin 1030/03 40849
SALTFICPID1 SP CV Control Module 05 10/3003 40849
SWATERFICPIDPV CV Control Module 593092 Lhmn 10/3003 408 49


tem response to saltwater flowrate changes from 0 to 0.5 then
to 1.0 L/min (plus a few more). The effective tank-time con-
stant varies with the flow.


LABORATORY TASKS

Myriad tasks can be done with the aforemen-
tioned apparatus. The purpose of this labora-
tory portion of the course is to allow students
the opportunity to evaluate a variety of control
schemes. To initiate this with the mixing-tank
dV
product =- experiment, students set a tank level and then
dt perform three step tests, where each step test is
/y) either an increase or a decrease from a nominal
t value. Tuning parameters (PI) are then calcu-
lated from the resulting process-reaction curves,
using the students' choice of method
(Cohen-Coon, Ziegler-Nichols, or
IMC open-loop rules). The calculated
tuning parameters are then compared
rized Characteristic with the tuning parameters obtained
using the DeltaV automated tuning
an2.h. h program (DeltaV tune), and both sets

are tested by making setpoint changes

K, or disturbances in the saltwater flow-
ct 2 rate. The "best" set of tuning param-

eters is then chosen based on visual ob-

dy dV servations of the system response, in-
Vdt Ydt cluding time to steady state, for each
set of tuning parameters. With the best












44,. Figure 4. A time-
I domain plot from the
DCS demonstrating
I the system response
,--' J1'I'- ,'-,- to saltwater flowrate
changes from 0 to 0.5
then to 1.0 L/min
(plus a few more).




500 1530 1600








I NUM Eveh BIRlBTROLChorlcb Hst- u RBTROL


Winter 2006


I





















The simulator

demonstrates

in real time

what the

effects of...

"disturbances"

would be on

the plant

operation,

and it allows

the student to

evaluate the

strengths and

weaknesses

of a given

process

control

scheme.


tuning parameters entered into the system, the level in the mixing tank is then changed signifi-
cantly, for example from 65% to 35%, which would mean moving from the cylindrical (linear)
to the conical (nonlinear) section of the mixing tank or vice versa. Setpoint changes) are
then made in order to allow students to examine the process response.
The students are then asked to perform a full analysis of the process behavior in both open and
closed loop, including comments on linearity, order of response, and possible better control
strategies for the apparatus. As well, the students are given an additional open-ended problem:
to calculate the amount of salt initially added to the storage tank. The information given to the
students to complete these tasks includes printouts of process data (e.g., flowrates, conductivity)
and the initial height of water in the storage tank. Students are also able to measure the tank
dimensions if they so desire.

EVALUATION
Along with an analysis of the process behavior, the students were asked to provide some
general comments on the laboratory. Overall, the laboratory was found to provide good expo-
sure to the latest process equipment, along with demonstrating different tuning methods (includ-
ing those done using the built-in autotuner). Students were able to recognize the nonlinearity in
the system and provide an explanation, as well as provide explanations for the changes in time
constant and deadtime with different flowrates. System noise was well demonstrated in this
laboratory and its effect on the graphical method for calculating tuning parameters was noted.
As well, the effect of capacity was seen. Many students also attempted the open-ended prob-
lem-to calculate the initial mass of salt-and used a number of approaches in attempts to solve
it. General student comments and laboratory reports indicated that students enjoyed working
with the new laboratory experiment, and that it was helpful to see a real process that could
provide them with a feel for what types of disturbances can be made in a plant. (Whereas, in the
simulation workshops, unrealistic disturbances are quite possible and it is sometimes difficult to
measure the actual time effect a disturbance would have.)
Because it was a real process, the students did find the experiment was a little long, as it
usually ran slightly in excess of four hours (the time period scheduled for the experiment). A
smaller process could be considered, but long time constants are a reality of industrial plants and
this is an important fact for students to realize that is often somewhat overlooked in their process
control education.
In general, it was felt that the laboratory was well received by students, and that it provided
them with good exposure to state-of-the-art control hardware. The students were also exposed to
instrumentation they had not seen before, such as magnetic flow meters and conductivity cells.
The experiment also effectively displayed the difference between a simulation and a real pro-
cess, in that it took up to 30 minutes to achieve steady state in closed loop, depending on the
tuning parameters and the setpoint change made. Some ways in which this "down" time could
be used more effectively include:
l Quizzes
E Lab discussions
E Tutorial support
E Additional reading material
E Increased time to explain the apparatus
These options could be used to keep the students focused on the experiment since it is felt that
what was actually going on in the process was often overlooked due to other distractions during
the time lags. Despite this, students did seem to take note of some pitfalls that can be encoun-
tered when tuning controllers, such as the errors associated with the graphical methods and the
importance of proper input design.
The experiment also reaffirmed the value of a DCS in the teaching environment. Unit opera-
tions laboratories had previously had DCS systems integrated into them, but the DCS was not
used in a control context and students did not need to make use of all of the data-collection and


Chemical Engineering Education










handling capabilities of the system. This experiment also showed a practical application of
cascade control as the fresh-water supply pressure was not regulated-therefore changes in the
water system would propagate through the system but would be quickly compensated for by the
slave fresh-water flow-control loop that is manipulated by the master-conductivity control loop.
It was felt that the bonus question worked well and that it should be made mandatory for future
labs. It was also convenient for the teaching assistants that the lab could be run differently for
each group by simply changing the initial salt concentration or flowrates. As well, this change-
ability provided the teaching assistants with an opportunity to learn more about process control.
Overall, it was thought the lab performed very well and showed much promise as well as
many other areas of potential use. For instance, it would be useful in a more advanced process
control course where it could be used to demonstrate system identification and model predictive
control in a practical setting.

CONCLUSIONS
The introduction of this new lab was successful from the students' point of view. They enjoyed
working with the latest process control instrumentation. They also gained a new appreciation of the
problems associated with real plants, in the form of noise and unexpected disturbances. The com-
parison of conventional open-loop tuning methods and an automated tuning package was appreci-
ated, as was the chance to show their creativity in the solution of the open-ended bonus question.
From the instructors' point of view, the laboratory was considered successful. The only real
concerns with the lab were based on the length of time it took to complete. This will be ad-
dressed in coming years with the introduction of quizzes and discussion while waiting for the
process to reach steady state. Despite these concerns the lab provided an effective demonstra-
tion of a nonlinear and MIMO system. Most importantly, it was felt the students were better able
to understand process behavior by being able to see many of the classroom concepts on an
actual process. The department also gained a valuable tool for additional process control courses
due to this lab's ability to have the control configuration changed, the ease in which it can be
upgraded or modified, and its extensive data-collection and data-handling capabilities.

ACKNOWLEDGMENTS
Financial support from the Calgary Engineering Endowment fund is gratefully acknowledged
for the purchase of the laboratory hardware. Bernie Then is thanked for the assembly of and
assistance in commissioning the laboratory hardware.

REFERENCES
1. Rivera, D.E., K.S. Jun, V.E. Sater, and M.K. Shetty, "Teaching Process Dynamics and Control Using an Indus-
trial-Scale, Real-Time Computing Environment," Comp. Apps. in Eng. Ed., 4(3), 191 (1996)
2. Badmus, O.O., D.G. Fisher, and S.L. Shah, "Real-time, Sensor-based Computing in the Laboratory,"(
Ed., 30(4), 280 (1996)
3. Braatz, R.D., and M.R. Johnson, "Process Control Laboratory Education Using a Graphical Operator Interface,"
Comp. Apps. in Eng. Ed., 151 (1998)
4. Rusl, E., S. Ang, and R.D. Braatz, "A Quadruple-Tank Process Control Experiment," ( 38(3), 171
(2004)
5. Svrcek, W.Y., D.P. Mahoney, and B.R. Young, "A Real-Time Approach to Process Control Education-A Para-
digm Shift," ASEE99 Conference, Charlotte, NC, June (1999)
6. Aspen Dynamics and HYSYS, Products of AspenTech Inc., and subsidiaries, Boston (2002)
7. Svrcek, W.Y., D.P. Mahoney, and B.R. Young, A Real-time Approach to Process Control, John Wiley and Sons
Ltd., Chichester, UK (2000)
8. Young, B.R., D.P. Mahoney, and W.Y. Svrcek, "Real-Time Simulation Workshops for Undergraduate Process
Control Education," Proceedings, ACE2000, 5h IFACIIEEE Symposium on Advances in Control Education, Nara,
Gold Coast QLD, Australia, December (2000)
9. Skliar, M., J.W. Price, and C.A. Tyler, "Experimental Projects in Teaching Process Control," Chem. Eng. Ed.,
32(4), 254 (1998)
10. Bequette, B.W., K.D. Schott, V. Prasad, V. Natarajan, and R.R. Rao, "Case Study Projects in an Undergraduate
Process Control Course," Chem. Eng. Ed., 32(3), 214 (1998)
11. Ang, S., and R.D. Braatz, "Experimental Projects for the Process Control Laboratory," Chem. Eng. Ed., 36(3)
(2002) 1
Winter 2006


General

student

comments and

laboratory

reports

indicated

that students

enjoyed

working with

the new

laboratory

experiment,

and that it was

helpful to see

a real process

that could

provide them

with a feelfor

what types of

disturbances

can be made

in a plant.











[fj^ class and home problems


DATA ANALYSIS MADE EASY

WITH DATAFIT



JAMES R. BRENNER
Florida Institute of Te. !r, *., \ Melbourne, FL 32901


Shortly after starting as an assistant professor, I real-
ized that quite a few of our students were unable to
analyze laboratory data at a level consistent with that
expected when I had worked in industry. Having been put in
charge of the Florida Institute of Tc, liii. ._- 's introductory
chemical engineering course and its materials science and
engineering laboratory course, I decided that a strong em-
phasis on data analysis would be added to each of these
courses in order to satisfy ABET's requirement regarding the
ability of students to analyze data.
Most departments emphasize spreadsheet calculations and
plotting of data in Microsoft Excel as part of their introduc-
tory chemical engineering course. Experience in our depart-
ment has shown that unless sufficient time is spent on data
analysis instruction such that spreadsheet calculations, plot-
ting, and curve fitting become second nature, such skills are
either forgotten or are never learned properly.


We have incorporated DataFit from Oakdale Engineering1 ]
throughout the entire curriculum at Florida Tech, beginning
with CHE 1102, an eight-week, one-day-per-week, two-hour,
one-credit-hour, second-semester Introduction to Chemical
Engineering course in a hands-on computer classroom. The
syllabus for CHE 1102 is shown in Table 1. The examples


Copyright ChE Division of ASEE 2006
Chemical Engineering Education


The object of this column is to enhance our readers' collections of interesting and novel prob-
lems in chemical engineering. Problems of the type that can be used to motivate the student by
presenting a particular principle in class, or in a new light, or that can be assigned as a novel
home problem, are requested, as well as those that are more traditional in nature and that eluci-
date difficult concepts. Manuscripts should not exceed 14 double-spaced pages and should be
accompanied by the originals of any figures or photographs. Please submit them to Professor
James O. Wilkes (e-mail: wilkes@umich.edu), Chemical Engineering Department, University
of Michigan, Ann Arbor, MI 48109-2136.
<___________________________________________


James R. Brenner received his B.S. degree
from The University of Delaware and M.S. and
Ph.D. degrees from The University of Michi-
gan. After a postdoc at Argonne National
Laboratory and industrial experience at
Westinghouse Savannah River Company, he
became an assistant professor of chemical
engineering at Florida Institute of Technology
His research interests are in hydrogen purifi-
cation and sensing, electronic noses, and
nanoporous materials.











chosen, shown in parentheses, are selected so as to be consis-
tent with concepts that students learn concurrently in other
courses. DataFit also has become commonly used in our
Physical Chemistry Lab and Materials Science and Engineer-
ing Lab courses, as well as in several courses in other engi-
neering departments. Our experience at Florida Tech is that


students retain data
analysis concepts
best when such con-
cepts are formally
taught to them in
this short course and
then periodically re-
inforced throughout
their academic ca-
reers. Several ex-
amples covered in
weeks three through
eight will be dis-
cussed here.

An introduction
to basic statistics is
included in nearly
all introductory
ChE courses and
will not be dis-
cussed in this ar-


TABLE 1
Data Analysis Curriculum

1) Statistics and Confidence Intervals
2) Introduction to Plotting and
Calculations in Excel
3) y = ax + b Fitting in DataFit
(Pressure Transducer Calibration)
4) y = ax Requires User-Defined
Models (Hygrometer Calibration)
5) Semi-Log Functions (First-Order
Rate Laws Felder and Rousseau
2.34)
6) Plotting and Curve Fitting of
Power-Law Functions (Crystal
Growth Felder and Rousseau
2.37)
7) Nonlinear Functions (Vapor
Pressures)
8) Curve-Fitting in 3-D (Rate Laws
With Two Reactants)


Figure 1. Calibration of a Span Instruments' pressure transducer against an NIST-
traceable Paroscientific pressure transmitter.
Winter 2006


shown in Figure 1, includ-
ing error bars, I ask the stu-


Experience in our department has shown
that unless sufficient time is spent on data
analysis instruction such that spreadsheet
calculations, plotting, and curve fitting
become second nature, such skills are either
forgotten or are never learned properly.



tide. Students in CHE 1102 cover basic statistics during the
first week of the course and get constant reinforcement of these
concepts through the use of DataFit.111 The second half of CHE
1102 consists of problems that require Polymath- or Excel-
based solutions to either sets of linear and nonlinear algebraic
equations or numeric integration, as suggested by Clough.E21
All Excel and DataFit files are available at ~jbrenner/dataanalysispaperl .zip>.


SOLVING PROBLEMS WITH DATAFIT

Problem 1. Calibration of a Pressure Transducer
Following the introduction to basic statistics, the first prob-
lem that I assign students is the calibration of a 0-250 psig
Span Instruments' NTT-204 (now Millipore) pressure trans-
ducer against a 0-1000 psia Paroscientific pressure transmit-
ter. In addition to being useful for teaching students how to
make plots with error bars and determine the difference be-
tween absolute and gauge
pressures, it provides a rela-
tively simple problem for
studying linear regression
with DataFit. The repeatabil-
ity and lack of drift of
Paroscientific pressure trans-
mitters is even superior to
that of a deadweight tester
that was calibrated at
NIST.131 The repeatability of
the quartz oscillator that the
Paroscientific pressure
transmitters use is certainly
+ .05)*x + (-15.2 + .8) within the quoted 0.01% of
full-scale precision (i.e., 0.1
psia fixed error for a 1000-
psia transmitter). Span In-
S struments' pressure trans-
250 300 ducers output a signal that
ranges from 4-20 milliamps
to within 0.08 milliamps.
After having the students
nrpnar a nalnt nf the data


300


250


S200

150


100


1 50


0


y =(.998


0 50 100 150 200
Transmitter pressure (psia)











dents to copy and paste the data into DataFit, click on the
Solve Regression option, click on OK, select the y = ax + b
option, and let DataFit do the work for them. By clicking on
Results Detailed, the Fit Information output is obtained (Table
2). Included in the output are the residual sum of squares
(RSS), which is the sum of the squares of the differences
between the calculated values of Y, the pressure in psig as
determined by the pressure transducer, and the correspond-
ing experimental values. Also evaluated are the commonly
seen R2 correlation parameter as well as several more-
advanced goodness-of-fit parameters. Most importantly,
the 68%, 90%, 95%, and 99% confidence intervals are
conveniently tabulated. This is an excellent opportunity
to reinforce basic statistics, most notably the Gaussian
distribution, which is typically taught at the beginning of


98% replied correctly to a similar question during hourly
and final exams.
Once the students have realized that b is unnecessary, it is
time to teach them how to create a user-defined model in
DataFit, as y = ax is not one of the built-in models (one of
DataFit's few shortcomings). This can be done by returning
to DataFit's main menu and clicking on the Define User Model
option under the Solve menu. The user defines a Model ID,
(which I defined as "Linear, no intercept," in this case). The
user also inputs the Model Definition, in this case Y = a*x.
Mathematical functions in DataFit, such as multiplication and
exponentiation, work in the same way as Excel.

In many cases, including this one and all cases where the
fitting is of a linear function, initial estimates are unneces-


CHE 1102.

Problem 2. Calibration of a
Hygrometer
The second problem that I as-
sign is Problem 2.32 from Felder
and Rousseau's textbook.[4] This
problem involves the correlation
of a signal from a hygrometer
versus the mass fraction of water
in the inlet stream to the hygrom-
eter. For this problem, first ask
students to do the y = ax + b fit as
described in the previous section.
The 95% confidence intervals on
the slope, a, and the intercept, b,
are as follows: a = 470 + 20; b =
0 + 2, at the appropriate number
of significant figures (proper use
of significant figures is an ex-
tremely difficult concept to get
students to consistently apply).
Then ask them whether the inter-
cept, b, is mathematically signifi-
cant (i.e., nonzero within the 95%
confidence interval). They should
answer that b is not mathemati-
cally significant at the 95% con-
fidence level. Out of a sample of
100 students asked over the last
five years as part of an in-class
exercise, only 50% have an-
swered correctly to this question;
25% of students replied "don't
know." This is a surprisingly dif-
ficult concept to master that re-
quires consistent reinforcement
throughout CHE 1102. Yet, of
the same sample of students,
62


TABLE 2
Fit Information for Pressure Transducer Calibration

DataFit version 6.1.10 Sum of Residuals = 4.08562073062058E-14
Results from project Average Residual = 3.14278517740044E-15
"F:\brenner\datafitpcalib.dft" Residual Sum of Squares (Absolute) = 5.20799168906741
Equation ID: a*x+b
Equation ID: a*x+b Residual Sum of Squares (Relative)= 5.20799168906741
Number of observations = 13 Standard Error of the Estimate = 0.688079784556427
Number of missing observations = 0 Coefficient of Multiple Determination (RA2) = 0.99994096
Solver type: Nonlinear
Nonlinear iteration limit = 2000 Proportion of Variance Explained = 99.994096%
Diverging nonlinear iteration limit =10 Adjusted coefficient of multiple determination
Number of nonlinear iterations performed = 1 (Ra2) = 0.9999355927
Residual tolerance = 0.0000000001 Durbin-Watson statistic = 2.88469613789683

Regression Variable Results
Variable Value Standard Error t-ratio Prob(t)
a 0.998001779 0.002312177 431.6287071 0
b -15.1762779 0.359917526 -42.1659876 0

68% Confidence Intervals
Variable Value 68% (+/-) Lower Limit Upper Limit
a 0.998001779 0.002408132 0.995593648 1.000409911
b -15.1762779 0.374854103 -15.551132 -14.8014238

90% Confidence Intervals
Variable Value 90% (+/-) Lower Limit Upper Limit
a 0.998001779 0.004152438 0.993849342 1.002154217
b -15.1762779 0.646375884 -15.8226538 -14.529902

95% Confidence Intervals
Variable Value 95% (+/-) Lower Limit Upper Limit
a 0.998001779 0.005089101 0.992912679 1.00309088
b -15.1762779 0.792178474 -15.9684564 -14.3840995

99% Confidence Intervals
Variable Value 99% (+/-) Lower Limit Upper Limit
a 0.998001779 0.007181158 0.990820622 1.005182937
b -15.1762779 1.117831851 -16.2941098 -14.0584461

Variance Analysis
Source DF Sum of Square Mean Square F Ratio Prob(F)
Regression 1 88206.02278 88206.02278 186303.3408 0
Error 11 5.207991689 0.47345379
Total 12 88211.23077

Chemical Engineering Education











sary, but they become critical when doing some nonlinear
fitting. The default values of each of the curve-fit parameters
are unity in all cases. I look at this as one of DataFit's very
few design flaws. When one goes through a Taylor series
expansion, terms involving higher-order parameters are sup-
posed to be corrections to the previous terms, meaning that
the product of the curve-fit coefficient multiplying a high-
order term and that higher-order term (i.e., d*x3) should be
less than those of previous terms. Without some exceptional
physical justification, it would be difficult to throw out con-
stant, linear, or quadratic terms and keep a cubic term.

After manually assigning initial estimates and/or constraints
on the curve-fit coefficients, clicking OK, clicking Solve Re-
gression, and OK again, the user will need to locate his or her
user-defined model in the list of models. After locating your
recently defined model, click on Solve, click OK, and then
click on Results Detailed to return to the Fit Information
screen once again. The models are ranked by the RSS, and so
the Fit Information that pops up first is the one with the low-
est RSS, not the one for the most recent fit. By clicking on
the uppermost dialog box to locate the user-defined model,
one will get the Fit Information associated with the user-de-
fined model, "Linear, no intercept." Interestingly, scrolling down
to the 95% confidence interval shows that the confidence inter-
val for the one-parameter model (a = 473 8) is narrower than
the slope from the two-parameter model (a = 470 + 20).


Problem 3. Fitting Water Vapor Pressures to the
Clausius-Clapeyron and Antoine Equations
Fitting water vapor-pressure data to the Clausius-Clapeyron
equation is challenging for underclassmen, but usually can
be done successfully if the previous examples have been
worked out in class or for homework. This problem, along
with the follow-up fitting of the same data to the Antoine
equation, typically is either the final in-class or homework
problem that students are asked to solve during CHE 1102.
Data for the vapor pressure of water is tabulated in Appendix
B.3 of Felder and Rousseau.41 The Clausius-Clapeyron equa-
tion is as follows, and requires conversion of temperatures
into Kelvin:


B
logloP= A-
T


At this point in the course, the students know that they
should plot pressure on a logarithmic scale on the y-axis and
reciprocal temperature on the x-axis. Students are asked to
plot 1000/T so that the values on the x-axis are between a
more aesthetically pleasing 0 and 10, to estimate the slope (-
B) and the intercept (A) graphically, to use DataFit to deter-
mine A and B, and finally to superimpose the curve fit (the
solid line) on top of the experimental points (Figure 2).
The Clausius-Clapeyron equation is a reasonably good fit














Figure 2.
Clausius-
Clapeyron plot
for water vapor
pressures.[4


2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8

1000/Temperature (KI1)



Winter 2006


1000


X
1 100

i2
I-^


C
sa
1 10

I-











of the vapor pressure of water data from 0 to 60 C, but one
can see that there is a systematic deviation from linearity at
low temperature and pressure. By graphically extrapolating
a straight line through the portion of the data that appears to
be linear, one can estimate the slope (-B) as -2200 and the
intercept (A) as 109 from Figure 2. Interestingly, there are
slight differences in the DataFit estimates of the curve fit pa-
rameters, depending on whether the logarithm of the pres-
sure data and the inversion of the temperature data are taken
before curve fitting in DataFit or not (Table 3). In the case
where the data are not so linearized before entry into
DataFit and then a nonlinear model is generated in DataFit,
the points at low vapor pressures are de-emphasized rela-
tive to the other points.
If one tries to fit the Antoine equation for water vapor pres-


TABLE 3
Clausius-Clapeyron Constants for Vapor Pressure of Water
from 0 to 60 "C

Clausius-Clapeyron Linear Fit of Nonlinear Fit
Constants Linearized Data of Raw Data

A 9.091+ 0.004 9.003 + 0.004
B 2301+ 1 2274+ 1



'rA fl 1 AI


sures either below 60 C or above 60 C, in either case if one
does not manually change the default parameter guesses of
unity, DataFit's .. IIIII. n" will require more iterations than
the default number of iterations, which is 250.

logl P=A- (2)
(T +C)
This problem can be changed using Edit Preferences. I have
changed the default number of iterations permanently to 2,000.
The problem with using the results for A and B from the
Clausius-Clapeyron equation as initial guesses for A and B
for the Antoine equation fit is that the Antoine equation re-
quires temperatures to be in degrees Celsius instead of in
Kelvin. In fact, if one uses the Clausius-Clapeyron equation
constants to fit the water-vapor pressures above 600C and
lets DataFit set the default value of C to 1, then
even after having made the appropriate conversion
of the data from Kelvin into Celsius, DataFit will
erroneously return a "successful" result after only
one iteration that contains errors larger than the val-
ues of the parameters themselves. The Antoine
of equation cannot be solved for temperature ranges
in which the denominator, (T+C), switches from
negative to positive over the range of temperatures.
If one uses the values of A and B from the Clausius-
Clapeyron equation and an initial guess for C of
273.15, then the Antoine equation does converge
properly to the answers below in Table 4 in the
"Proper Convergence" column.


le
th
A
is
e

A


This discrepancy proved a difficult chal-
nge for even the best students. At a minimum,
ie number of significant figures reported for
ntoine equation constants in the literature4, 5]
grossly overstated, and, for some mol-
cules, is just not quite right (see Table 5).

ASSESSMENT
In the first class exposed to this curricu-
lum, 17 of 20 students successfully com-
pleted both the Clausius-Clapeyron and
Antoine problems. Two of the three stu-
dents who failed to make a proper plot
and a proper fit in DataFit attended class
less than one-third of the time, and the
other student, although in good atten-
dance, turned in less than half of the
homework assignments and had signifi-
cant language problems. The past four
years of classes have had similar results.
A similar problem, for butane vapor
pressures, has been assigned to sopho-
mores and graduate students, using data
from the NIST Chemistry WebBook.[121
Chemical Engineering Education


Antoine Curve Fitting of Vapor Pressure of Water from 0 to 60 C


Constants 250 iterations Proper Convergence Literature Data[4'5]

A 6.95+ 0.08 8.124 + 0.002 8.10765
B 1180 +40 1759.8+ 0.6 1750.286
C 186 + 4 235.8+ 0.1 235.000


TABLE 5
Clausius-Clapeyron Equation Parameters*


Molecule AL BL AN BN

Carbon Dioxide 7.58 + 0.02 865 + 4 7.58 + 0.01 864 + 3
Ethane 7.37+ 0.05 837 + 9 7.127 + 0.008 785 + 2
Propane 7.71+0.08 1130+14 7.191 + 0.007 1128+3
Isobutane 7.69+ 0.07 1274+ 16 7.198 + 0.007 996+2
Butane 7.61 + 0.06 1306 +7 7.256 + 0.009 1193 + 4


*Pressures in mm Hg and temperatures in Kelvin
LLogarithm of pressure taken first
NLogarithm of pressure not taken first











All but one of 12 sampled students who came to Florida Tech
from other countries for ChE graduate school sought me out
for help. None of the eight students that went to Florida Tech
for both bachelor's and master's degrees needed help. Ninety
percent of sophomore students who took CHE 1102 as fresh-
men were also able to solve the butane problem successfully.
With the default guesses, DataFit failed to converge be-
cause it cannot handle the denominator changing from nega-
tive to positive, depending on temperature. When the second
term exceeds A, the solution also diverges. Under some sets
of initial estimates, DataFit "converges" to a flat line! When
the initial estimates are reasonably close to what DataFit re-
ports as the correct answer (A = 7.44 + 0.04; B = 1330 + 30;
C = 294 4), the solution converges to what is shown in
Figure 3. Even this is clearly incorrect, as the low vapor pres-
sure data is de-emphasized, because the magnitude of the er-
ror in such a small quantity is dwarfed by a small percentage
error in the high vapor pressure points. This kind of error is not
unique to DataFit. I have seen it in Polymath curve fits as well.

CONCLUSIONS
Of the international graduate students asked to fit vapor-
pressure data for the previous problem, none had previous
exposure to either Polymath or DataFit. While each of them
also learned how to use Polymath in graduate school, 11 of


the 12 polled said that they found DataFit easier to use. The
reason that I downloaded DataFit in the first place was not
because of its excellent curve-fitting capabilities, but because
when I first started using it in industry in 1998, DataFit was
the only program that did proper 3-D scientific plotting for
less than $500. In 1999, when Florida Tech bought a site li-
cense for DataFit version 6.1, it cost only $750 for the entire
campus (albeit a relatively small campus), whereas a single
copy cost $100. Moreover, the site license allowed for stu-
dents and faculty to use DataFit at home as long as they were
doing academic work. A comprehensive set of solutions to
similar problems can be found at dataanalysispaperl .zip>.

REFERENCES
1. Gilmore, J., DataFit, v 6.1, Oakdale Engineering, 23 Tomey Road,
Oakdale, PA 15071, (724) 693-0320, sales@curvefitting.com, /www.oakdaleengr.com>
2. Clough, D., "Spreadsheets Across the Curriculum," ASEE Summer
School for ChE Faculty, July (2001)
3. Brenner, J.R., and E.F Dyer, Westinghouse Savannah River Company,
unpublished results, December (1997)
4. Felder, R.M., and R.W. Rousseau, ,' Principles of Chemi-
cal Processes, John Wiley and Sons, 3rd Ed., New York (2000)
5. Dean, J.A., Lange's Handbook of( McGraw-Hill Compa-
nies, Inc., 14th Ed., New York (1992)
6. National Institute of Standards, NIST Chemistry WebBook, webbook.nist.gov/chemistry> O


1UUUUU


10000


1000


o 100-















-150 -100 -50 0 50 100 150 200
,- 10
0


1
a-
I-
a 0.1


0.01


0.001
-150 -100 -50 0 50 100 150 200

Temperature (Degrees Celsius)


Figure 3. Antoine fit of butane vapor pressure data clearly shows bias against low vapor pressure points.

Wirrte 2006











classroom


ENGINEERING ANALYSIS

IN THE CHEM-E-CAR COMPETITION








RANDY S. LEWIS, ALIAKBAR MOSHFEGHIAN, AND SUNDARARAJAN V. MADIHALLY
Oklahoma State University Stillwater, OK 74078

ince 1999, Chemical Engineering undergraduate stu- Randy S. Lewis is a professor of chemical
dents have had the opportunity to participate in the engineering at Brigham Young University and
t te rgiol ad ntiol an adjunct professor of chemical engineering
Chem-E-Car Competition at the regional and national at Oklahoma State University. He received his
level under the direction of the American Institute of Chemi- B. S. and Ph.D. degrees in chemical engineer-
cal Engineers (AIChE). The competition was initiated by ing from Brigham Young University and Mas
sachusetts Institute of Technology, respectively.
AIChE members to (1) provide an opportunity for students He currently serves as past-chair of the Ca-
to participate in a team competition at the national level, (2) reer and Education Operating Council of
AIChE. His research interests include
encourage professional society interaction, and (3) increase biomaterials development and the utilization of
the awareness of chemical engineering in the public.[11 Ex- renewable resources for the production of
amples of national competitions in other engineering disci- chemcals.
plines include the concrete canoe race (civil engineering), mini-
baja race (mechanical engineering), and International AIAA/ Aliakbar Moshfeghian graduated with a
B.S. in chemical engineering from Okla-
ONR Design, Build, Fly contest (aerospace engineering). homa State University in 2003. He placed
second in the 2003 AIChE National Stu-
The Chem-E-Car competition involves the design and con- dent Paper Competition after winning the
struction of a chemically powered car that has to travel a speci- AIChE Mid-America Student Paper Com-
petition. Currently he is completing a
fied distance (50-100 ft) while carrying a certain amount of master's degree in chemical engineering
water (0-500 ml). The car must fit into a box no larger than .- ..- at Oklahoma State University.
40 cm X 30 cm X 18 cm and the team must be composed of
members from at least two undergraduate classes. Additional Sundararajan V Madihally is an assistant
professor of chemical engineering at Okla-
rules are applicable to the competition.[11 The objectives of homa State University. He received his B.E.
the competition are applicable to numerous ABET educational from Bangalore University and his Ph.D.
from Wayne State University, both in chemi-
outcomes including "an ability to design a system, compo- cal engineering. He held a research fellow
nent, or process to meet desired needs," "an ability to func- position at Massachusetts General Hospi-
tion on multidisciplinary teams," and "an ability to commu- tal/Harvard Medical School/Shriners Hos-
pital for Children. His research interests in-
nicate effectively."[2] clude tissue regeneration and the develop-
ment of therapies for traumatic conditions.


Chemical Engineering Education


Copyright ChE Division of ASEE 2006










To promote the competition among Oklahoma State University
(OSU) chemical engineering students and to provide an additional
design experience in the undergraduate curriculum, the competi-
tion was implemented in 2000-2001 as part of a spring sophomore
course (Introduction to Chemical Process Engineering) and a spring
junior course (Chemical Reaction Engineering). The juniors ini-
tially worked on designing the cars and were eventually joined by
the sophomores who primarily helped with the calibration, poster,
and safety aspects. The teams (six-eight students) currently com-
pete in the middle of the spring semester for the opportunity to
represent OSU at the AIChE Regional Chem-E-Car Competition.
The evolution of the competition at OSU was recently presented.[31
In brief, funding for the OSU competition was initially provided by
the department, but ChevronPhillips now provides funding for
equipment costs, T-shirts, the awards banquet, and travel to regional
and national competitions. Further, ChevronPhillips personnel pro-
vide extensive safety reviews on students' reports. Liquid effluent


Figure 1. Picture (A) and diagram (B) of the Chem-E-Car. The
left chamber was used to generate gas from a sodium bicarbon-
ate (NaHCO3) and vinegar reaction. The right chamber con-
tained water that was forced from the chamber following the
removal of the clamp. The expelled water propelled the car
forward. The parameters and values are shown in Table 1.
Winter 2006


The Chem-E-Car competition involves
the design and construction of a
chemically powered car that has to travel
a specified distance while carrying
a certain amount of water.

was allowed to discharge from the car in 2001, only
water was allowed in 2002, and no liquid discharge
has been allowed since 2003.
In 2003, an additional fall junior course (Thermody-
namics) was included in the competition to enable the
students to spend more time working on their cars.
As part of the integrated sophomore and junior team,
the students are required to write a safety and environ-
mental report, provide a detailed sketch of the car, build
a prototype, provide preliminary and final calibrations,
provide an engineering analysis, give a poster presen-
tation, and participate in the department competition.
The engineering analysis is performed solely by the
junior students, although they have traditionally pro-
vided a vague analysis such as using empirical equa-
tions, providing detailed equations without any solu-
tions, and identifying fundamental equations that may
not be applicable.
Engineering analysis is not required at the national
competition and is often not applied. Rather, students
rely on calibration data and trial and error to predict
the distance traveled by their cars.
To demonstrate and encourage the use of detailed
engineering analysis among the students in predicting
the distance traveled by a car, a model was developed
for a car (previously used in the competition) in which
pressure generated by a chemical reaction resulted in
car movement via the discharge of water. This work pre-
sents the model for predicting the travel distance based
on the initial pressure and various car parameters.
Although discharged water must now be contained
such that the model may not be applicable to current
car designs, this work provides an example of how stu-
dents can effectively apply engineering analysis. An
advantage to engineering analysis is that it allows stu-
dents an opportunity to determine the effects of design
components (e.g., vessel size, car weight, liquid vol-
ume, nozzle size, for this example) on the distance the
car travels.

MATERIALS AND METHODS
Car Design and Experimental Runs The car,
shown in Figure 1, was designed and built by Ali
Moshfeghian, Christ Schulte, and Kyle Sharon (junior
chemical engineering students at the time) and was used
in the 2002 competition at OSU. The key car param-
67












eters are provided in Table 1. The left chamber, shown in
Figure IB, was initially filled with 125 ml of a saturated aque-
ous solution of sodium bicarbonate (NaHCO3). Glacial ace-
tic acid (vinegar) was then added to the solution, causing a
chemical reaction to form CO2 that increased the chamber


pressure. The acetic acid was added according to the amounts
shown in Table 1 and, when necessary, additional water was
added so that the acetic acid/water addition equaled 10 ml.
Although acetic acid and sodium bicarbonate were used to
generate the gas pressure, any pressure-generating chemical


TABLE 1
Parameters and Values Used in Engineering Analysis


Run # Acetic Acid Initial gas Adjusted initial Distance
(ml) pressure (atm) pressure (atm) (feet)

1 2.5 4.39 2.78 4.3
2 5.0 8.14 4.76 19.3
3 7.5 11.20 6.37 32.6
4 10.0 13.24 7.43 41.7

Parameter Description Value Units Note

A Area of water chamber 11.4 cm2 Constant

Ao Nozzle area 0.087 cm2 Constant

C Head loss coefficient 0 to 0.2 unitless

g Gravitational constant 9.81 cm/s2 Constant

h Water height above nozzle

ho Nozzle height

m Mass of water 374 g Initial value

m Mass of car 2470 g Initial value

nga Moles of gas

Pgas Adjusted initial pressure See above atm Initial value

po Atmospheric pressure 1.0 atm Constant

R Gas constant 82.06 cm3atm/molK Constant

T Temperature 298 K Constant

v Car velocity 0 cm/s Initial value

V Gas volume 390 cm3 Initial value
gas

V Water volume 375 cm3 Initial value

Gas and initial3
Vto water volume 765 cm Constant

vo Water velocity Eq. (8) cm/s Variable

x Car distance 0 feet Initial value

Pliq Water density 0.997 g/cm3 Constant

P k Friction coefficient 0.07 unitless Figure 4


Chemical Engineering Education











reaction would be sufficient for operation of the car. It is im-
portant to note that the generated pressure should not exceed
the pressure limits of the materials to prevent material fail-
ure. A clamp was used to keep the gas pressure in the left
chamber until the pressure reached equilibrium. The right
chamber was filled with 400 ml of water (V h). Following
pressure equilibration, the clamp was removed and the rear
nozzle opened. The pressure above the water (Pgas), related to
the moles of gas (n g), forced the water to exit the rear nozzle
(cross sectional area of Ao) at a given velocity (v0). P0 repre-
sents the atmospheric pressure. The separator was added to
minimize foam, generated in the left chamber, from entering
the water chamber. The exiting velocity produced a thrust
that moved the car forward. When the water ran out, the car


v n
Pg,
\as



Vhq A
vbq Po90
ho ..... P0 0


Figure 2. Diagram of model
used for the engineering
analysis. The model repre-
sents the chamber that
contains the water. The
parameters and values are
shown in Table 1.


rolled to a stop and the
distance the car trav-
eled was measured. As
noted in Table 1, four
experiments were per-
formed. The experi-
ments were performed
on a smooth brick
surface.
Model Development
A model was devel-
oped to predict the dis-
tance the car would
travel based upon the
initial pressure above
the water. The model
was used as a compari-
son with the experi-
mentally measured


distance. Figure 2 shows a representation of the water cham-
ber that was used for the model. Vhq is the water volume, ho is
the height of the nozzle (assigned a value of zero), h, is the
height of the water above the nozzle, v, is the surface veloc-
ity of the water at hi, A0 is the nozzle cross-sectional area, v0
is the water velocity leaving the chamber, and P0 is the pres-
sure of the surrounding atmosphere. Pgas V ,g and ng, repre-
sent the pressure, volume, and moles of the gas above the
water, respectively.
A material balance on the total mass of the car (mc), which
is equivalent to a constant mass plus the mass of the water in
the chamber (m), shows that the mass changes with time ac-
cording to

dm car dm
dt dt =-PqVoA (1)
where Pliq is the liquid density. The right term represents the
mass flowrate of water leaving the water chamber (and the
car). Since m=Vq pliq, and if pliq is assumed constant, the
water material balance shows how V1q changes with time ac-
Winter 2006


cording to


dV ii
dVliq
dt


dV
vAo= gas
dt


The change in V is also shown with time in Eq. (2) since
any water volume decrease results in the same increase in the
gas volume (i.e., the total volume, Vtot, is constant and equal
to V gas+V hq).
gas lqa
To assess how the gas pressure (Pga) changes with time,
the ideal gas law was assumed where P sV g= P g(V -
Sga s ga s gas to
Vhq)=n asRT. Since Vtot is constant and nasRT is constant as
the water is leaving the nozzle (assuming negligible tempera-
ture change and no new gas is generated once the experiment
starts), the time derivative of the ideal gas law gives

dPgas Pgas dVliq -v0 A0
dt tot Vliq dt (Vtot Vliq (3)

Eq. (2) was substituted into the middle term of Equation 3
to obtain the term on the right.
The velocity of the car with time was predicted from a
momentum balance on the car. The momentum balance states
that the change of momentum (mass of the car, mer, times the
velocity of the car, Vc.) is equal to the sum of the forces act-
ing upon the car:


d(mcarVcar) dvcar dm, car
-mcar- +v+vcar -, =PliqVOA0- kmcarg
dt car dt car dt 0 Piq o -
(4)

The first term on the far right side of Eq. (4) represents the
thrust force that pushes the car forward.[4] Only thrust occur-
ring when water leaves the chamber was considered. Once
the water runs out, residual gas pressure greater than atmo-
spheric pressure will cause some thrust but the thrust is likely
negligible since the gas density is small compared to liquid.
The second term on the far right side represents the friction
force between the car and the ground, with P k as the friction
coefficient.[4] The negative sign signifies a force that decreases
the car velocity. The drag force between air and the car was
neglected. Substitution of Eq. (1) into Eq. (4) gives

dvcar Pliqv0Ao(v +Vcar) (5)
dt m ar

Once the water runs out of the chamber, the first term on
the right side is zero and the velocity of the car will decrease as
a result of friction until the car stops. The distance (xca) at which
the car stops was predicted from the definition of velocity,

dxcar
= vcar (6)
dt
To predict the velocity of water leaving the car (v0) for ap-
69










plication in Eqs. (1)-(5), the mechanical energy balance, 51
with the inclusion of frictional head loss due to the exit nozzle,
was utilized such that

C 2 1 2 (Pi -P) (7)
v 0+'v -vi2) = g(h h0o)+ (7)
2 2 p liq
The subscripts i and 0 refer to the values at the gas-liquid
interface and the nozzle exit, respectively. C is the head loss
constant. Since P = P s, (h-h0) = Vh/Ac (where Ac is the cross-
sectional area of the water chamber), and if v, << v (the liq-
uid velocity leaving the chamber is much faster than the ve-
locity of the water surface at the gas-liquid interface) then


vo=0 2 (iq + (8)
(1+C) Ac Pliq

Eqs. (7) and (8) are only valid when water is present in the
chamber. Thus, once the water completely runs out of the cham-
ber, Eqs. (7) and (8) no longer apply and v0 is zero in Eqs. (1)-(5).
Eqs. (1)-(3), (5), and (6) [with the definition of Eq. (8)]
were numerically integrated using Polymath[6] to obtain val-
ues of the integrated parameters as a function of time. When


60


50


40


30


20


2.78


4.76


the model results showed that the water ran out (Vhq = 0), vo
was set to zero for reasons stated above. At this point, only
Eqs. (5) and (6) were numerically integrated.
The initial values for solving the model were mc.=2470 g,
V q=375 cm3, v =0, and x =0. For the water volume, the
volume initially added to the chamber was 400 ml. Since 25
ml was below the nozzle and did not leave the chamber, the
initial water volume was modeled with a value of 375 ml.
The values of Pgas for the four experimental runs are shown in
Table 1. Since the initial gas pressure (Pini), as shown in Table
1, was measured prior to opening the clamp, the adjusted ini-
tial pressure was determined by Pga =(205/390)*Pnnit+(185/
390)* latm. The adjustment was based on the assumption that
the pressure above the water chamber (with a volume of 185
ml) was 1 atm and that the initially measured pressure (with
a volume of 205 ml) equilibrated (in the total volume of 390
ml) after the clamp was opened and prior to the opening of
the rear nozzle. A value of C=0.1 is consistent with fluid leav-
ing a large reservoir and entering a small rounded-edge en-
trance (i.e., similar to liquid leaving the chamber and enter-
ing the nozzle). 51 Table 1 summarizes the model parameters
with their associated values. Unit consistency was ensured
when solving the equations.


6.37


7.43


Initial Pgas (atm)


Figure 3. Measured (dashed lines) and predicted distance (bars with Pk = 0.069) traveled by the car
as a function of initial gas pressure (Pgs) above the water. The error bars show the predicted range
with 0.066 < lk < 0.072.
Chemical Engineering Education


0 C=O
3 C=0.1
3 C=0.2








4- -










Friction Factor Analysis The friction coefficient (Pik)
shown in Eq. (5) was needed for solving the system of differ-
ential equations. The coefficient is dependent upon the type
of surface and the type of wheels contacting the surface. Thus,
the coefficient can vary and must be measured for each sur-
face upon which a car is tested. For this work, the friction
coefficient was measured by pushing the car by hand, mea-
suring the initial car velocity (Vca,0), and then measuring the
final distance (xf) at which the car stopped from the point at
which the initial velocity was measured. The initial velocity
was measured a short distance from where the car was pushed
to ensure that the car was decelerating during the analysis. A
ruler was placed at the initial velocity measuring point while
a video camera recorded the time for the car to travel a given
distance of the ruler (5-13 inches). An average initial veloc-
ity was obtained by dividing the distance by the time.
Since there was no thrust between the initial velocity point
and when the car stopped, Eq. (5) states that dvc dt = tk g.
Integration of Eqs. (5) and (6) gives


v car
dv car
v car,O


tf
c(v ar,o
0


Pk gdt = V car Vcar, -Pkgt



2


(9)



(10)


Since vc.=0 at tf (the time for the car to travel the entire
distance), tf= Vca,0( Pk g) according to Eq. (9). Substitution
into Eq. (10) gives

1 V2
Xf= car,o(11)
Pk 2g

Thus, a plot of x, versus v car, /2g gives an inverse slope of
the friction coefficient.

RESULTS AND DISCUSSION
Experimental Runs The distances the car traveled dur-
ing the four experiments are shown in Figure 3 with the dashed
lines. The furthest distance traveled was 41.7 feet at an ad-
justed pressure of 7.43 atm as shown in Table 1. The traveled
distance increased with initial pressure as expected.
Friction Factor The results of the friction factor experi-
ments are shown in Figure 4. Six experiments were performed
such that the distance traveled varied between 10 and 30 feet.
The wide range of distances allowed for a more complete analy-
sis of the friction coefficient. The plot of xf versus v ar,0/2g
yielded a straight line, which is in agreement with Eq. (11).
Regression analysis resulted in an inverse slope of P k = 0.069
+ 0.003 (95% confidence) for the friction coefficient.
Model Predictions and Comparison Figure 3 shows the
model predictions based on P k =0.069 and a head loss coeffi-
cient (C) ranging from 0 to 0.2. The error bars show the range
of model predictions when k k ranges from 0.066 to 0.072 (the


0 0.5 1.0 1.5


V2car,0/2g (feet)


Figure 4. Friction
coefficient analysis as
described by Eq. (11).
The inverse of the
slope represents the
friction coefficient. The
distance traveled (xf) is
shown as a function of
the initial car velocity
(vcor,


Xf
Sdx car
0


35

30-

25 y = 14.42x
25
R2= 0.9899
020

S15

10

5

0


Winter 2006











As part of the integrated sophomore and junior team,
the students are required to write a safety and
environmental report, provide a detailed sketch of the car,
build a prototype, provide preliminary and final calibrations,
provide an engineering analysis, give a poster presentation,
and participate in the department competition.


95% confidence interval). As shown in Figure 3, the model
predictions were in good agreement with the experimental
results when C=0.1. With C=0, the model predictions were
much higher than experimental measurements for the three
highest initial pressures. However, C=0 is unreasonable since
head loss occurs as a result of the nozzle. Model predictions
with C=0.2 are lower than experimental measurements for
the three highest initial pressures. The model predictions with
a range of C values are shown to demonstrate the effect of C
on model predictions.
With C=0.1, the predictions had a difference of 1.0%, 7.3%,
and 8.6% from experimental values at initial pressures of 4.76,
6.37, and 7.43 atm, respectively. It must be remembered,
however, that the only fitted parameter in the model was the
friction coefficient, and the coefficient was measured via a
different experiment than the experiment for which the model
was used. All other parameters were car dimensions, the ini-
tial starting pressure, or the value of C. Thus, considering all
of the model assumptions, the model did a reasonable job in
predicting the traveled distance.
There are several possibilities as to why the model had some
disagreement. The first possibility was that the initial start-
ing pressure was lower than the adjusted initial pressure used
in the model. In the future, the measurement of the initial
pressure following the removal of the clamp would be ben-
eficial. A second possibility was a potential gas leak, such
that the contributing pressure to the thrust of the car would
be lower. No noticeable gas leaks were observed when run-
ning the car, however. A third possibility is a change in the
value of the friction coefficient, P k, during the course of ex-
periments due to wind conditions and axle friction (since P k
was a function of the experimental conditions). No notice-
able wind changes were observed and the distances utilized
in the evaluation of P k were similar to the experimental runs.
The effects of changing P k, however, are noticeable by the
error bars in Figure 3. The validity of assumptions is an area
that could be further explored.
With the successful demonstration of the model predictions
with the experimental results, the impact of car parameters
on the traveled distance can be explored. For instance, the
effects of varying the rear nozzle diameter, water volume,
initial pressure, or friction coefficient (representing an increase


or decrease in friction due to changing the type of wheels or
the type of surface on which the car travels) can be assessed
with regard to distance traveled. This type of exercise allows
a student to have a better understanding of how engineering
design can affect the function of the car, without the need for
numerous experimental designs.

CONCLUSIONS
This work describes the effective utilization of engineer-
ing principles in a model to predict the distance traveled by a
Chem-E-Car using the acetic acid/baking soda reaction.
Although the model is specific for one type of car-propul-
sion system, this work demonstrates how engineering analy-
sis is applicable to the Chem-E-Car competition.
One could extend the engineering analysis to include cal-
culation of the theoretical pressure build-up in the reactor,
and correlate the theoretical pressure to the experimentally
observed pressure in the chamber. Similar analysis could be
performed for hydrogen peroxide-catalase reaction systems
that generate pressure. Engineering analysis is also applicable
to other Chem-E-Car models, such as the iodide clock reac-
tion used to stop a car via breaking an electronic circuitry.
For example, the kinetics of the reaction could be incorpo-
rated with the momentum equation to predict the time at which
the reaction stops the circuitry and the distance at which the
car stops. In conclusion, engineering analysis concepts intro-
duced through the Chem-E-Car competition not only pro-
vide an opportunity to reinforce theoretical concepts but also
provide a tool for the design of the cars.

REFERENCES
1. (2005)
2. Criteria for Accrediting Engineering Programs, 2003-2004 Accredita-
tion Cycle, Engineering Accreditation Commission; Baltimore, p. 19
3. Madihally, S.V., and R.S. Lewis, "Evolution of the Chem-E-Car Com-
petition at Oklahoma State University," Session 1413d, Am. Soc. of
Eng. Ed. Annual Meeting, Salt Lake City, UT (2004)
4. Halliday, D., and R. Resnick, Fundamentals ofPhysics, 2nd Ed., John
Wiley and Sons, New York (1981)
5. Bober, W., and R.A. Kenyon, Fluid Mechanics, 1st Ed., John Wiley
and Sons, New York (1980)
6. (2005) [


Chemical Engineering Education




Full Text













PAGE 1

W inter 2006 1 Chemical Engineering Education Volume 40 Number 1Winter 2006 CHEMICAL ENGINEERING EDUCATION (ISSN 0009-2479) is published quarterly by the Chemical Engineering Division, American Society for Engineering Education, and is edited at the University of Florida. Correspondence regarding editorial matter, circulation, and changes of address should be sent to CEE, Chemical Engineering Department, University of Florida, Gainesville, FL 32611-6005. Copyright 2006 by the Chemical Engineering Division, American Society for Engineering Education. T he statements and opinions expressed in this periodical are those of the writers and not necessarily those of the ChE Division, ASEE, which body assumes no responsibility for them. Defective copies replaced if notified within 120 days of publication. Write for information on subscription costs and for back copy costs and availability. POSTMASTER: Send address changes to Chemical Engineering Education, Chemical Engineering Department, University of Florida, Gainesville, FL 32611-6005. Periodicals Postage Paid at Gainesville, Florida, and additional post offices. EDITORIAL AND BUSINESS ADDRESS:Chemical Engineering Education Department of Chemical Engineering University of Florida Gainesville, FL 32611PHONE and FAX : 352-392-0861 e-mail: cee@che.ufl.eduEDITOR Tim Anderson ASSOCIATE EDITOR Phillip C. Wankat MANAGING EDITOR Lynn Heasley PROBLEM EDITOR James O. Wilkes, U. Michigan LEARNING IN INDUSTRY EDITOR William J. Koros, Georgia Institute of Technology CHAIRMAN E. Dendy Sloan, Jr. Colorado School of Mines MEMBERS Pablo Debenedetti Princeton University Dianne Dorland Rowan University Thomas F. Edgar University of Texas at Austin Richard M. Felder North Carolina State University Bruce A. Finlayson University of Washington H. Scott Fogler University of Michigan Carol K. Hall North Carolina State University William J. Koros Georgia Institute of Technology John P. O'Connell University of Virginia David F. Ollis North Carolina State University Ronald W. Rousseau Georgia Institute of Technology Stanley I. Sandler University of Delaware Richard C. Seagrave Iowa State University C. Stewart Slater Rowan University Donald R. Woods McMaster University DEPARTMENT 8 Columbia University Carl C. Gryte, Lenora Babb, Edward F. Leonard EDUCATOR 2 Kirk Schulz of Mississippi State University CLASSROOM 14 Numerical Problem Solving Using Mathcad in Undergraduate Reaction Engineering, Satish J. Parulekar 66 Engineering Analysis in the Chem-E-Car Competition Randy S. Lewis, Aliakbar Moshfeghian, and Sundararajan V. Madihally CLASS AND HOME PROBLEMS 60 Data Analysis Made Easy With DataFit, James R. Brenner RANDOM THOUGHTS 38 The Way to Bet, Richard M. Felder LABORATORY 24 Experimental Air-Pressure Tank Systems for Process Control Education, Christopher E. Long, Charles E. Holland, and Edward P. Gatzke 40 A Flexible Pilot-Scale Setup for Real-Time Studies in Process Systems Engineering, Chanin Panjapornpon, Nathan Fletcher, and Masoud Soroush 46 Mechanical Testing of Common-Use Polymeric Materials With an In-House-Built Apparatus, Cristiana Pedrosa, Joaquim Mendes, Fern‹o D. Magalh‹es 54 A Nonlinear, Multi-Input, Multi-Output Process Control Laboratory Experiment, Brent R. Young, James H. van der Lee, and William Y. Svrcek LEARNING IN INDUSTRY 32 Partnering With Industry for a Meaningful Course Project, Rhonda Lee-Desautels, Mary Beth Hudson, Ralph S. Young 23 Call for Papers PUBLICATIONS BOARD

PAGE 2

2 Chemical Engineering EducationBalance. It's one thing that all of us strive for daily, but it seems to come quite naturally for Kirk Schulz, dean of the James W orth Bagley College of Engineering at Mississippi State University. If you know him or have ever worked with him, then you know what I'm talking about. He seems to get more accomplished in one day than most of us do in a week. And although his dedication to his work is obvious, his commitment to his family is even more evident. Kirk has been called approachable, accessible, and a great listener, and he has also been characterized as quick-witted, decisive, and driven quite a combination for a 42-year-old dean. One MSU engineering alumnus and advisory board member recently spoke of Kirk saying, "He is open to new ideas and new ways of doing things, but more than that he follows through and implements those ideas." Kirk is not one to sit around and wait for things to happen; he is a man of action and integrity with a strong desire to move his engineering program into the Top 50. Faculty excellence is at the top of every dean's list of priorities. Deans want to see their professors excel as mentors and leaders, knowing that their efforts are molding the young minds of our world's future leaders. Kirk knows that faculty excellence is the foundation of a top-tier institution and such excellence is crucial to the success of all its programs. Recognizing faculty for their accomplishments and successes in teaching, research, and service is important to him and to the continued excellence of the engineering program at Mississippi State. Simply put, Kirk is an administrator who enables faculty to do what they do bestteach others to become contributing leaders in their fields. MSU Provost Dr. Peter Rabideau remarks, "Kirk is one of the most enthusiastic and energetic administrators with whom I have been associated. He supports quality as the number one issue, and he wants to work with faculty, staff, and students to develop a vision that will move MSU engineering to the next level of excellence." As the new dean of the Bagley College of Engineering, Kirk has established the Academy of Faculty Fellows to recognize those who have risenJULIE M. LEMONSMississippi State University Mississippi State, MS 39762-9595 Copyright ChE Division of ASEE 2006 ChEeducator Mississippi StatesKirk Schulz Kirk updates alumni and friends of the Bagley College during Engineering Day on campus this past fall.

PAGE 3

W inter 2006 3to the rank of Fellow in their respective engineering professional societies. This past year, 15 faculty members were inducted into the academy. Another way that the college is recruiting new faculty and celebrating current faculty members' excellence is by honoring them through endowed positions. Currently, the Bagley College has 23 endowed positions, and in Kirk's inaugural year as dean he has made it a priority to fill these important faculty positions in the college. He has already filled nine endowed chairs or professorships as well as several other key leadership positions around the college including two associate dean positions, two department head positions, and a center d irector. With a stellar leadership team now in place, Kirk hopes to help Mississippi State's Bagley College of Engineering gain the recognition it deserves as one of the top research universities in the country.HIS BACKGROUNDBorn into a family of educators in Portsmouth, Va., during the summer of 1963, Kirk was destined to become a teacher. He was raised in Norfolk, Va., by parents who were both university employeeshis father a mathematics professor at Old Dominion University, his mother an associate registrar and director of compliance. In speaking of his parents, Kirk characterizes his father as an "outstanding tea cher who does a great job engaging his students," and his mother as "very active as a researcher, doing very creative and innovative research in historical geography." As a student at Norfolk Christian High School, Kirk first realized that he had a natural bent toward mathematics and chemistry. He excelled as a leader throughout his high school years, and as a graduating senior he was awarded both the leadership and science awards. It was during his high school years that Kirk first gained hands-on engineering experience while participating in the Soapbox Derby. He learned how to machine different steels, use a lathe, and paint. Building a car and steering it down a long track seemed to suit him. He won the local Tidewater Virginia race, going on to represent his home state in the World Championships in Akron, Ohio, in 1977. After graduating from Norfolk Christian in 1981 with 44 of his peers, Kirk actually wanted to study medicine and was advised to work as a volunteer in the emergency room to see if he would really like being a doctor. He recounts, "After about one year of volunteering at Norfolk General Hospital, I decided that I really wanted to go into engineering and not medicine." Kirk's father had always spoken highly of the engineering profession, which initially planted the thought in his head. After a family friend took it upon himself to give Kirk a tour of the chemical plant where he worked, the decision to pursue engineering was solidified. In 1984, with three years of undergraduate work under his belt, Kirk decided to transfer from Old Dominion to Virginia Tech to pursue his chemical engineering degree, receiving his bachelor's in 1986. At Virginia Tech, Kirk was actively involved in the Baptist Student Union, serving as president and statewide vice president. Kirk realized as an undergraduate that he wanted to continue his education at the graduate level and become an educator himself. While in graduate school, Kirk remained active in various organizations, even hel ping start a Scout troop at his church in Blacksburg, Va. The troop is still active today. Kirk was the first Ph.D. student that Dr. David F. Cox ever advised at Virginia Tech, so needless to say they both learned much during their time together. Kirk was conducting research in metal oxide surface chemistry. Dr. Cox recounts his time with Kirk: "He helped set up my laboratory, did all the interfacing of computers and experimental apparatus, and wrote all the code for data collection and analysis. When he wasn't "Kirk is one of the most enthusiastic and energetic administrators with whom I have been associated. He supports quality as the number one issue, and he wants to work with faculty, staff, and students to develop a vision that will move MSU engineering to the next level of excellence." MSU Provost Dr. Peter Rabideau

PAGE 4

4 Chemical Engineering Education W ith wife Noel at the Tower of London, above. Left, Kirk presents Dr. Rand German with a medallion in honor of his new position as director of the Center of Advanced Vehicular Systems and holder of the CAVS endowed chair. Kirk as a Ph.D. student under Dr. David F. Cox at Virginia Tech. working in the lab, he was founding a new graduate student association in our department, working on committees with the dean's office, and generally helping out with any departmental, college, or university task with which he was approached. It drove me crazy. "I was a young assistant professor, and I kept thinking he would be able to accomplish so much more if only he would focus more of his effort on his research. In the end, we published 12 papers from his graduate research. Fifteen years and many students later, I feel extremely lucky when I have a graduate student that manages to produce one-third of what Kirk did. With each passing year my appreciation increases for Kirk's skill, dedication, and technical abilities. Unfortunately for him, I did not realize how good I had it when he was working in my lab." Dr. Cox explains, "When Kirk was still a Ph.D. student, I asked him about his career goals in the hope that I could offer him some sound career advice. He told me even then (I kid you not) that he wanted to become a dean. I suggested to him that such a career path would be a waste of a good scientific career. Thank goodness he ignored my advice. "Throughout his academic career he has continued to turn out excellent scientific work even as he became more and more involved in administration. Kirk is the most wellrounded academician I know. I continue to be amazed by his ability to perform so well in so many different arenas. These days, I go to Kirk for career advice rather than the other way around. Whenever I am asked, I take credit for all of Kirk's success, but the bottom line is I have learned more from him than he ever learned from me." Kirk feels blessed to have had a large number of people play an integral part in the success of his career. "My father and my research adviser, Dr. David Cox, both stressed the need to work hard and to finish the things I started. Tom Owens, my first department chair at the University of North Dakota, really stressed the need to communicate with people clearly and to set high standards. Ed Fisher, my department head at Michigan Tech, and Wayne Bennett, former dean at MSU, both stressed the need to aggressively seek external support and private gifts for big, visionary ideas." After receiving his Ph.D. from Virginia Tech in 1991, Kirk accepted a position as an as sistant professor of chemical engineering at the University of North Dakota in Grand Forks, N.D. His wife, Noel, an associate professor in electrical engineering and TVA Professor at MSU, recalls, "For two years during my Ph.D. work, I lived 325 miles away from our son T imothy and him. He was a single parent during the week so I could get my Ph.D. It was a challenging time, but we made it through because of Kirk's commitment to my advancement." The two met during Noel's freshman year at Virginia Tech and connected during a mission trip to Kentucky that Kirk led. "She has been my number one fan and has been willing to pick up her research program and move it when an opportunity came up for me," says Kirk of his wife of 18 years. She echoes the sentiment saying, "Kirk has always been an extremely supportive spouse. Since he is several years ahead of me in his professional career, he has been a mentor all alongsometimes making the mistakes, then warning me about them."

PAGE 5

W inter 2006 5Above, as director of the Swalm School of Chemical Engineering, Kirk taught the unit ops labs. Here, he is assisting students as they review their results. Right, all active in Scouting, Kirk and sons Tim (now 15) and Andrew (now 11) have long enjoyed spending time together outdoors. Kirk spent four years at the University of North Dakota before moving to Michigan Tech University as assistant professor in 1995. His leadership abilities were quickly recognized, and he was promoted to associate professor in 1998. That same year, he assumed the chairmanship of the Department of Chemical Engineering. After several years heading chemical engineering at Michigan Tech, Kirk accepted a position at Mississippi State University in 2001 as director of the Dave C. Swalm School of Chemical Engineering and holder of the Earnest W. Deavenport, Jr., Chair. Dr. Wayne Bennett, dean emeritus of the Bagley College of Engineering, recalls, "I recognized his leadership skills when he interviewed for the director's position of the Swalm School of Chemical Engineering. Under his leadership, the school progressed on every front. The undergraduate programs flourished, graduate enrollments increased, and the research set new records." Success doesn't come without ample opportunities. This idea is something that Kirk recognizes, and he acknowledges that there have been several people throughout his career who have given him an opportunity when conventional wisdom would have said otherwise. One such individual was Bob W arrington, dean of engineering at Michigan Tech, who was willing to give a 35-year-old associate professor a chance to be a department chair very early in his career. Another administrator willing to take a chance was MSU Provost Dr. Peter Rabideau, who agreed to let a 41-year-old lead the university's flagship college. "Both of these individuals gave me an opportunity in my career that many people never get," comments Kirk.HIS MOTIVATIONHis motivation is simple: to make a difference in the lives of others, especially the students. Working with students and faculty is what Kirk enjoys the most about his job. "Teaching at the university level in my mind is the real chance to make a difference in someone's life. If you ask an engineer who they had for chemical engineering reactor design, most can give you a nameeven 20 years later. When you talk with alumni, you realize just how big an influence we have on a person during their formative years," says Kirk. Kirk receives great joy and satisfaction from seeing his former doctoral students become successful in their careers. Dr. Alan Nelson, one of Kirk's first doctoral students, is now an associate professor and associate chair in the University of Alberta's Department of Chemical and Materials Engineering. Regarding Kirk's abilities as a professor, Dr. Nelson says, "I have a great deal of respect for Kirk, not only because he is a scholar and educator, but because he has been and continues to be a benevolent mentor to so many individuals. I would certainly not be where I am today without the research supervision and professional guidance Kirk has provided to me over the years." Dr. Nelson goes on to say, "Kirk's ability to maintain personal and professional balance should be a model for all new chemical engineering faculty. He is a case study of how to be an effective and efficient educator, while not sacrificing his personal goals or time with his family." Professors work hard to impart some vast wisdom or knowledge to the students they teach. The students will, of course, remember some of a teacher's meager efforts in the lab and

PAGE 6

6 Chemical Engineering Education "Whenever I am asked, I take credit for all of Kirk's success, but the bottom line is I have learned more from him than he ever learned from me." Dr. David Cox,Kirk's Ph.D. advisor at Virginia Tech classroom and how he or she graded them, but in the end it is one's character that students take note of the most. It is how they are treated as students and how faculty respond to life and all its clichŽs that demonstrate what true "balance" in life is all about.HIS FAMILYFamily is the most important part of Kirk's life. One favorite family activity at the Schulz house is Scouting, which probably has a little to do with having an Eagle Scout for a dad. Currently, Kirk is the assistant Scoutmaster of Troop 14 in Starkville. Son Timothy (15) is a Life Scout, while Andrew (11) is in Webelos. The Schulz family also enjoys spending time outdoors as well as traveling to a wide variety of destinations, such as Disney W orld, London, and San Diegonot to mention following the MSU Bulldogs to the SEC Basketball Tournament in Atlanta. Kirk shares a good relationship with Noel's parents, and they have always been supportive of his professional career. While at Virginia Tech, Kirk was a student w orker for Dr. Charles Nunnally, the assistant dean of engineering at Virginia Techand Noel's father. Being a dean and having a wife who is an associate professor oftentimes means that the weeknights are booked with college or university events. It is not unusual for Timothy and Andrew to be present at some of these functions, and they are often a crowd favorite at such events. Kirk made it a point when he first became dean of the Bagley College to let his staff and faculty know that the dean's office was a family-friendly environment. "We all have families and from time to time there are family situations that come before work. I want my staff and faculty to know that I support them personally and am here if they ever need anything," said Kirk. Throughout the academic year Mississippi State will often host MathCount competitions and Science Fairs, and you will see the Schulz boys in and out of the dean's office visiting Dad. One of the most enjoyable visits to campus for his sons was during Engineering Week at MSU this past year. It was toward the end of the week and just so happened to be the same day as Timothy's MathCount competition. All week, engineering students had purchased tickets to "pie" a faculty member or fellow student as part of a fund-raiser. The activity on the Drill Field drew a c rowd of curious onlookers, as Kirk and two other department heads sat bravely in anticipation of the firing squad of students and faculty that stood before them. It is not often that students, or faculty for that matter, get to throw a pie at their dean. Needless to say, Kirk took it in stride and Timothy and Andrew were greatly amused.HIS CAREERThroughout his career Kirk's talent for administration and leadership have been recognized and have afforded him some wonderful opportunities. Of course, no one begins their career as a dean of a college; there are dues to pay and work to be done along the way. Kirk has traveled that path, working his way from a summer school chemical engineering instructor at Virginia Tech all the way to dean of a major state university. Each engineering program that Kirk has been involved in has benefited from his leadership. As an administrator, much of his time and energy have been devoted to improving alumni relations, growing graduate programs, increasing the diversity of faculty and students, recruiting new faculty, and increasing external funding. With a strong desire to see faculty collaborating across the college at Michigan Tech, he assisted in the initiation of the Carbon Technology Center. This is a multidisciplinary research center involving faculty from chemical engineering, mechanical engineering, civil engineering, and chemical engineering technology in research focused on polymer composites. At Mississippi State, Kirk saw a need to make pursuing a Ph.D. in chemical engineering more enticing. So as director of the Swalm School of Chemical Engineering, he led the efforts to establish the first directadmit doctoral program, increasing the number of Ph.D. students from three to 15 in three years. Kirk has shared his knowledge and research findings through 42 journal articles that he has authored or co-written; he has presented numerous conference papers and given 100 presentations. He has had over $1.8 million in funded research projects and has provided guidance to six doctoral students and 14 master's students. Kirk is currently co-chairing the Chemical Engineering Division of ASEE's 2007 Summer School for Chemical Engineering Faculty, wh ich will be hosted on the campus of Washington State University. His service to engineering professional organizations does not end here. He is very involved with ABET and has served as a program evaluator, a member of the AIChE Education and Accreditation Committee, and now a member of the Engineering Accreditation Commission (EAC). He has served as the division chair and program vice chair for ASEE's New Engineering Educators; he has held offices such as secretary/treasurer, program chair, and director for the Chemical Engineering Division; and he is a senior member of AIChE. Since 2003, he has been a member of the Advisory Board for the University of Tennessee's Department of Chemical Engineering.

PAGE 7

W inter 2006 7Kirk, center, along with Drs. Tony Vizzini (head of aerospace engineering) and Glenn Steele (head of mechanical engineering), getting piefaced to the delight of students, faculty, and especially Kirk's sons Timothy and Andrew during last year's E-week activities. Along the way, Kirk has been recognized in numerous capacities for his work. Early in his career, he was named Outstanding Professor of the Year at the University of North Dakota. In 1997, while at Michigan Tech, Kirk was recognized by the Chemical Engineering Division of ASEE with the Raymond W. Fahien Award for his outstanding accomplishments and commitment to his profession. He has also received an NSF CAREER Development Award (1995-1999), the Dow-ASEE Outstanding Young Faculty Award (19951996), and ASEE's Outstanding Teaching Award for the North Midwest Section (1999). His alma mater, Virginia Tech, named him as the Outstanding Young Alumnus in 2001. The next year, Kirk was recognized by Mississippi State University as Outstanding Professor in Chemical Engineering.HIS VISIONOver the past 20 years, Kirk has observed a dramatic change taking place in the field of engineering education, noting the increased importance now placed on the quality of undergraduate teaching and faculty development at research universities. He has also seen the communication skills of engineering graduates improve dramatically from what they were 20 years ago. In the Bagley College of Engineering a strong emphasis is placed on both of these areas. Through the newly established Center for Engineering Student Excellence, programs such as the Shackouls Technical Communication Program, Six Sigma, Study Abroad, congressional internships, and the Entrepreneurship Program emphasize the need to take technical degrees a step further. Students are encouraged to complement their technical degrees with programs such as these, providing them with better global awareness as well as improved communication and business skills. Kirk's vision for the Bagley College is to see it recognized as one of the top research institutions in the United States. He knows that this must be done by investing resources in carefully selected areas where MSU can be internationally renowned. He strongly believes that MSU will be one of the leading institutions in providing a diverse engineering workforce, and he is committed to the education of African-American engineering students at all degree levels. The Bagley College of Engineering, in fact, already ranks among the top 15 schools in graduating African-American engineers, and Kirk wants his college's ranking to go even higher. Ultimately, Kirk wants to see the MSU Bagley College of Engineering thrive and become one of the Top 50 engineering colleges in the country. The core of his strategic plan focuses on providing first-rate education while continuing to recognize faculty for their research endeavors and teaching excellence. He believes that MSU has an obligation to the state of Mississippi and the nation, and to support growth and economic development with the expertise and knowledge found in the faculty of the Bagley College of Engineering.HIS MOTTOThe advice he extends to each of the engineering students at Mississippi State is the same adage he chooses to live by: "Seek out challenging opportunities during your careerlook for something that people say can't be doneand then go out and do it." Now 42 and almost a year into his deanship, is there anything that Kirk would have done differently along the way? Not a chance.

PAGE 8

8 Chemical Engineering EducationColumbia Chemical Engineering is quintessentially New York, a central part of a university whose legal name is Columbia University in the City of New York. It is a university united to its city perhaps more than any other urban university in the United States. In 2005 Columbia Chemical Engineering celebrated its 100th anniversary, but it is rooted even farther back into the chemical, financial, and public works history of its home city; the special characteristics of contemporary chemical engineering at Columbia trace far into the department's, the city's, and the country's past.A STORIED HISTORYColumbia's engineering school was founded in 1864, initially named the School of Mines. It originated out of science departments that had participated in the 19th-century struggle in much of the W estern World to reconcile philosophical and practical views of science. In 1896, separate schools of engineering, chemistry, and architecture were set off from the School of Mines, resulting, finally, in Columbia's first curriculum in chemical engineering being offered by the School of Chemistry in the fall of 1905 (having been approved the preceding February). A towering figure at Columbia and in New York at the time was Charles Frederick Chandler a Bostonian whose pivotal education in chemistry was, notwithstanding his origins, in Germany. There he met some of Europe's leading scientists, including Wšhler, Liebig, von Humboldt, and Pasteurall of whom had progressed from a purely philosophical to a decidedly practical bent. On his return with a doctorate from Gšttingen, Chandler joined Columbia, where he taught for 46 years. ChEdepartmentChemical Engineering atColumbia UniversityCARL C. GR YTE,with contributions fromLENORA BABB ANDEDWARD F. LEONARDColumbia University New York, NY 13699-5705 Copyright ChE Division of ASEE 2006 CELEBRATING 100 YEARS

PAGE 9

W inter 2006 9Professor Chandler actually campaigned for 14 years before 1905 for a program that would produce what we could now only call chemical engineers. He was strongly resisted by professors who saw chemistry as pure science, beautiful in its own right and with deep philosophical meaning. Such battles raged in many universities and account for the earliest realization of chemical engineering as a distinct discipline being born in technological institutions such as MIT.INFLUENCES FAR AND NEARNew York in 1905 was one of the high-thinking, intellectual centers of the young United States. It was an international business center where agents of Europeanespecially Germanchemical firms issued and oversaw limited licenses to operate processes developed in Europe. In reaction, strong incentives existed for establishing an American capacity to develop and improve chemical processes, even before this became a desperate priority when the first World War broke out. Professor Chandler was close friends with Nicholas Murray Butler Columbia's president from 1902 until 1949. Both men consulted and were on good terms with the New Y ork business community centered on Wall Street. Columbia's role in international business and politics was then, and remains today, preeminent, affecting every department of the university. These connections, and Professor Chandler's popularity as a teacher deeply involved in his subject, impelled the Columbia program to flourish. New York City after the Civil War was also a dirty, overcrowded, unhealthy, and unsafe place. In addition to his extensive involvement with the chemical industry, Professor Chandler played a central role in the public health of New Y orkers, dealing officially for the city with "the adulteration of milk, kerosene accidents, gas-factory nuisances, and general sanitation," as well as an issue that persists todaylead in drinking water. Professor Chandler was also very concerned with the chemical education of physicians and pharmacists and presented lectures to those professions regularly. We at Columbia like to think of him as our first biomedically oriented chemical engineer.BUILDING THE PROGRAMRecords show the first chemical engineering curriculum at Columbia laid out four solid years of unremitting "chemistry, engineering, metallurgy, mathematics, mechanics, physics, and mineralology," having presumed prior preparation in "algebra, geometry, plane trigonometry, chemistry, physics, freehand drawing, English literature, composition and grammar, American and English history, French, and German." Professor A.W. Hixson who joined the faculty in 1922 and later became the preeminent department historian of Columbia, put forth the claim that despite the preexistence of other programs entitled chemical engineering, Columbia's was "the first well-balanced and completely integrated curriculum in chemical engineering to be established in America." In what is arguably a less disputable first, only five years after its 1905 founding Columbia admitted students to study for the degree "doctor of philosophy in chemical engineering." Professor Chandler's handpicked colleague and later successor, Milton C. Whitaker also became a leading figure in chemical engineering education. He was recognized with two honorary degrees and was an early president of AIChE. Engendered by these early innovators, Columbia Chemical Engineering's current specializations all have origins and histories that reach back to the department's founding.Polymer SurfacesIn the earliest years, polymers were mostly natural and the coursework was concerned with materials such as cellulose, gutta percha, and rubber. The esters of cellulose were already in wide use, however, and as the department was being founded Leo Baekeland was inventing the phenol-formaldehyde resin that was to bear his name. Indeed, Baekeland was an honorary professor in the department, an advisor to Columbia's President Butler, and an overseer of the chemical engineering program nearly until his death in 1944. In 2005 Columbia Chemical Engineering celebrated its 100th anniversary, but it is rooted even farther back into the chemical, financial, and public works history of its home city. An early view of the large electrical generators at Columbia ChE's now-closed Heat Transfer Research Facility. For more than 50 years the laboratory tested electrically heated models of nuclear fuel-rod assemblies. Practically every configuration used in the W estern World's boiling-water nuclear reactors was tested at this facility. Tests were run late at night to reduce dimming of lights in Manhattan.

PAGE 10

10 Chemical Engineering Education L eft, Columbia's Unit Ops Lab, circa 1929, with students dressed "properly" for lab work in those days. Columbia's department was one of the earliest proponents of the unit operations concept and such laboratories evolved continuously along these lines through the first half of the 20th century. Above, Professor Elmer Gaden and family. Gaden was named "Father of Biochemical Engineering" by Chemical Engineering News in 1971; this photo appeared on the magazine's cover. W ith the nation's drive for independence from European technology, major emphasis was placed on process and plant design. A steady stream of doctoral theses based on process and plant design flowed out of the department from 1915 through the beginning of the second World War. While these dissertations covered a wide range of processes, many were concerned with raw materials for synthetic substances. In 1939, James M. Church arrived at Columbia and for more than 20 years ran an undergraduate unit processes laboratory in which students conducted carefully scaled-down versions of industrial, mostly organic, processes. The real resurgence of interest in polymers began in the mid-'60s, however, with the hiring of George Odian in 1966 and Harry Gregor in 1967. They were joined by Carl Gryte in 1972 and Christopher Durning in 1983. The trend continued when Ben O'Shaughnessy a condensed-matter physicist, joined the faculty in 1988, followed by Rasti Levicky in 1998, and Jeffrey Koberstein in 2000. The lasting theme of this resurgence has been an interest in polymer surfaces in an exceptionally wide range of applications.Electrochemical EngineeringElectrochemistry and electrochemical engineering have had a similarly long run through the department's history. For a long time the ability to generate electricity from the potential energy of water far outweighed the ability to transmit electricity over long distances. In that era, a major center of electrochemical manufacturing evolved at Niagara Falls, N.Y. The first professor of electrochemical engineering, Samuel A. Tucker was appointed in 1910 and rapidly built up what historian Hixson has called the most complete electrochemical laboratory in the country. The strength of a great university was brought to bear on this enterprise through the influence of Columbia's Department of Physics, with its interests in electricity. In 1922, Colin G. Fink joined the department to begin a long and distinguished career in electrochemical engineering. Professor Fink was a 1903 Columbia graduate who subsequently received his Ph.D. in chemistry (from the University of Leipzig) in 1907. Fink's personal research accomplishments were extraordinary, includingduring earlier employment with General Electricthe process for drawing tungsten wire that was essential to light-bulb manufacture, as well as the development of chromium plating. He became the executive secretary of the Electrochemical Society, revitalized it, and negotiated a home for it on the Columbia campus. Fink (who retired in 1950) was joined by Henry B. Linford in 1942. Linford, too, served as executive secretary of the Electrochemical Society, retiring in 1976. Joining late in Linford's tenure was Huk Y. Cheh (who retired in 2001 to become director of research for the Duracell Company). Cheh was honored in 1982 with the Ruben-Viele chair named in honor of Samuel Rubena protogŽof physics professor Michael Pupinwho made important contributions to the electrochemistry of metals. Cheh was joined in 1991 by Alan W est a specialist in electroplating. West was joined in 2000 by Scott Calabrese Barton specializing in fuel cells.

PAGE 11

W inter 2006 11 Professor Carl Gryte (on staircase) with Isao Noda (top left) and other doctoral students installing the Cobalt-60 radiation source used in polymer research (about 1971).BioengineeringBioprocessing, biochemical engineering, and biomedical engineering have also long figured in the department's history. Professor Chandler's influential involvement with the healthcare community and public health has already been mentioned. Following in his footsteps was D.D. Jackson who lead the department for 23 years, from 1917. Jackson was trained in chemistry, engineering, and biology, and had a major interest in biochemical processes (second only to his interest in processes for the production of heavy chemicals). Professor Jackson was succeeded by the aforementioned Professor Hixsonwho had a major interest in yeast chemistry. Such chemistry was fundamental to much early bioprocessing. The real prominence of Columbia in the area of bio processing, however, came with the rise of Elmer L. Gaden in th e years immediately following World War II. The discovery of penicillin and its manufacture by fermentation, combined with the extensive demand for it during the war, had enormously accelerated interest in bioprocessing. Professor Gaden, an eminently practical man but also an ideologue, quickly grasped the significance of oxygen delivery in fermentations and developed, over many years, methods for measuring and increasing it. His students were soon continuing his efforts, both in other schools and in industry. Juan Asenjo and Alex Seressiotis followed Gaden, who left Columbia in 1974 to ultimately become a professor of chemical engineering at the University of Virginia. Such was the influence of his work that, on the cover of its May 31, 1971, issue, Chemical and Engineering News declared Professor Gaden, "the father of biochemical engineering." But beyond Gaden's contribution to biochemical engineering was his early recognition of the development of biomedical engineering. Largely through his efforts, by 1965 Columbia had graduate and undergraduate programs in "bioengineering" with a decidedly medical orientation. The graduate program was run by an interdisciplinary committee, but the undergraduate program remained within chemical engineering until 1997, when a separate department of biomedical engineering was established. Many faculty members contributed to the bioengineering program, which was seen as a broad effort to focus the tools and methods of chemical engineering on biological and medical problems. These influential individuals included Jordan Spencer, Harry Gregor, Frank Castellana, Mary Anne Farrell-Epstein Huk Cheh, and Rakesh Jain No faculty member was more involved in this effort than Edward Leonard however, who has worked on problems related to artificial organs since 1956two years before he joined the Columbia faculty. In more recent times, the department has initiated a program in genomic engineering the first of its kind in the country. Professor Jingyue Ju is the director of sequencing in the Columbia Genome Center, while Professors Ju, Levicky, Banta and Leonard are all involved in research that relates to the modeling and manipulation of gene expression.SHAPED BY WORLD EVENTSThus, the three areas of current concentration in the depa rtment have extensive histories. The full story, however, is necessarily a bit more complicated. Two great wars stamped the department indelibly. W orld War I matured chemical engineering throughout the country. Europe, most notably Germany, no longer served as the fountainhead of chemical engineeringprofessors were no longer "finished" in European universitiesand the American chemical industry moved rapidly toward reliance on chemical engineers wholly formed in the United States. This shift lead inexorably to the dominance that American chemical engineering now possesses. W orld War II had more specific effects. Columbia was the home of the Manhattan Project. While the project later moved to other universities and to the national laboratories, its beginnings were at Columbia, and no other university was as

PAGE 12

12 Chemical Engineering EducationElectrochemical Engineering1911192219421970198419912000 SamuelColin FinkHenry B.Huk YukUlrich Alan WestScott Calabrese T uckerLinfordChehStimming BartonBiomedical Engineering18661946195819771983 1988 2000 20012004 CharlesElmer Gaden, Edward RakeshJuan Alex Jingyue Ju NinaScott Chandler"Father of Leonard JainAsenjo Seressiotis ShapleyBanta Biochemical Engineering"Polymer Engineering1914 19171939 1966 1967 19721983 1988 1998 2000 Leo Baekeland, D.D. JacksonJames George Harry P. CarlChris Ben Rasti Jeff inventor ofChurch Odian Gregor GryteDurning O'Shaughnessy Levicky Koberstein Bakelite, the first important thermosetting resin T ABLE 1Eminent Faculty in Columbia's Three Principal Areas (current faculty in bold) The current faculty. Seated: Scott Banta, Rasti Levicky, Nina Shapley, Jingyue Ju, Jeff Koberstein. Standing: Edward Leonard, Alan West (chair), Ben O'Shaughnessy, Scott Calabrese Barton. Not pictured: Carl Gryte.much affected. Chemical engineers participated, especially in the early conceptualization of the gaseous diffusion process for the separation of uranium isotopes. While the detailed story remains to be told, Professor Thomas Drew was pivotal in these efforts. He remained at Columbia until 1965. Another legacy of the Manhattan Project was Columbia's Heat Transfer Research Laboratory. This laboratory, founded in 1951, served as the major research and testing facility for thermal-hydraulic design of nuclear reactors until its closure in 2003. In major tests it could consume 13 mW of electrical energy, which had to be accessed out of peak usage times yet could still dim lights on the west side of Manhattan during tests! The first director was Professor Charles F. Bonilla Later directors included a number of chemical engineering professors, notably Huk Y. Cheh in the laboratory's later years.THE BIG PICTUREThroughout the history of chemical engineering at Columbia there has been a steady concern with the "core" of chemical engineering. Notwithstanding the historical specialties emphasized above, Columbia Chemical Engineering has always been a broad endeavor, not a boutique dedicated to select applications. The more than 50 individuals who have held professorial positionstoo many to mention herehave represented almost every area of research: process design and development; energy conversion; particular unit operations such as distillation, heat transfer, fluid mechanics, solids separations, extraction, and most of the rest, as well as kinetics and reactor design; process control and optimization; and oil and gas recovery. Columbia Chemical Engineering today has 10 faculty, currently chaired by Alan West. Table 1 lists their interests.

PAGE 13

W inter 2006 13 An Innovative Introduction to Chemical Engineering ComputingIntroduction to Chemical Engineering ComputingBruce A. Finlayson0-471-74062-4 Paper December 2005 360 pages $54.95 US / $70.99 CAN / £31.50 / C = 45.80 As chemical engineering technology advances, so does the complexity of the problems that arise. The problems that chemical engineers and chemical engineering students face today can no longer be answered with programs written on a case-by-case basis. Introduction to Chemical Engineering Computing teaches professionals and students the kinds of problems they will have to solve, the types of computer programs needed to solve these problems, and how to ensure that the problems have been solved correctly. Each chapter in Introduction to Chemical Engineering Computing contains a description of the physical problem in general terms and in a mathematical context, thorough step-by-step instructions, numerous examples, and comprehensive explanations for each problem and program. This indispensable text features Excel, MATLAB, Aspen PlusTM, and FEMLAB programs and acquaints readers with the advantages of each. Perfect for students and professionals, Introduction to Chemical Engineering Computing gives readers the professional tools they need to solve real-world problems involving: € Equations of state € V apor-liquid and chemical reaction equilibria € Mass balances with recycle streams € Mass transfer equipment € Process simulation € Chemical reactors € T ransfer processes in 1D € Fluid flow in 2D and 3D € Convective diffusion equations in 2D and 3D Order a Copy Today!North, Central & South America T el: 877.762.2974 Fax: 800.597.3299 Email: custserv@wiley.com Internet: www.wiley.com Europe, Middle East, Africa & Asia T el: +44 (0) 1243 843 294 Fax: +44 (0) 1243 843 296 Email: cs-books@wiley.co.uk Internet: www.wileyeurope.com Germany, Switzerland, & Austria T el: +49 (0) 6201 606 152 Fax: +49 (0) 6201 606 184 Email: service@wiley-vch.de Internet: www.wiley-vch.de

PAGE 14

14 Chemical Engineering EducationWith the development and availability of fast, efficient computers, the role of computing in analysis and solution of engineering problems and graphical communication of results has increased dramaticallyleading to greater need for computer-application skills in the curricula and practice of various engineering disciplines.[1] Efficient solution of problems is essential for enhanced understanding of chemical engineering principles at all course levels.[1] Commercially available computational packages, such as Maple, Mathcad, Mathematica, and Matlab, have considerably reduced the time and effort required for engineering calculations. Such programs allow engineers with limited or no formal training in programming to solve relatively complex problems.[2-4]One of these packages, Mathcad, combines some of the best features of spreadsheets and symbolic math programs, allows efficient manipulation of large data arrays, and provides a good gra phical user interface.[2, 4, 5] Ability to perform calculations with units is an important feature of Mathcad for engineering students.[2] While students need to understand the problem they are trying to solve, they may know little or nothing about numerical analysis; Mathcad allows them to work on problems even if they know very little of the program's syntax.[4] Some of the advanced and special capabilities of Mathcad, such as solution of stiff dif ferential equations, statistical methods for nonlinear parameter estimation, and programming, have been used in undergraduate courses.[3, 5-8]Experience in using Mathcad in the undergraduate chemical reaction engineering course at the Illinois Institute of Technology (IIT) is discussed here. Pertinent illustrations are provided to demonstrate the ease with which problems with varying complexity can be solved using Mathcad. Example problems considered for illustration deal with simultaneous solution of: linear algebraic equations ( i.e. kinetic parameter estimation); nonlinear algebraic equations ( i.e. equilibrium calculations for multiple reactions and steady-state behavior of isothermal/nonisothermal CSTR with single/multiple reactions); integral equations ( i.e. design of steady-state plug flow reactor, or PFR); integral-algebraic equations; and nonlinear ordinary differential equations ( i.e. solution of conservation equations for steady-state PFR and unsteady state CSTR). Based on these illustrations, the benefits of this user-friendly software in accelerating learning and strengthening the fundamental knowledge base should be evident. With hand calculations being replaced by computation, it is more imporSatish J. Parulekar is a professor of chemical engineering at Illinois Institute of Technology. His research interests are in biochemical engineering and chemical reaction engineering. His research publications include five book chapters and the book Batch Fermentation: Modeling, Monitoring, and Control. He has held visiting appointments at the University of Minnesota and the National Chemical Laboratory, India, and faculty associate and faculty research participation appointments at Argonne National Laboratory.NUMERICAL PROBLEM SOLVING USING MATHCADin Undergraduate Reaction EngineeringSA TISH J. PARULEKARIllinois Institute of Technology Chicago, IL 60616 ChEclassroom Copyright ChE Division of ASEE 2006

PAGE 15

W inter 2006 15tant than ever to consistently validate and verify the results.[9]This is done, where appropriate, in the illustrations that follow. The Mathcad worksheet for each illustration is provided in a table and contains problem input, solution algorithm, and presentation of results in appropriate (numerical and/or graphical) format.NUMERICAL ILLUSTRATIONSIllustration 1This illustration pertains to estimation of kinetic parameters using linear regression, which requires sol ution of several simultaneous equations that are linear in unknown parameters. Consider the following relation among varia bles y and xj(j = 1, 2, . m) that is linear in terms of the unknown parameters qj (j = 1, 2, . m). yyexxxpmm=+=+++(), yp11221 qqq K I nformation on y and xj (j = 1, 2, . m) is available in the form of n samples (n > m). The parameter estimation problem then involves finding the parameter set Q j=1, 2, . ., mjqq =()j for which ei i n 2 1 = is minimized. After some algebra, the necessary and sufficient condition for this can be deduced to be AbXY XXXYTT ,, QQ = =()()A=X b=X TT 12 with Y y y y X xxx xxx xxxn m m nnnm m= = = 1 2 11121 21222 12 1 2 , M L L MMOM L M Q q q q ()3 Each column in array X represents the collection of values of a particular variable xj (j = 1, 2, . ,m) for different samples. The goodness of fit of the least squares can be examined by calculating the relative error for each data point or sample ( Œi ) defined as Œi = (yi yip)/yi, i = 1, 2, . n. The specific example considered here pertains to Problem 5-13 of Fogler[10] and involves a three-dimensional linear fit (m = 3). The dependence of the rate r of a solidcatalyzed association reaction between A and B on partial pressures of A and B, pA and pB, respectively, is expressed as rkppAB=a b with k, a and b being the kinetic parameters to be estimated. The units for k, pA, pB, and r are mmol/{g cat.h.(atm)( a + b )}, atm, atm, and mmol/{g cat.h}, respectively. Upon linear transformation of the expression, a relation linear in terms of three unknown parameters can be obtained as in Eq. (1), with x1 = 1, x2 = ln(pA), x3 = ln(pB), y = ln(r), q1 = ln(k), q2 = a and q3 = b The data for pA, pB, and r are listed in Table 1, where the Mathcad worksheet for this problem is also shown. Mathcad allows input only of column vectors. Two-dimensional arrays can be constructed from column vectors already introduced using the "stack" feature. The predicted reaction rates, rp, are compared with the reaction rates available from measurements, r, and provide a very close fit (Table 1). The reason for presenting the relevant equations in this and other illustrations, where necessary, is to enable the reader to see how the equations to be solved and the Mathcad syntax are almost identical.[4] In the illustrations that follow, the subscript 0 denotes variable values at the start of a batch reactor or in the feed for a flow reactor.Illustration 2This illustration pertains to an autocatalytic reaction and involves comparison of space times ( t ) required for steady-state isothermal operations of a CSTR and a PFR. The reaction ABB + 2 occurs as per the kinetics r = kCACB.T ABLE 1W orksheet for Illustration 1 : . . . : . . . : . . rppAB= ˆ = ˆ = 042 096 018 078 12 028 288 01 02 005 03 04 005 05 01 02 005 001 0 002 04 05 1 1 1 1 1 1 1 1111 23 12. . : : : : :, ˆ = ˆ =()=()=()= X Xn pX np Yn rQ stackXXXAB TT 3 3 1 123112233 36 65209970205 4 87110T TTT p rr p rXQXX XY krXXX k x()== () =()=== + + ()===-() = -: : : exp : : :exp . . q qaqbqqqq ab ˆ -1 48110 9 41710 1 07310 2 95910 6 00610 4 098103 3 3 3 3 3. . . x x x x x x

PAGE 16

16 Chemical Engineering EducationThe feed contains A and B in the ratio 100:1. For the feed composition under consideration, the reaction rate is expressed as rkCfXfXXXA=()()=-()+()()0 21001 4 A comparison of the req uired space times for a CSTR and a PFR is e quivalent to the comparison of the corresponding Damkohler numbers, Da (= kCA0 t ) which can be obtained explicitly in terms of the exit conversion Xe. The Mathcad worksheet for this problem is shown in Table 2. Rather than calculating Da for one value of Xe at a time, the Da's for CSTR and PFR are expressed as a function of Xe, a floating variable (Table 2). The Da's for a particular Xe are then readily obtained by plugging the value of Xe into the symbolic solutions. Keeping Xe floating also enables the student to represent the results graphically over a specified range of Xe (0 < Xe < 1 in Table 2). This beneficial feature in Mathcad is also used in Illustrations 7 and 8. For minimizing the required space time, a CSTR is the reactor of choice up to a critical conversion, Xc, and a PFR beyond this conversion (Table 2). Identifying Xc requires solution of an integr al-algebraic equation in Xc the numerical solution of which is certainly challenging for an undergraduate student. Using Mathcad, the solution is obtained rather easily and its accuracy is demonstrated in Table 2.Illustration 3The gas phase reaction, SOOSOC223 A + B1 2()()() occurs as per the kinetics r = kCACB. For the feed composition under consideration, the reaction rate is expressed as[10] rkCfXfX XX XA=()()= -()-()-() ()0 2 2105405 1014 5 .. with k being the kinetic coefficient, CA0 the feed concentration of A, and X the fractional conversion of A. The reaction is carried out in three CSTRs of equal volume in series with the exit conversion being specifi ed. Computation of the intermediate fractional conversions and the required total space time, or of the corresponding Da =()kCA0t requires simultaneous solution of design equations for the three reactors, viz. XXfXii i Da i--()==()-()1 301236 , In the above, Xi refers to fractional conversion of A in reactor i (X0 = 0) and Da corresponds to the total space time for the three-reactor battery. The Mathcad worksheet for this problem is shown in Table 3. The solution proceeds by providing initial guesses for Da, X1, and X2. The validity of the solution is verified by substituting Da, X1, and X2 generated by the solution into Eq. (6). 0 0.51 0 10 20 Da CSTR Xe () Da PFR Xe () Xe T ABLE 2W orksheet for Illustration 2 Given V erify DaX XX dX DaX X XX WhenisDaDa X XX dX X XX XPFRe X e CSTRe e ee CSTRPFR e e ee X e()= -() +()()= -() +()= = -() +()-() +() =: : ? :. .. 1 1001 1001 07 1 1001 1001 00 0 c ce c CSTRc c cc X cFindXXDaX XX dX X XX x : . .. =()=()= -() +()-() +()=--0 8416222 1 1001 1001 4 187108 0

PAGE 17

W inter 2006 17Illustration 4This illustration pertains to estimation of the equilibrium composition of a reaction mixture and is adopted from Problem 4.14 of Cutlip and Shacham.[11] The reactions HSAHCSDKP HSASOBSDHOEK yy yy PP y C y D y A P DE A B 2221 05 05 22222 152 2 0505 21527 , . .()¤()+()=()+()¤()+()=() occur in a gas phase batch reactor. In the above, KP1 and KP2 denote the equilibrium coefficients, P the total pressure, and yJ the mole fraction of species J. The initial pressure P0 is 1.2 atm and the initial composition is (I = inerts): yA0 = 0.45, yB0 = 0.25, and yI0 = 0.3. For KP1 = 0.45 atm0.5 and KP2 = 28.5 atm0.5, obtain the composition of the reaction mixture at equilibrium in constant volume and constant pressure operations of the reactor. Let nJ0 denote the initial number of moles of species J, while x1 and x2 equal the extents of reactions 1 and 2, respectively, and nJ equals the number of moles of J after certain extents of the two reactions. The expressions for yJ's in terms of x1 and x2 then, are: yJ = nJ/nt, nt = J nJ, nt0 = J nJ0, and J = A, B, C, D, E, I, with nA = (nA0 x1 2 x2 ), nB = (nB0x2 ), nC = (nC0+ x1 ), nD = (nD0+0.5 x1 +1.5 x2 ), nE = (nE0+2 x2 ), nI = nI0, and nt = (nt0+0.5 x1 +0.5 x2 ). The mole fractions in the equilibrium relations in Eq. (7) and the reactor pressure for constant volume operation are then expresse d as y y y y yy y nn PP constvolumeA A B B CD E tt= --()= -()== +()==++()== =01202 1 12 2 121 1 0 2 2 0 020515 2 10505 rr y r y r y rr y r y yrrr x r x , .. , .., , tan Y( ()=()(), tan PP constpressure08 T ABLE 3W orksheet for Illustration 3Given V erify XDaXX X Da XX X XX Da XX X312 1 11 1 2 21 22 2 20920306 3 105405 1014 0 3 105405 1014 :. : :. :. .. .. ==== -() -()() --() -()() = --() -()() ˆ =()=== -(= =0 3 105405 1014 0 1950206350823 3 132 33 3 2 1 2 12 12 1 1XX Da XX X Da X X FindDaXX DaXX X Da X .. :,, . ) ) -()() =- --() -()() =- --()-05405 1014 9 21210 3 105405 1014 3 98410 3 10541 1 2 9 21 22 2 2 9 32 3.. . .. . X X XX Da XX X XX Da X -()() = 05 1014 03 3 2. X X

PAGE 18

18 Chemical Engineering EducationThe Mathcad worksheet for constant volume operation is shown in Table 4. The equilibrium relations are nonlinear coupled equations in the dimensionless extents, r1 and r2 initial guesses for which need to be supplied (Table 4). The extents calculated are substituted into equilibrium relations to verify that these indeed are satisfied. Computations for constant pressure operation, not shown here, proceed in a similar fashion. Illustration 4 reveals to the students the uniqueness of the physically realizable equilibrium composition for a given initial composition.Illustration 5Illustration 5 pertains to multiplicity of steady states in an isothermal CSTR. The reaction under consideration, catalytic hydrogenation of ol efins, obeys the kinetics r = CA/(1+CA)2, with r being expressed per unit reactor volume. The operating conditions for the reactor are: CA0=13 mol/L, V=10 L, v0=0.2 L/s.[12] The Mathcad worksheet for this illustration is shown in Table 5. The symbolic solution of the steady-state mass balance for A, viz. CCrCAAA 0-()=()t reveals that the reactor can operate at three steady states. The students observe that the steady-state mass balance is a cubic equation in the unknown, CA, and therefore has three solutions, not all of which may be real. The verification of solutions of the steadystate mass balance, generated as a vector, follows as usual and is done at once for all three solutions. The start-up conditions are important in determining which steady state is eventually reached. This requires solution of the mass balance for the transient operation, viz. dC dt CC rCCCA AA AAAi= -()-()()=()009 t The results of computations pertaining to two CAi a re shownT ABLE 4W orksheet for Illustration 4 Equilibrium Composition Given V erify :. :. :. :. :. : :. :. ... .PKKy yyyyy PPPA BIABI 0120 00000 12 112 05 0 012045285045 025103 0 0850132 0515 ==== ==--= == + () rr rrr 5 5 012 12 15 2 2 0 05 012 2 02 1 2 1212 0450 05152 2 2850 006 y P yy FindA AB-()+ ()() -() -() ˆ =()== =rr rrr rrr r r rrrr .. : . 2 2 112 05 0 05 012 10 12 15 2 2 0 05 012 2 020 157 0515 2 0451 80510 05152 2 285 = + () -()-= + ()() -() -()--. .. .. .. .. . .rrr rr rrr rrr P y P yyA AB = =- = -()+ + ()= -()+ + ()= + + ()= +-4 70710 2 10505105 0510505 057 012 12 02 12 1 12 1. : .. : .. : .. : y y y y y yA A B B C Drr rr r rr r rr r 1 15 10505 2 105051 0505 0 0680083005402402 12 2 12 0 12. .. : .. : .. . . ()+ + ()= + + ()= + + ()===== r rr r rrrr yy y yyyyyEI I ABCDE284 2840271 1105 05110912 :.. y y yyyyyI A BCDEI= +++++==+ + = yrry

PAGE 19

W inter 2006 19T ABLE 5W orksheet for Illustration 5 Unsteady state CSTRWhich steady state do we reach? C1 = CA vCV V v rC C C fC CC rC Cf C solveCA A A A A AA A AsAA fC As 00 0 2 00213 10 1 75153576775218804425 2 1309325629587234809 8 1175316692890884749 :. : : : : : : . ====()= +() ()= -()-()=() ˆ ()t t = = ˆ ====()= -()+()=()== 0 0 0 00 10000020021305 1 10 7521 01 1 1 2 1 121tt NptsC DtC CC C C SolRkadaptCttNptsD CSoCifAi A Aiif AfNptsAf:. :. : :. ,: :,,,, : .,t :. :,,,, : .,C SolRkadaptCttNptsDCSolCAi AiifAfNptsAf 2 22222 1315 8 118 = =()== in Table 5. In this illustration and Illustration 6, integration of appropriate differential equations has been accomplished us ing the Runge-Kutta method with adaptive step size (Rkadapt). Let the steady state concentrations of A be denoted as CAs1, CAs2, and CAs3, with CAs1< CAs2< CAs3. The reactor operation started from CAi1 (very close to but less than CAs2) leads to the lowest concentration steady state ( CCAfAs 11 Table 5), while that started from CAi2 (very close to but greater than CAs2) leads to the highest concentration steady state ( CCAfAs 23 T able 5). The steady state corresponding to CAs2 is therefore unstable. Working with other values of CAi, the students deduce that for 0 < CAi < CAs2, CA converges to CAs1 at large times and for CAi > CAs2, CA converges to CAs3 at large times (additional computations not shown).Illustration 6This illustration pertains to a membrane reactor employed to obtain higher conversions for reversible reactions, and is adapted from Example 4-10 of Fogler.[10] A gas phase dissociation reaction ABC ¤+ is carried out in a steady-state plug flow reactor, the wall of which consists of a membrane which allows transport exclusively of B. The feed to the membrane reactor contains only A, with FA0 = 10 mol/min. The reactor and the feed are kept at 8.2 atm and 500 K. Since A and C remain in the reaction phase throughout the reactor, it follows from the reaction stoichiometry that FC = FA0 FA. As there are two independent unit operations (reaction and membrane separation), the two independent mass balances are those for A and B, viz. dF dV r dF dV rrAB B=-=-(), 10 The expressions for the volume-specific rates of reaction, r, and removal of B, rB, are[10] rkC CC K k F F C FFF F rkCk F F FFF kkCC C K kkCC P RTA BC C A T AAB T BBBB B T TAB T T C BBTT= ˆ =-() ===+()====( , , , 11 0 2 10 101 0 10011) ) with k = 0.7 min-1, KC = 0.05 mol/L, and kB = 1 min-1. It is desired to obtain profiles of FA and FB in a 300 L reactor via numerical integration of Eq. (10). The Mathcad worksheet

PAGE 20

20 Chemical Engineering EducationT ABLE 6W orksheet for Illustration 6 PRTC P RT C kKkFF kkCkkCC C K IC F FToTo CBAB ToBBTo To C A B:. :. : : :. :. :. : : : : : : ==== = ===== = = == ˆ 820082 50002 070051010000 111 0 0 Npts NptsVVFFFF DVF k F FF CF FF FF k F FF CF FF FifAB A A A A A A: :. :. & ,: =====()= + -()+() + -()+ 10000300012 1 1 02 12 01 02 2 1 1 02 12 01 0 F F k F FF SolRkadaptICVVNptsDVSolFSolFSol Equilibriumforreactiononlyoperation GF F FF CFB A ifAB A 2 2 1 2 02 123 1 1 02 12() +() =()=== -()= + :,,,, : : : : -()+() ()=- -()-()-()-()-()-() + FF FF GF substituteFFF F F F F F F F F solveFA A A 01 02 2 1201 1 1 1 2 1 2 1 1 1 2 1 2 120 4 0000000000000000000 10 20 20 4 10 20 102 ,. 5 5 1025 10255528 ˆ === : FFAeAe for this illustration is shown in Table 6. The students observe that the profile of FB exhibits a maximum (Table 6), since B is not supplied in the feed and is subject to two serial processes, namely generation by reaction and removal by membrane. If B is not removed by membrane separation (reaction-only operation, FB = FC = FA0 FA per reaction stoichiometry), working with the driving force for the reaction, the lowest FA (= FAe, corresponding to reaction equilibrium) is calculated via symbolic manipulations to be 5.528 mol/L (Table 6), which corresponds to 45% conversion of A. From the profile of FA in Table 6, the students observe that for the effluent from the membrane-wall reactor, FA is much lower than FAeand therefore the conversion of A is much higher. The last two illustrations deal with multiple reactions.Illustration 7This illustration, adopted from Example 6-7 of Fogler,[10]pertains to the series-parallel reactions MHXMrkCC XHTMerkCCeHM HX++= ++=(), 11 2212 with M, H, X, Me, and T being abbreviations for mesitylene, hydrogen, m-xylene, methane, and toluene, respectively. The reactions are carried out in a C STR. The feed contains only M and H. In view of the stoichiometry of these mole-conserving reactions, it can be deduced that the concentrations of species influencing the kinetics are related to one 0 100200300 0 5 10 F A F B V

PAGE 21

W inter 2006 21another as CaCC a CCCMHXMXH=+-()=+-()1 2 000213 Since there are two independent reactions, one needs to solve only two mass balances, e.g. those for hydrogen and m-xylene, in conjunction with the stoichiometric relation in Eq. (13). The Mathcad worksheet for solution of the design equations is shown in Table 7. The kinetic and operating parameter values are[10]: k1 = 55.2 (ft3/lb mol)0.5/h, k2 = 30.2 (ft3/lb mol)0.5/h, CH0 = 0.021 lb mol/ft3, and CM0 = 0.0105 lb mol/ft3. The profiles of CH and CX are shown in Table 7, with the space time for CSTR t cbeing in hours. The profiles reveal that the concentration of m-xylene, an intermediate, exhibits a maximum as expected, since it is not supplied in the feed. For each tc one has to provide initial guesses for CH and CX, and solve the mass balances iteratively. The same initial guesses may work for certain range of tc This happens to be the case in this illustration. The solutions of mass balances are therefore obtained using tc as a floating variable (Table 7). To verify the solution, the normalized residues associated with the mass balances for H and XreH and reX, respectivelyare calculated by substituting CH and CX generated by the solution into the mass balances. From the definitions of reH and reX and magnitudes of these displayed in the profiles in Table 7, it is evident that the profiles of CH and CX in T able 7 are indeed solutions of the mass balances.T ABLE 7W orksheet for Illustration 7 0 0.20.40.6 2 10 6 1 10 60 1 10 6 reH creX c c 0 0.20.40.6 0 0.02 0.04 C H c10C X c c kkCCC aCCCa CC CC kCaCCkHoMoXo MoXoHo HX HHo c HHX 12 1 05 255230200210010500 20 0 0089000312 05 :. :. :. :. :. : :. :. ..===== = +-= == -()() +-() -=Given t () () +-() ()()=()()=()()=()()= =CC C kCaCCkCC SolFindCCCSolCSol reH kCHX X c HHXHX cHXHccXcc 05 1 05 2 05 12 105 111. ... :, : : : t ttttt t t H HHXHX HHo HHXHaCCkCC CC reX kCaCCkC ttttt t t ttttt()() +()-()() ()()() ()-()= ()() +()-()() ()(05 2 05 1 05 205 1 05.. .. : ) )() ()-051.C CX Xt t

PAGE 22

22 Chemical Engineering Education T ABLE 8W orksheet for Illustration 8 CHHCT EE k E k E kTk EPA: : : :. : :. : : :. exp :.exp : exp ==-=-=== == = ˆ = ˆ ()= 300550007150003300001 990027000 33 1 987 1 300 458 1 987 1 5001200 12 10 1 20 2 110DDt 1 1 220 2 1 1 12 2 2 01 9871987 11 :exp : : ˆ ()= ˆ ()= ()+ ()() + ()+ ()() ()= -()T kTk E T GT kT kT HH kT kT RTCpTT difT t t t t DD ( ()=() ()-()=()=()=-()=-()=--:. .. .. .. .. GT RT dif dif dif dif dif 10 30959624910 35433278810 47385469710 54029581210 719585 5 6 61 1 027106. Illustration 8This illustration, adopted from Example 8-12 of Fogler,[10] pertains to elementary liquid phase reactions ABCkk 12 taking place in an adiabatic steady state CSTR. The expressions for CA and CB obtained from solution of mass balances for A and B are C C k C kC kA A B A= +()= +()()0 1 1 211 14 t t t The energy balance has the form (specific heats of all species being considered equal, CpA = CpB = CpC = Cp) CCTTkCHkCH kk E RT iAPAB ii i 001122 00 1215 -()++()= = ˆ =()DDt exp, Upon substituting Eq. (14) into the above, the master equation for the adiabatic reactor is obtained as GTRT RTCTT GT kT kT HH kT kTmmp()=()()=-()()=-()+()[]+()+()() (), ,0 1 1 12 2 21 1 16 t t t t DD with the reactor temperature T being the only unkno wn. The values of various parameters are: Cp=300 J/{mol.K}, D H1 = -55,000 J/mol, D H2 = -71,500 J/mol, CA0=0.3 mol/L, T0=300 K, t = 0.01 min, E1 = 9,900 cal/mol, E2 = 27,000 cal/mol, k1 = 3.03 min-1 at 300 K, k2=4.58 min-1 at 500 K. With the exception of Cp, all other parameter values have been taken from Fogler,[10] where Cp has been consid ered to be 200 J/{mol.K} and the reactor operation has be en considered to be nonadiabatic. The Mathcad worksheet for this illustration is shown in Table 8. By plotting G(T) and R(T) versus T, the students observe that the two profiles intersect at five T's for T > T0, implying existence of five steady states. The temperature at each steady state can be calculated via iterative solution of Eq. (16). Alternately, the relative error associated with Eq. (16), denoted as dif(T) in Table 8, can be calculated at various temperatures to directly zoom in on the steady state temperature. For the parameters under consideration, the steady state T values are 309.59, 354.33, 473.85, 540.29, and 719.58 K. 400600800 0 5 1041 1051.5 105 GT sRT sT s

PAGE 23

W inter 2006 23 CALL FOR PAPERSfor the Fall 2006 Graduate Education Issue ofChemical Engineering EducationWe invite articles on graduate education and research for our Fall 2006 issue. If you are interested in contributing, please send us your name, the subject of the contribution, and the tentative date of submission. Deadline for manuscript submission is April 1, 2006. Respond to: cee@che.ufl.eduDISCUSSIONThe students also use Matlab in parallel to Mathcad. Both packages are available on computers across the IIT campus and in the chemical engineering computer laboratory. The purpose of exposing students to different packages is to provide them with a broad spectrum of skills needed for solving engineering problems and to demonstrate the differences in the packages' capabilities for solving different engineering problems.[1] The students recognize that some of the problems can be formulated, but not solved, by hand. They can quickly develop worksheets for these problems and solve them, the emphasis thus being on understanding the fundamentals of the problems. Care must be taken to ensure that use of computational software enhances students' understanding and enriches their logic and problem-solving skills, rather than simply allowing them to solve problems with only a superficial understanding of the problems.[13] W ith this in mind, the undergraduate chemical reaction engineering course using this software at IIT includes handouts and tutorials providing an introduction to the software and to different numerical methods. Further, the author has integrated computational software throughout the course, with the use of software always following solution of related simpler problems by hand.[13]CONCLUSIONThe use of computational packages enhances teaching and learning, allowing the teacher to cover more material.[2, 14] In the process, the students learn more and faster and appreciate the course even more, while developing the skills and flexibility necessary for ready adoption of different software packages for professional activities in industry.[1, 4, 10, 14] The graphics capabilities of Mathcad help in quick visualization of results as well as in reinforcing expected results and understanding not-so-expected results. The capabilities of Mathcad in symbolic manipulations are of considerable use in developing analytical skills of students in solving complex problems. The time spent outside the courses on gaining further familiarity with different computational software and their applications will allow students to reap the benefits of these programs.[14]REFERENCES1.Al-Dahhan, M.H., "Computing in the Undergraduate ChE Curriculum," Chem. Eng. Ed. 29 (3), 198 (1995) 2.Abbas, A., and N. Al-Bastaki, "The Use of Software Tools for ChE Education: Students' Evaluations," Chem. Eng. Ed. 36 (3), 236 (2002) 3.Davis, R.A., and O.C. Sandall, "A Simple Analysis for Gas Separation Membrane Experiments," Chem. Eng. Ed ., 37 (1), 74 (2003) 4.Sandler, S.I., "Spreadsheets for Thermodynamics Instruction: Another Point of View," Chem. Eng. Ed. 31 (1), 18 (1997) 5. Aluko, M.E., and K.N. Ekechukwu, "Introducing Process Control Concepts to Senior Students Using Numerical Simulation," Chem. Eng. Ed. 33 (4), 310 (1999) 6.Chen, W.-I., "Rate Measurement with a Laboratory-Scale Tubular Reactor," Chem. Eng. Ed. 33 (3), 238 (1999) 7.Dickson, J.L., J.A. Hart IV, and W.-I. Chen, "Construction and Visualization of VLE Envelopes in Mathcad," Chem. Eng. Ed. 37 (1), 20 (2003) 8.Smith, W.R., M.Lisal, and R.W. Missen, "The Pitzer-Lee-Kesler-Teja (PLKT) Strategy and its Implementation by Meta-Computing Software," Chem. Eng. Ed. 35 (1), 68 (2001) 9.Brauner, N., M. Shacham, and M.B. Cutlip, "Computational Results How Reliable Are They? A Systematic Approach to Model Validation," Chem. Eng. Ed. 30 (1), 20 (1996) 10.Fogler, H.S., Elements of Chemical Reaction Engineering 3rd Ed., Prentice Hall PTR, Upper Saddle River, N.J. (1999) 11 Cutlip, M.B., and M. Shacham, Problem Solving in Chemical Engineering with Numerical Methods Prentice Hall PTR, Upper Saddle River, NJ (1999) 12.Froment, G.F., and K.B. Bischoff, Chemical Reactor Analysis and Design 2nd Ed., Wiley, New York (1990) 13.Dahm, K.D., R.P. Hesketh, and M.J. Savelski, "Is Process Simulation Used Effectively in ChE Courses?" Chem. Eng. Ed. 36 (3), 192 (2002) 14. Mackenzie, J.G., and M. Allen, "Mathematical Power Tools: Maple, Mathematica, Matlab, and Excel," Chem. Eng. Ed. 32 (2), 156 (1998)

PAGE 24

24 Chemical Engineering EducationProcess control education is a significant aspect of the chemical engineering curriculum, as it provides a fundamental basis for modern chemical process operation. The subject is highly applied yet rooted deeply in theory. Bridging the gap between the theory and application is often a difficult task, particularly in the classroom setting. Experimental laboratories have been shown to be useful in motivating students and reinforcing the information taught in the classroom,[1-4] often with the additional benefit of small-group learning.[5,6] The use of hands-on experimental laboratories that are closely tied to the traditional process control lecture course allows students to actually link the theoretical content of the courses to its use on real-world systems. For this reason, process control experiments have been developed across the country.[7-9]The development of useful, dynamic, process control experiments requires a number of considerations. Safety is the primary consideration because an environmentally friendly system that can be operated with minimal risk to both the equipment and the user is necessary. The ideal system would also be a cost-effective means to demonstrate the pertinent material with some industrial relevance. It should be of moderate complexity, as simple systems may be too trivial to motivate students while a full-scale industrial process may be too overwhelming. Giving it flexible configuration options will allow for its use in a variety of contexts. Reasonable process time constants are also essential so that the system dynamics are slow enough to demonstrate that process changes are not instantaneous, while also reacting quickly enough to limit student boredom when examining dynamic process transitions. Undergraduate students typically have very limited experience with dynamic systems since many undergraduate courses work under assumptions of steady-state operation. The use of the dynamic experiment(s) provides this experience and demonstrates all aspects of the textbook theory.[10-17]There are a number of well-designed, low-cost experiments available commercially, from vendors such as Lego, for useEXPERIMENTAL AIR-PRESSURE TANK SYSTEMSfor Process Control EducationCHRISTOPHER E. LONG, CHARLES E. HOLLAND, AND EDWARD P. GA TZKEUniversity of South Carolina Columbia, SC 29208 Copyright ChE Division of ASEE 2006 ChElaboratory Christopher E. Long is currently a Ph.D. candidate in the Department of Chemical Engineering at the University of South Carolina. His research interests lie in the field of process systems engineering, focusing specifically on the applications of nonconvex optimization to process control and identification. He holds a B.S. (2001) in chemical engineering from Clemson University. Charles E. Holland is the staff engineer for the Department of Chemical Engineering at the University of South Carolina. He earned both his B.S. (1997) and M.S. (2003) in chemical engineering at the University of South Carolina. He designed and built the experimental systems described in this article. Edward P. Gatzke is currently an assistant professor in the Department of Chemical Engineering at the University of South Carolina. His research examines a variety of topics in process systems engineering, including process identification and process control. He holds a B.ChE. (1995) from the Georgia Institute of Technology and a Ph.D. (2000) from the University of Delaware.

PAGE 25

W inter 2006 25in process control education.[18] These systems, however, fail to offer the flexibility to be utilized in many different contexts. Furthermore, they often fail to provide any semblance of being industrially relevant. At the University of South Carolina, both a simple, dynamic, nonlinear, two-tank, air-pressure system and a more complex, multivariable, four-tank, air-pressure system have been developed. These pressure-tank systems prove quite useful in process control education, as they address the objectives for an ideal process control experiment. Inspired by experimental liquid-level systems,[19-23] these experiments are exceptional instructional tools for chemical engineers. As opposed to liquid-level systems, in these systems pressure differences drive the flow. This variation removes the limitations in system flexibility typically associated with gravitydriven liquid systems. The two-tank system is quite portable, thus lending itself well to classroom and outreach demonstrations. A variety of undergraduate topics including openloop modeling and traditional single-input, single-output (SISO) closed-loop control strategies can be readily demonstrated on the two-tank system. The more complex, multiv ariable, four-tank system can be used in a small group setting to illustrate more advanced topics such as multiinput, multi-output (MIMO) modeling, interacting systems, and multivariable decoupling, to name a few. This paper presents a detailed description of both systems and summarizes their current and future uses for both educational and research purposes.THE TWO-TANK SYSTEMA compact, experimental, air-pressure tank system involving a pair of tanks in series has been developed (). A schematic and photograph of the system are provided in Figures 1 and 2. This section describes the system its elf as well as presenting its uses in the context of undergraduate process control education.System DescriptionThe two-tank pressure system is comprised of two constant-volume aluminum tanks assembled in series supported by aluminum framework (22 inches long 24 inches high 17 inches wide). The two cylindrical tanks are each a foot in length. Their diameters are two inches and one inch, respectively. Supply air enters the system through a single onehalf-inch, air-actuated, BadgerMeter control valve.[24] The air flows through quarter-inch tubing into the two tanks in series and exits to the atmosphere. A small muffler is utilized at the exit to reduce system noise. The tanks are separated by Swagelok[25] metering valves with repeatable vernier handles. This provides a means to accurately transform the system between various system configurations. Note that completely opening a valve between the two tanks effectively "joins" the tanks, resulting in one large tank of uniform pressure, as opposed to two tanks in series. Pressure measurements are available from each of the two pressure tanks. Gauges are installed on each tank to provide visual indications of the pressures while pressure transducers are used to more accurately measure and transmit pressure readings to a computer. The larger tank is also fitted with a small release valve that vents to the atmosphere. This valve can be used to create a disturbance on the system that might simulate a leak in the given tank, providing the opportunity to examine disturbance rejection as a possible control objective in addition to referP2P1T ank 2 T ank 1 V1 CV Muffler V2 Vd Figure 1. T wo-tank schematic. Figure 2. Photograph of the two-tank system.

PAGE 26

26 Chemical Engineering EducationFigure 3. Schematic of the four-tank pressure system. Figure 4. Photograph of the four-tank pressure system. High Pressure Air Feed TANK 1 TANK 3 T A N K 4 T A N K 2 # # $ % ( ) + / 0 1 2 3 3 4 6 6 ence tracking. The apparatus is equipped with a National Instruments Data Acquisition system which can be interfaced to both Matlab/Simulink[26] and LabVIEW.[27] A com plete materials listing can be obtained by contacting the authors. It should be noted that initially the control valve exhibited substantial hysteresis, making accurate modeling impossible. A valve positioner was required in order to generate reproducible open-loops results on the system. This also helps introduce students to cascade control and the complexity of real industrial systems. In the lab environment, the feed air pressure can be supplied in a more permanent manner from a compressor. On the other hand, small compressed-gas cylinders or lecture bottles can be used so that the system can be taken into the classroom for demonstrations. Similarly, a dedicated desktop computer can be used in the labs, while a laptop can be conveniently carried to the classroom.Educational UsesThis new experimental system is quite valuable for educational purposes. In the classroom setting, it lends itself well for demonstration to larger audiences. Alternatively, smaller groups can experiment with the system in a laboratory setting and reap the benefits of learning in a hands-on environment. The typical undergraduate class can be broken into small groups that can be rotated between the actual pressure-tank system and nearby computer labs. In the computer labs, students can use a high-fidelity model of the system to carry out simulation work that closely parallels what is to be done experimentally. This way, those entering the computer labs first can prepare for the actual experiment, while those that see the actual system first can later reaffirm what has been done experimentally. These advantages are supported by the rapid dynamics of the system. Note that the open-loop time constant is on the order of 30 seconds. In an extended class period, it is possible that numerous groups could get a substantial amount of time working with the apparatus.

PAGE 27

W inter 2006 27 Figure 5. Flow diagram for alternative configurations of the four-tank system. Using this system, many topics from the undergraduate process control curriculum can be illustrated. Open-loop modeling can be performed to identify both firstand second-order SISO models of the two tanks, depending on the configuration. Both frequencyand time-domain models can be considered, including input/output descriptions such as Autoregressive Moving Average (ARMA) models. Linearization of an available nonlinear first-principles model can also be carried out. Traditional closed-loop control methodologies such as Proportional-Integral-D erivative (PID) and Internal Model Control (IMC) can be implemented. Additionally, related topics such as closed-loop stability can be demonstrated.THE FOUR-TANK SYSTEMThis section describes the four-tank system in comparison to the two-tank apparatus. A schematic and photograph of the system are provided in Figures 3 and 4. This system's uses for undergraduate, intermediate, and advanced process control education are presented along with its utility in process systems engineering research.System DescriptionThe MIMO experimental system consists of four interconnected air tanks arranged in two parallel trains of two tanks, in series, built upon a steel framework. Each tank is 35 inches in length with diameters of 4 inches and 2.5 inches for the upstream and downstream tanks, respectively. Supply air flows into the system through two air-actuated BadgerMeter control valves which serve as the manipulated variables for the system. The air flows through copper tubing and the tanks before exiting to the atmosphere. Again, mufflers have been installed at the system exit to reduce the noise level. Specifically, the air flowing through control valve 1 (CV1) proceeds into tank 1 and subsequently into tank 2 downstream before exiting the system. Additionally, a portion of the flow from the control valve can be routed into the downstream tank of the adjacent train (tank 4). In a similar manner, control valve 2 (CV2) affects the pressure in tanks 3 and 4, with cross-flow effects on tank 2. Valves V14 and V32 are directly responsible for the cross-train flow. In some cases, the interacting nature of the system as a result of the cross-train flow leads to the presence of an adjustable, multivariable, right-half plane zero and inverse response. Physically, this is a result of the fast and direct response of the downstream t ank pressures to cross-train flow, in contrast to the slow indirect effects of the flow from the large upstream tanks into the smaller downstream tanks. The flow of air through the system is driven by pressure gradients. Check valves are not used, therefore air could flow back upstream provided that the pressure gradient is in the appropriate direction. (Similar liquid levels have limitations in these regards as the flow path is dictated by gravity.) The result is a more flexible, dynamic experiment. As with the two-tank system, the various tanks are separated by a number of Swagelok metering valves; their placement allows the system to be configured in a variety of ways. By opening or closing select valves between the tanks, the system can be quickly transformed from one configuration to another. The possible configurations include: a single tank of numerous possible sizes (depending on the number of tanks utilized), two to four tanks in series, a pair of tanks in parallel, and other setups that would have tanks in both series and parallel. For example, V14, V22, and CV2 can be completely closed, resulting in an SISO fourth-order system with air flowing through all tanks in series (see Figure 5b). Note that in the interest of saving laboratory space, the system is "folded" so that the smaller tanks are placed above the larger ones, leaving a system with total dimensions of 72 inches long, 22 inches high, and 22 inches wide.Educational UsesAlthough not portable enough to be taken to the classroom, this system is well suited for use in the laboratory environment. This apparatus can again be used for large group demonstrations or in a more personal setting for individual-to-

PAGE 28

28 Chemical Engineering EducationFigure 6. Students performing lab on the tank system. Figure 7. Simulink interface showing closed-loop control of the four-tank system. small-group work (see Figure 6). The multivariable, four-tank, pressure system can be configured in such a manner that it closely mimics the operation of the simple two-tank system, thus allowing one to demonstrate similar concepts. The additional complexity and flexibility of the four-tank system, however, also allow for its use in a wider variety of contexts, particularly with regard to its multivariable nature. The system can be configured such that one control valve acts as a measured disturbance into the downstream tankthus allowing for feedforward control. This configuration is shown in Figure 5a. Input/output modeling of multiple tanks in series can be carried out given the appropriate configuration, but MIMO modeling techniques such as continuous and discrete-time, linear-time-invariant (LTI), state-space approaches can also be applied. Interacting systems can be demonstrated as well as dynamic decoupling. The simulink interface showing PI control of the fourtank system is shown in Figure 7. In this feedback arrangement, the two downstream tank pressures are being controlled by manipulating the two control valves at the inlet. The disturbance rejection capabilities of this control scheme can be shown by simulating a leak in either of the upstream tanks or by changing the supply air pressure. In addition to aiding in the presentation and reinforcement of the undergraduate material, more advanced undergraduate and graduate topics can be covered using this sytem. Linear and nonlinear state and parameter estimation routines can be developed for the system. Advanced control schemes can be used including multivariable IMC, H• and linear Model Predictive Control (MPC). With some tank configurations, the system can exhibit a multivariable right-half plane zero thus inverse responsemotivating the examination of input directionality and control performance limitations.[16]Student AssignmentsFor illustrative purposes, two relevant assignments typically given to students in the undergraduate and advanced (intermediate and graduate-level) courses are provided. Undergraduate Assignment Configure the four-tank system into an SISO arrangement that involves two tanks in series. Develop a transfer function representation of the relationship between the control valve and the pressure of the downstream tank. Using this model, implement an Internal Model Control scheme on the system in Matlab/Simulink and test the closed-loop performance of the system by introducing both setpoint changes and disturbances. Advanced Assignment Configure the four-tank system into a 2-by-2 MIMO arrangement that involves two parallel trains of two tanks in series with cross flow. Consider the two downstream tanks as process outputs and the two control valves as the manipulated variables. Use

PAGE 29

W inter 2006 29subspace identification methods in Matlab to develop a linear state-space representation of the system. Using this model, implement a traditional Model Predictive Controller on the system and test the closed-loop performance of the system by introducing both setpoint changes and disturbances. Test the impact of the various tuning parameters on the stability and performance of the controller.These assignments exemplify those used in the different control courses. They provide students with the opportunity to explore the modeling and control the experimental pressuretank system. Again, note that in the interest of time, some students can develop their control methodology using a highfidelity process model as the system to be controlled before implementing their work on the actual system.Related ResearchIn addition to its utility in the instruction of process control theory, this fourtank system has potential for use in research in the field of systems engineering. To date, this particular system has been the focus of a number of research endeavors. For instance, system modeling is an important precursor to many advanced model-based control schemes. In limited regions of operation a simple linear model could suffice. Process nonlinearities, however, often require more complex model forms. The nature of this system is such that the process can exhibit hybrid dynamic behavior as the flow of air through the valves of the system can discretely switch between distinct, multiple, continuous regimes of operation. Under low pressure-drop conditions, the air flowrate across a given valve is dependent on both the upand downstream pressures. In high pressure-drop conditions, however, a sonic, or choke, flow regime is encountered in which the flowrate across a valve becomes solely dependent on the upstream pressure. The respective valve manufacturers, Swagelok[25]and BadgerMeter,[24] provide "hybrid" flow expressions based on first principles to capture these dynamics. For the BadgerMeter control valves the flow can be described by: qNC PP GT ifPPv a ga ba=" D 05 or qNC P GT ifPPv s ga ba=()£ 32 052 while for the Swagelok needle valve the flows can be described by: qNCP P P P PGT ifPPva aaga ba= ˆ 1 2 3 051 DD .() or qNCP GT ifPPva ga ba=£ 0 471 1 052 .() where q is a volumetric air flowrate across the valve at standard conditions, N is a numerical constant for units, Cvis the valve coefficient, Pa is the upstream pressure, Gg is the specific gravity of the fluid, and Ta is the temperature of the system. Temperature measurements are not available at the various points in the system. For convenience it is assumed that the temperature of the air in the system is approximately constant throughout. The first flow expression defines the low pressure drop regime where the flow across the valve is a function of both the upstream and downstream pressures. The second flow expression defines the choked flow regime where the downstream pressure has no influence on the flowrate. Under ideal conditions, these flow expressions can be used in conjunction with the ideal gas law to develop discrete-time models of the pressure in each tank. To model the rate of change of pressure in a given tank ( ) Pi the ideal gas law is assumed as the system is operated at both a reasonable temperature and pressure. () P nRT Vi i i= 3 where Vi is the volume of the tank, ni is the molar rate of change of air in the tank, R is the gas constant, and T is the temperature inside the tank. Provided that flow expressions define a volumetric flow across a va lve at standard conditions, the ideal gas law can be utilized a second time to convert to a molar flow across a valve. () n P RT qatm std= ˆ 4 where Patm is the standard (atmospheric) pressure, Tstd is the standard temperature, and again q is a volumetric flowrate. Thus () P P V T T qqi atm i std inout= ˆ ˆ -()5 PaPbqva,b Figure 8. Schematic showing the relationship of the pressures involved in calculating the gas flowrate across a valve. Adapted from Reference 25.

PAGE 30

30 Chemical Engineering Education 2000 2500 3000 3500 4000 15 20 25 30 Low P Model of P2 (psia) 2000 2500 3000 3500 4000 15 20 25 30 Hybrid Model of P2 (psia) 2000 2500 3000 3500 4000 0 20 40 60 80 Time (sec)Control Valve (% Open) model data model data Figure 9. Comparison of a fundamental low-pressure-drop flow model and a hybrid dynamic model in their ability to describe the pressure in a downstream tank. Based on this general expression, a discrete-time model of the system can be developed. Using the switching conditions prescribed by the valve manufacturers, a least squares regression can be performed to identify model coefficients that represent parameters such as the valve coefficients, temperature influences, etc. For the simple case of modeling the pressure within a single tank, the results are presented in Figure 9. It can be seen that the hybrid model that considers both lowpressure drop and choke flow regimes is better able to capture the system dynamics than a model based solely on lowpressure drop flow. Alternatively, mixed integer methods[28-30] can be used to develop strictly empirical hybrid descriptions of the process. Propositional logic can be used to formulate Mixed Integer Linear Programs (MILP) whose solution yields optimal coefficients and switching conditions for a variety of model forms including hybrid Volterra, autoregressive moving average (ARMA), and more general nonlinear state-space representations. On a similar note, six process states can be considered in the modeling of the dynamics of the system. The pressures in each of the four tanks can act as states in the model, as well as two states that are not so obvious. The placement of the two supplemental valves leading into the two larger tanks causes some resistance to air flow, regardless of their position. This, in effect, makes the small sections of entrance tubing between the control valves and the supplemental valves act as two additional but very small tanks. The pressure in these two regions will act as the remaining process states. No pressure measurements are available in the areas, yet the size of these "tanks" and the nature of the system imply that the associated dynamics are extremely fast. A set of ordinary differential equations (ODEs) can be developed for the tank system to describe each respective state. Under the assumptions that these two extra tanks exhibit fast dynamics in comparison to the rest of the system, however, an approximation can be made that reduces the respective ODEs to algebraic relationships as the derivative term can be approximated as

PAGE 31

W inter 2006 31zero. This leads to the use of a system of differential algebraic equations (DAE) to describe the system, as well as motivating studies in the area. Additionally, the system has been utilized as a testbed for the development of advanced control strategies. In one case, the prioritized objective inferential control of unmeasured process states is considered. The system is operated in a 2by-2 fashion with measurements of the downstream tank pressures available. The two upstream tank pressures are considered as the unmeasured process states to be controlled. Traditional MPC methods are often limited to the control of measured outputs and typically rely on a heuristic tuning to address the trade-off between satisfying different control objectives. A state-space modeling approach can be utilized to explicitly describe unmeasured process states. Using information from this state-explicit model and using propositional logic, a mixed-integer MPC algorithm[31] can be developed that relies on the online solution of an MILP or MIQP for the optimal control move. Such a formulation can allow for a more intuitive tuning in which control objectives, possibly involving unmeasured states, are met in order of their assigned priority.CONCLUSIONSChemical process control education is often limited by the availability of practical hands-on educational tools. Few industrially relevant systems are available that offer both reasonable size and cost while providing interesting dynamics with the flexibility to be used in numerous contexts. This paper describes two such systems that provide students with the opportunity to actually apply and demonstrate experimentally many of the theoretical concepts that are fundamental to the subject. A small, experimental, two-tank system has been developed for use as a tool in process control education. The size and simplicity of the system lend themselves well to particular use in the undergraduate classroom. A similar yet more complex multivariable four-tank has also been developed. Its flexibility enables its use in a variety of applications. Many aspects of both the undergraduate and graduate-level process control curriculum can be presented. Additionally, the system is the focus of a variety of interesting research problems. Among these are studies on the hybrid dynamic nature of the flow through the system, and the systems' use as a testbed for advanced control schemes such as prioritized objective MPC.ACKNOWLEDGMENTThe authors would like to acknowledge financial support from the National Science Foundation Early Career Development grant CTS-0238663.REFERENCES1.Doyle, F.J., III, E.P. Gatzke, and R.S. Parker, Process Control ModulesA Software Laboratory for Control Design Prentice Hall (1999) 2.Doyle, F.J., III, E.P. Gatzke, and R.S. Parker, "Practical Case Studies for Undergraduate Process Dynamics and Control Using the Process Control Modules, Comp. App. in Eng. Edu. 6 (3),181 (1998) 3.Jung, J.H., M. Lee, J. Lee, and C. Han, "A Development of Experimental Education Program: Computer Control of Multi-Stage Level Control System," Comp. Chem. Eng. 24 (2), 1497 (2000) 4.Marlin, T.E., "The Software Laboratory for Undergraduate Process Control Education," Comp. Chem. Eng. 20 S1371 (2000) 5.Millis, B.J., and P.G. Cottel, Cooperative Learning for Higher Education Faculty Oryx Press, Phoenix (1998) 6.Johnson, D.W., R.T. Johnson, and K.A. Smith, Active Learning: Cooperation in the College Classroom Interaction Book Co., Edina, MN, (1998) 7.Gatzke, E.P., R. Vadigepalli, E.S. Meadows, and F.J. Doyle III, "Experiences with an Experimental Project in a Graduate Control Course," Chem. Eng. Ed. 33 (4), 270 (1999) 8.Joseph, B., C. Ying, and D. Srinivasagupta, "A Laboratory to Supplement Courses in Process Control," Chem. Eng. Ed. 36 (1), 20 (2002) 9.Ang, S., and R.D. Braatz, "Experimental Projects for the Process Control Laboratory," Chem. Eng. Ed. 36 (3),182 (2002) 10.Ogunnaike, B.A., and W.H. Ray, Process Dynamics, Modeling, and Control Oxford University Press (1994) 11 Riggs, J.B., Chemical Process Control Ferret Publishing, 2nd Ed. (2001) 12.Seborg, D.E., T.F. Edgar, and D.A. Mellichamp, Process Dynamics and Control John Wiley and Sons (1989) 13. Astrom, K.J., and B. Wittenmark, Computer-Controlled Systems: Theory and Design Prentice Hall, Inc., 3rd Ed. (1997) 14.Bequette, B.W., Process Control: Modeling, Design, and Simulation Prentice Hall (2003) 15.Marlin, T.E., Process Control: Designing Processes and Control Systems for Dynamic Performance McGraw Hill, 2nd Ed. (2000) 16.Skogestad, S., and I. Postlewaite, Multivariable Feedback Control Analysis and Design John Wiley and Sons, New York, 1st Ed. (1996) 17.Stephanopoulus, G., Chemical Process Control: An Introduction to Theory and Practice Prentice Hall (1984) 18.Moor, S., P. Piergiovanni, and D. Keyser, Design-Build-Test: Flexible Process Control Kits for the Classroom," in Proceedings of the ASEE Annual Conference Nashville, TN (2003) 19.Johansson, K.H., and J.L.R. Nunes, "A Multivariable Laboratory Process with an Adjustable Zero," Proc. American Control Conf. 2045 2049, Philadelphia (1998) 20.Johansson, K.H., and A. Rantzer, "Multi-Loop Control of Minimum Phase Systems," Proc. American Control Conf. 33853389, Albuquerque, NM (1997) 21.Vadigepalli, R., E.P. Gatzke, and F.J. Doyle III, "Robust H-infinity Control of an Experimental 4-Tank System," Ind. Eng. Chem. Res. 40 (8), 1916 ( 2001) 22.Dai, L., and K.J. stršm, "Dynamic Matrix Control of a Quadruple T ank Process," Proceedings of the 14th IFAC 295300, Beijing (1999) 23.Gatzke, E.P., and F.J. Doyle III, "Use of Multiple Models and Qualitative Constraints for Online Moving-Horizon Disturbance Estimation and Fault Diagnosis," J. Proc. Cont. 12 (2), 339 (2002) 24.BadgerMeter, Inc.: Industrial Division, Tulsa, OK, Research Control V alves: Installation, Operation, and Maintenance Procedures 25.Swagelok Company, Swagelok: Valve Sizing August (2000) 26.The MathWorks, Matlab 6.5 Prentice Hall (2002) 27. LabVIEW 6.1 National Instruments Corporation (2003) 28.Roll, J., A. Bemporad, and L. Ljung, "Identification of Piecewise Affine Systems via Mixed-Integer Programming," Automatica 40 37 (2004) 29.Bemporad, A., and M. Morari, "Verification of Hybrid Systems via Mathematical Programming," Lecture Notes in Computer Science 1569 31 (1999) 30.Frerrari-Trecate, G., M. Muselli, D. Liberati, and M. Morari, "A Clustering Technique for the Identification of Piecewise Affine Systems," Automatica 39 205 (2003) 31.Long, C.E., and E.P. Gatzke, "A Mixed Integer Model Predictive Control Algorithm for the Prioritized Objective Inferential Control of Unmeasured States," Ind. and Eng. Chem. Res. 44 (10), 3575 (2005)

PAGE 32

32 Chemical Engineering EducationEngineering students can gain valuable benefits from an industry-sponsored project. Not only do students gain exposure to a full-scale chemical process, they also work closely with process engineers to collect and evaluate data. Students may even be allowed to collect data themselves by running product-testing equipment on-site. Once the data are analyzed, students can present their findings in a formal environment in front of industry personnel. Many chemical engineering programs provide opportunities for students to tour regional industries, thus exposing them to the complexities of a full-scale chemical process. Rarely, however, are students given the chance to do coursework on a real problem with an actual state-of-the-art industry process. Yet such experience is especially valuable to students who do not receive a co-op or internship opportunity. The University of Kentucky at Paducah has an advantageous location in close proximity to many industries. Calvert City, 17 miles east of Paducah, is home to 16 multinational industrial plants including Arkema Chemicals (formerly Atofina Chemicals), ISP Chemicals, Degussa Corporation, Celanese Chemicals, Westlake Vinyl Corporation, Wacker Polymer Systems, and Air Products and Chemicals. Many of these industries were involved in establishing the UK-Paducah engineering program, and now part icipate on an Industrial Advisory Board (IAB) that provides input into course content. Through the IAB, contact was made with one member interested in collaborating on a course project. Wacker Polymer Systems, whose manufacturing site is on the Air Products plant site, provided the opportunity for an industry project applicable to Introduction to Particle Technology, a course offered biannually to upper-level undergraduates. Air Products is a minority partner in a joint venture with Wacker Polymer Systems on the operation of a spray-dryer system. The system manufactures a powder used in dry-mix mortars and other construction-related products.PARTNERING WITH INDUSTRYfor a Meaningful Course Project Copyright ChE Division of ASEE 2006This column provides examples of cases in which students have gained knowledge, insight, and experience in the practice of chemical engineering while in an industrial setting. Summer internships and co-op assignments typify such experiences; however, reports of more unusual cases are also welcome. Description of the analytical tools used and the skills developed during the project should be emphasized. These examples should stimulate innovative approaches to bring real-world tools and experiences back to campus for integration into the curriculum. Please submit manuscripts to Professor W.J. Koros, Chemical Engineering Department, University of Texas, Austin, TX 78712. ChElearning in industryRHONDA LEE-DESAUTELSUniversity of Kentucky at Paducah Paducah, KYMARY BETH HUDSONW acker Specialties Calvert City, KYRALPH S. YOUNGAir Products and Chemicals, Inc. Calvert City, KY Rhonda Lee-Desautels is an assistant professor of chemical and materials engineering at the University of Kentucky at Paducah. She received her Ph.D. in 1994 from The Ohio State University, under the direction of L.-S. Fan. Before taking a position in academia, she was employed by International Paper for seven years. Her research areas include particle-particle interactions, gas-solid fluidization, and advanced battery materials. Mary Beth Hudson is the site manager of W acker Polymer Systems in Calvert City, Ky. She received a B.S. in chemical engineering from the University of Kentucky in 1989. She began her career as a process engineer for Air Products and Chemicals in 1989 and joined W acker Polymer Systems in her present role in 1998. Ralph Young is the environmental manager at the Air Products and Chemicals plant in Calvert City, Ky. He received a B.S. in chemical engineering from Cornell University in 1971 and an M.B.A. from State University of New Y ork (SUNY) at Buffalo in 19 81. In 1991 he received a master's in environmental technology from Murray State University in Murray, Ky.

PAGE 33

W inter 2006 33Three projects were identified that: were of interest to W acker; involved the spray-dryer system; and applied to the course content. One important project-selection criterion was that students would have the opportunity to perform particlesizing anal yses using the company's Beckman Coulter Counter laser diffraction analyzer. Therefore, each student would be involved in data collection on a real project, and would gain experience running a particle-sizing instrument. This industry project, taking the place of the usual term paper assignment, counted as 20% of the final grade. The requirements of the industry project were: to tour the process site; obtain all available data from sponsors; collect additional data; compile and analyze the data; formulate conclusions and recommendations; wr ite the report; and present to sponsors. One of the first steps was separating the 10 students enrolled in the courseall undergraduate seniorsinto one of the three projects. The industry projects were introduced during the fourth week of class, after students had been exposed to particlesize analysis, mixing and segregation of particles, and separation of particles from a gassubjects related to the three chosen projects. Given a form containing a short description of the projects, the students were asked to rank their interest in each. All students were then assigned to their first or second project choice. One project group had four students and the other two groups each had three. The industry tour of the spray-dryer process site (See Figure 1) took place during the fifth week of the course. The regular class meeting time was at 2 p.m. on Tuesdays/Thursdays for 75 minutes each. Arrangements were made to carpool on a Thursday to the Air Products plant site, leaving at the beginning of regular class time, and returning before 5 p.m. (one student had a 5 p.m. class). This three-hour time span allowed for 20 minutes travel to plant site, 30 minutes for introductions and a safety/orientation video, a one-hour plant tour, a 30-minute break-out session with engineers to discuss specific projects, and 20 minutes return travel. On the day of the tour, students were instructed to wear long pants, no opentoe shoes, and no sleeveless shirts. Our industry contacts provided flame-retardant smocks, hard hats, and safety glasses for the students at the plant site. After the tour, groups were responsible for making arrangements with a Wacker engineer for any experiments or analyses required by the projects.THE INDUSTRY PROJECTSFigure 2 shows a schematic of the Wacker spray-dryer process indicating the locations of the three projects.[1] In this process, the facility produces vinyl acetate-ethylene copolymer redispersible powders.[2] The conglomerated polymer powder that forms during the process is redispersed when contacted with water. These powders are used to improveFigure 2. Spray-dryer process flow diagram. Figure 1. The Wacker spray-dryer system in Calvert City, Ky.

PAGE 34

34 Chemical Engineering Education Figure 3. Student Melissa Barrett and Professor Lee-Desautels use the Beckman Coulter Counter at the plant. Figure 4. T ypes of agglomeration occurring throughout the spray dryer. Figure 5. Pneumatic ductwork to main baghouse. adhesion, impact resistance, flexible strength, water and freeze-thaw resistance, and abrasion resistance properties of Portland cement and other architectural coatings. In the process, polyvinyl alcohol (PVOH) is mixed with emulsions and fed to the spray dryer. High-pressure air and the solution are supplied to the top of the tower through spray nozzles. In the tower, water is driven from the mix leaving a dry powder at the bottom of the tower. The dried powder is pneumatically transported from the spray dryer to the main baghouse, where particles are separated from the gas before being transported to the product baghouse; there particles are screened and then stored in a silo. From the silo, the product powder is packaged and warehoused until delivery to the end user. Project 1. Nozzle Configuration versus Particle-Size Distribution (PSD) of Spray Dryer Product In the spray-dryer tower, polymer is supplied to the top of the tower through a high-pressure ring of spray nozzles. The high pressure forces the liquid droplets through a small orifice, causing them to atomize into a fine spray. The first project investigates the effect of the nozzle configurationthat is, the sequence of nozzles that are operationalto the final PSD of the product. Students measured the PSDs based on three different spray-nozzle configurations using the Beckman Coulter Counter (See Figure 3). Students compared the PSDs and analyzed the results based upon differences in trajectories between the various configurations. The students found little variation between sample distributions for the three nozzle configurations. Wacker provided an airflow model of the spray dryer to aid the students in their analysis.[3] The airflow model showed a vortex forming in the tower, causing much turbulence. The students attributed the small variation in PSDs to the presence of this highly turbulent vortex region, which formed in the tower independently of nozzle configuration. The students connected the project to their coursework by proposing the various forms of agglomeration that can occur throughout the tower (See Figure 4) with capillary (c) and droplet (d) occurring at the top of the tower, nearer to the atomized liquid spray, and pendular (a) and funicular (b) agglomeration dominating toward the bottom of the column, where much of the liquid has evaporated.[4]This student group recommended a study to maximize polymer feed to the tower without causing excessive agglomeration by controlling nozzle configuration, nozzle pressure, and airflow. Project 2. Baghouse Segregation Analysis Once the polymer powder has exited the spray tower, it has an average diameter of about 100 microns. It is mixed with clay particles (average size 60 microns) and pneumatically tran sported down flexible ductwork to the main baghouse. The main baghouse serves to separate the transport gas from the powder while controlling particulate emissions. The pneumatic ductwork splits into six separate ducts (labeled A, B, C, D, E, and F as shown in Figure 5) before entering the main baghouse. The second project involved analyzing the uniformity of particle loading on the main baghouse after the splitting of the ductwork. Samples were collected by industry personnel at each of the six separate ducts leading into the baghouse. The students analyzed the samples with the Beckman Coulter Counter and compared distributions. The students found that the mean particle size differed widely a mong the ducts. Duct A contained the largest particles at a median size of 159 microns; Duct B particles had a median size of 76 microns; Ducts E and F averaged 60 microns; and Ducts C and D averaged

PAGE 35

W inter 2006 35Figure 6. PSD of particles sampled from top of container. Figure 7. PSD of particles sampled from bottom of container. Note fraction of large particles Note lack of large particle fraction 45 microns. The students realized they couldn't explain these results in terms of inertial considerations alone, as the larger particles would be more likely to settle out when making the turn to Ducts A, B, E, and Fan effect that would lead to smaller particles in those ducts. The students decided they needed to gather more information about the ducting. On speaking with plant personnel, they were made aware that the lines had never been cleaned. The students also learned that directional plates had been installed in the transport lines to direct powder flow, but were nonfunctional due to buildup of wet productessentially "gluing" them in place. Students proposed in their analysis that blockage due to material buildup was occurring in the pneumatic lines, and proposed it was concentrated around Ducts C and D, creating a region of restricted flow and high pressure drop. This restriction to flow in turn resulted in the smaller average particle sizes in these ducts, they theorized. In addition to regular sampling of the transport lines to monitor particle distributions, the students recommended the directional plates in the ductwork be made operational to control powder fed to each duct. To prevent recurring problems, students proposed that since the majority of this buildup occurred during start-up of the process, developing stricter process start-up guidelines was recommended. Project 3. Product Segregation During Transport Once the powder has been sent through both sets of baghouses, it is transported to a silo where it is bagged and transported to consumers by truck. The third project investigated the segregation of powder product during the transport process. Some additional PVOH powder is added to the spray dryer product before reaching the product baghouse, and the company suspected some segregation might be occurring with handling and transport due to the PVOH having a smaller average particle size than the product. Having learned about the mechanisms of particle segregation,[4] students decided the mechanism of percolation was responsible due to the rise of coarse particles with agitation. To test if segregation could occur, the students used a RoT ap device to agitate a sample container for a given amount of time to simulate the transport process. The students then took samples from the top and bottom of the shaken sample container and measured PSDs in the Beckman Coulter. The students also had an unshaken control sample that was measured. They found that the control had little difference in particle-size distributions between the top and bottom samples, with mean sizes of 95 and 96 microns, respectively. The shaken samples showed a greater percentage of large particles in the top samples than in the bottom samples, indicating the percolation and coarse particle-rise phenomena. In one shaken sample, after shaking for 30 seconds particles removed from the top of the container had a mean size of 83 microns, while particles from the bottom had a mean size of 69 microns (See Figures 6 and 7).

PAGE 36

36 Chemical Engineering Education "One of the most valuable aspects of this assignment from an industry perspective was the Presentation to Plant Technical Professionals.' Many entry-level engineers do not have the communication skills to clearly share their ideas with technical management. In many cases, engineering supervisors spend significant amounts of time working with entry-level engineers on their presentation and communication skills." Industry feedback The students concluded that particle segregation is a negative effect for a product intended to meet certain requirements and specifications for its end use. Because this product had received no complaints, however, the students recommended no changes to the transportation of these powders. In spite of this concession, they furt her recommended making customers aware that this phenomenon occurs as a courtesy in case end users might want to homogenize the powder post-transport.PROJECT PRESENTATIONSAt the end of the semester, each team presented its project findings to industry personnel at a seminar held in the Air Products Engineering Building conference room. Attending the proceedings were the three industry participants plus an additional invited engineer. All students were required to participate in the presentation, and were given an outline on the required presentation format: Background (Define the System and the Problem) Experimental (What You Did to Collect Data) Results/Analysis (Present the Data and Analysis) Discussion (Your Interpretation of the Results) Conclusions RecommendationsThe students in each group took turns presenting portions of the findings and were graded on the quality of the visual aides and delivery. The conference room was equipped with state-of-the-art audiovisual equi pment including a projector and screen. The students were told to bring their presentations on a CD, with additional copies to hand out to industry attendees. Most students had never presented in this kind of corporate environment.INDUSTRY PERSPECTIVEIn an effort to capture the industry viewpoint on the project experience, industry participants were asked to submit comments on the project. Their comments are summarized below. The comments are valuable, not only for students, but also for faculty to gain insight into what qualities industry values from their engineering employees. From the responses, it is obvious that the industry participants looked at the project more as a way to prepare students for the workforce, offering words of advice and critique, than a means of obtaining free labor. The industry participants had a genuine desire to provide a distinctive learning experience for our engineering students.THE COMMENTSConcerning the Performance of the Students "From an industry perspective, I found the students enthusiastic and ready to do a hands-on' project. I'm not sure if everyone was trying to build their r esume, but each student approached the project with an open mind and was prepared to learn something new. They quickly learned how to operate the test equipment and collect useful data." "In most cases, once the newness' of running the Coulter Counter and other test equipment wore off, the tedium of repetitive testing and analysis was apparent. In this respect each student was exposed to real industrial experience: 10-25% new and exciting opportunities versus 75-90% less exciting work. Every student has their own threshold of tedious, repetitive work. These types of assignments provide the opportunity to help students decide career paths such as process engineering in a plant environment or research assignments in lab environments." "In this project, it was obvious each student had some prior presentation

PAGE 37

W inter 2006 37training and experience. Many engineering curricula include this training in their degree requirements. Project leaders divided the presentation so that it flowed logically and used graphics to help the audience understand the project and results." "The only element that was lacking in these presentations was the business case that would make or break a decision to allocate more resources. Since this aspect was not expected from the students, the technical staff was able to question the students and guide their thinking during the presentations. When the business case was made for a projectsuch as to increase plant production yields or benefit customersmany light bulbs seemed to go on in students' minds about the importance of the work. The interaction between students and industrial professionals was invaluable and one of the most important aspects of these projects."Concerning What is Valued in an Employee "One of the most valuable aspects of this assignment from an industry perspective was the Presentation to Plant Technical Professionals.' Many entry-level engineers do not have the communication skills to clearly share their ideas with technical management. In many cases, engineering supervisors spend significant amounts of time working with entry-level engineers on their presentation and communication skills." "Most new engineers get bogged down in project details and sophisticated analysis, and cannot summarize pros and cons to drive a management decision." "Key qualities I value in employees are: problem solving ability, creativity, communication, teamwork, ability to accomplish goals with minimal direction, initiative, dependability, time-management skills, and the ability to successfully manage multiple constraints. The students' analytical ability is proven by their successful completion of the engineering curriculum. This project allowed them to demonstrate the other key qualities above as well." "Among the biggest constraints in industry are time and personnel. We are expected to accomplish more with less. Therefore, we need goal-oriented employees who can drive projects to completion. I have seen many engineers spend too much time evaluating options in trying to find the best' solution, only to create more problems by not achieving anything. I was told as a young engineer that you will be seen as more successful if you attempt to solve a problem five times over a year and only succeed on the fifth try than if you spend the whole year developing the perfect solution for the first try." "We do not have clearly defined problems with one correct answer in our work environment. Often, data to analyze the problem are missing or incomplete. Resources such as money, personnel, and time are limited. Engineers are challenged to determine the best solution to the problem based on the information and resources at hand. There is always an economic impact that has to be evaluated."Concerning the Benefit to Industry "The results from the three projects reinforced our knowledge and confidence in what was happening." "The data will be useful to support the allocation of r esources to cleaning the ducting to the main baghouse, alleviate any concerns with nozzle configuration influencing final product quality, and increase awareness of product segregation with transport." "The particle-size data collected in these projects have been used to address customer issues associated with particle size. Examples are a recent modification to a powder grade to decrease particle size/increase bulk density in response to a bulk handling issue with one of our largest customers, and a recommendation of powder grades to address an application which will require a coarser particle size." "One of the main benefits to industry in participating in these programs is that we get a better introduction to the students who will be entering the job market."STUDENT EVALUATIONSThe students were asked to evaluate the industry project in the optional-items section of the evaluation form. Four queries were made on the project. Students were also asked to provide personal comments specifically about the industry project. Eight of the 10 students taking the course were present for the evaluation. Query 1. Rate your overall perception of the industry project. Response: One rated it outstanding, four rated it good, two rated it average, and one rated it poor. Query 2. The industry project has allowed me to learn more about a specific area of particle technology. Response: One rated it outstanding, four rated it good, two rated it average, and one rated it poor. Query 3. The industry project has helped me feel better prepared to seek employment with a company that manufactures/uses particles. Response: Two rated it outstanding, three rated it good, two rated it average, one rated it poor. Query 4. The industry project was a valuable component of the course. Response: One rated it outstanding, four rated it good, and three rated it as poor. Continued on page 53

PAGE 38

38 Chemical Engineering EducationThere is no such thing as certainty in scienceeverything we believe ultimately rests on unprovable assumptions and imprecise observations. Our current theory may seem to work beautifully, but if we really understand science we know that new data can overthrow it at any time. Nevertheless, if there's enough evidence to back it up, we can base predictions on it and sleep peacefully without worrying that we might be wrong. If I pick up a heavy object and drop it, I feel comfortable predicting that it will fall down. I can't prove Newton's theory of gravitational attraction and I'm clueless about why gravity works the way it does (as was Newton), but I'm confident that down is the way to bet. As much uncertainty as there may be in science (and by extension, engineering), there is far more in education. Students are infinitely more complex and unpredictable than cantilever beams and airborne projectiles and fruit flies. Even in education, however, there are some propositions that give you a great chance of coming out ahead if you bet on them often enough. I've got a few like that to offer you.STUDENTS If a student who fails a test claims afterwards that he/she really understood the material, then either he/she really didn't understand it or the test was unfair (too long, too tricky,...) The first one happens far more often than most students believe and the second far more than most professors believe. Students who argue vehemently for additional points on every test will have difficult lives as both students and professionals I also worry about their marriages. Students who routinely come up with bizarre but valid ways of approaching problems may struggle in school but will do very well as researchers and engineers (if they survive school). Students who drop out of engineering are on average no worse academically than students who stay in. We like to believe that our absurdly high dropout rates in engineering mean we are eliminating weak students and retaining good ones, but that's not how it goes. Lots of students who leave have fine academic records but just don't like what they see in our classes. (Don't bet against this oneI've got the data to back it up.)GOOD AND BAD TEACHERSAn engineering faculty member is a good teacher ( i.e. a teacher who motivates his/her students to learn and facilitates their learning) or a bad teacher ( i.e. a teacher who does not motivate or facilitate learning and may even interfere with it) if he or she: ( good ) gets all of his/her students actively involved in class and knows all of their names (or at least most of them in large classes). ( bad ) makes classes PowerPoint shows, or spends most of every period deriving equations, or puts high-level problems on exams that are qualitatively different from anything students have seen in class or on homework "to see if they can think for themselves."THE WAY TO BETRICHARD M. FELDERNorth Carolina State University Raleigh, NC 27695 Copyright ChE Division of ASEE 2006Random Thoughts . .The race is not always to the swift, nor the battle to the strong, but that's the way to bet .Damon Runyon Richard M. Felder is Hoechst Celanese Professor Emeritus of chemical engineering at North Carolina State University. He received his B.ChE. from City College of CUNY and his Ph.D. from Princeton. He is coauthor of the text Elementary Principles of Chemical Processes(Wiley, 2000) and codirector of the ASEE National Effective Teaching Institute.

PAGE 39

W inter 2006 39 ( good ) always has students waiting in the hall during office hours and coming with questions before and after class. ( bad ) uses words like spoonfeeding and coddling when talking about student-centered teaching methods (e.g., problem-based learning) and colleagues who use those methods, and dismisses all educational research as nonrigorous. I'm also betting that individuals who do this have never read an educational research study and could not name a journal that publishes them. ( good ) gets consistently excellent student ratings It's possible that the ratings are high because of easy grading or whatever other spin colleagues with lousy ratings put on it, but I'm betting (again with a lot of research backup) that a hi ghly rated engineering instructor would also show up as a good teacher in peer ratings and assessments of learning outcomes. ( bad ) gets consistently poor student ratings Someone who is regularly shredded by most students may claim it is because he is "rigorous" or "unwilling to lower standards," or because she "refuses to be an entertainer," and "the students don't appreciate me now, but after they graduate they'll see how good I was." Maybe, but if I always bet that those instructors are simply poor teachers I say I'll come out way ahead in the long run. ( good ) has students coming back years later saying what an outstanding teacher he or she was. I'd bet my life savings on this oneand I'd do so even if that individual has never gotten a grant or published a research paper.MISCELLANY Little or nothing meaningful will be accomplished at a faculty committee meeting The more frequently the committee has regularly scheduled meetings, the more I would be willing to bet on this one. Furthermore, the larger the committee, the less it will accomplish. New faculty members who get some formal training and/or mentorship will be better teachers and more successful researchers after two years than their counterparts who get the traditional amount of training and mentorship (none). More and more schools are choosing to bet my way by giving their new hires meaningful orientation and formal mentorship. Departments that decide to give tenure and promotion to qualified faculty members who focus on teaching and educational scholarship will have stronger teaching programs than they had before and their r esearch productivity and quality will not suffer A high school senior contemplating engineering will get a better education by avoiding schools where much of the administration and faculty think ABET is the enemy T extbooks with CD supplements will soon be replaced by interactive DVDs that may or may not have text supplements, which will lead to improved learning. The present generation of faculty and students may find the adjustment difficult, but the next generation will have no trouble with it at all. T raditional campus-based departments will find it increasingly hard to compete with excellent distance programs for good applicants. An online course that includes user-friendly interactive tutorials, electronic interactions between students and instructors and among students, and individual conferencing with the professor and tutors, provides a better educational experience than a campus-based course that is mostly chalk and talkand distance programs are getting better at those things all the time. More and more traditional engineering jobs will be handled by computers, technicians, and engineers in India and China (and Malaysia and Croatia and.. .). Graduates of schools that continue to focus on traditional content will have a harder and harder time finding and keeping jobs. Graduates of schools that focus more on entrepreneurship, critical and creative thinking, multidisciplinary project management, and global economics will do fine. I have undoubtedly tipped over some sacred cows here. Some of you will tell me that "Professor X dumbed his tests down and started to get great student evaluations," or "Professor Y's students burn her in effigy every year but as alumni they create multimillion-dollar endowments in her name," or "I can so name an educational research journal!!!" You don't have to send me an angry e-mail message about itI'll cheerfully concede right now that if I bet against Professor X or Y or against you I'll lose. In Vegas the casinos lose thousands of gambles every hour. They make many thousands of gambles, though, and the odds are with them. In the long run, they always win. All of the Random Thoughts columns are now available on the World Wide Web at http://www.ncsu.edu/effective_teaching and at http://che.ufl.edu/~cee/

PAGE 40

40 Chemical Engineering EducationThe inclusion of process control experiments in chemical engineering curriculums and the introduction of new process control experiments[1, 2, 3] indicate recognition of the importance of real-time experiments in process systems engineering. The experiments allow an instructor to reinforce and demonstrate theoretical systems concepts presented in lectures. Laboratory systems experiments in an academic setting provide students with an invaluable opportunity to familiarize themselves with important practical issues ( i.e. nonideality of industrial processes), such as processmodel mismatch, measurement noise, inadequate number of measurements, digital measurements, actuator saturation, unmeasured disturbances, and process nonlinearityissues often neglected in computer simulations. This manuscript describes a low-maintenance, low-safetyrisk, flexible, 0.9-m 1.5-m 2.4-m, pilot-scale setup that can be used for training students and carrying out research in process systems engineering. It briefly states typical applications of the setup. Detailed specific sample applications of the setup, together with real-time results, will be presented in forthcoming paper(s). The setup was built in the Department of Chemical and Biological Engineering at Drexel University and is located in the Process Systems Engineering Laboratory. The setup allows one to study a variety of processsystems engineering concepts such as design feasibility, de-A FLEXIBLE PILOT-SCALE SETUP FOR REAL-TIME STUDIESIN PROCESS SYSTEMS ENGINEERINGCHANIN PANJAPORNPON, NA THAN FLETCHER,* AND MASOUD SOROUSHDrexel University Philadelphia, PA 19104*Current address: Fluor Enterprises, Inc., Rose Tree II, Suite 5000, 1400 N. Providence Rd., Media, PA 19603 Copyright ChE Division of ASEE 2006 Chanin Panjapornpon is currently a Ph.D. candidate in the Department of Chemical and Biological Engineering at Drexel University. He received his B.Sc. from Chulalongkorn University, Thailand, in 1995 and his M.S. from Drexel University in 2002. His industrial experience includes five years with a petrochemical company in Thailand, and his research interests are in the areas of nonlinear model-based control, optimization, computer control, and controller-design software. Nathan W. Fletcher received his B.S. in chemical engineering from Drexel University in 1999. He was with Automation Application Inc., in Exton, Pa., from 1999 to 2004. He implemented DCS, PLC, and hybrid systems for the specialty chemical, oil and gas, pulp and paper, and food industries. In mid-2004, he joined Fluor Life Sciences in Media, Pa. His professional interests are in instrumentation and control. Masoud Soroush received a B.S. (chemical engineering, 1985) from Abadan Institute of T echnology, Iran, and two M.S. (chemical engineering, 1988, and electrical engineering: systems, 1991) and a Ph.D. (chemical engineering, 1992) from the University of Michigan. He is now a professor of chemical and biochemical engineering at Drexel University, and has worked as a visiting scientist at DuPont Marshall Lab, Philadelphia. His current research interests are in nonlinear modelbased control, high-temperature polymerization, nonlinear state and parameter estimation, fault detection and identification, and fuel-cell modeling, optimization, and control. ChElaboratory

PAGE 41

W inter 2006 41sign flexibility, co ntrol configuration selection, parameter estimation, process and instrument fault detection and identification, controller design and imp lementation, instrument calibration, and process modeling. Notable features of the setup are its flexibility and low safety risk (because it uses water only). The setup can be single-variable or multivariable, mildly or strongly nonlinear, interacting or noninteracting, and/or singleor multi-tank. It has features of both apparatus # 4 and 10 d escribed by Ang and Braatz[1]; it can be configured to be the same as apparatus #4 or 10, or a combination of apparatus #4 and 10. The setup can be used in both undergraduate and graduate process control laboratories to reinforce, through hands-on experiments, the concepts taught in process control and process analysis lectures.PILOT-SCALE SETUPA picture of the 0.9-m 1.5-m 2.4-m pilot-scale setup is shown in Figure 1a, and a schematic in Figure 1b. The setup has two identical, clear-plastic, cylindrical tanks. Each ta nk has an outside diameter of 0.2 m and a height of 1.0 m. The tanks can be connected to each other (by easy-connect/ disconnect flexible hoses) in several ways, which allows one to operate the setup as a system of a single tank, two parallel tanks, two interacting tanks in series, or two noninteracting tanks in series. The elevation of the second tank can be adjusted (via a jack) to alter the level of the interaction between the two tanks. Inside both tanks, there are helical copper tubes ( i.e. coiled copper tube banks) that can be used for heating or cooling, depending on the temperature of the water flowing into the copper tubes. One end of each copper tube is connected by a hose to a city water supply that is cold, hot, or a mixture of bothallowing adjustment of the inlet temperature of the water stream flowing into the copper tubes. Thermal energy can also be supplied to each tank by an electrical heater consisting of two heat cartridges inside the tank. Each tank has a variable-speed agitator. The setup has eight resistance temperature detectors (RTDs), two flowrate sensors, two level sensors, and one control valve. The RTDs measure the temperature of the inlet and outlet streams of the tanks and the cooling/heating copper tubes. The level sensors measure the level of water in the tanks. The flowrates of two inlet streams are measured by the two online and two off-line (rotameter) flow meters. A control valve adjusts the flowrate of a water stream flowing into Tank 1.ELECTRONIC HARDWAREAnalog Input DevicesEach of the sensors measures a process variable and generates a 4-20 mA analog signal, which is then sent to an analog input channel of a data acquisition board. The board then converts the analog signal to a digital signal. There are three Figure 1. The pilot-scale setup, in photograph (a) and schematic (b) a b

PAGE 42

42 Chemical Engineering EducationFigure 2. Front-end interface for the level control. types of analog instruments in this setup Pulse-Output Flow Meter. The paddle flow meter generates a positive on/off pulse signal when its rotor is rotated by the fluid flow. The pulse signal is then converted to 4-20 mA signal proportional to the flowrate (0-5 gpm). RTD Temperature Sensors. The resistance temperature detectors are connected to a Wheatstone bridge circuit that uses a reference r esistor of 100 ohms, which corresponds to 0 Celsius. Level Sensors. The level sensors used in the setup are of the pressure-transducer type. The liquid static pressure in a tank presses on the diaphragm of the transducer, generating a proportional analog signal.Analog Output DevicesThe proportional control valve is an analog output device that receives an analog input signal (4-20 mA) and sets the flowrate proportionally. It is a failto-close control valve. The power of the heaters is adjusted by a solid-state relay (SSR), which is connected to a pulse-controller module (both from O mega Engineering, Inc.). The module allows simple conversion of the on/off SSR to a proportional power regulator. Therefore, the average power to the heater is proportional to the input 4-20 mA analog signal to the module. Each of t he electrical heaters consists of two High Watt Density Cartridge Heaters (rated power of each: 1.5 kJ/s at 240 volts).Data-Acquisition BoardA DAS-1 701ST-DA data-acquisition board[4, 5] and an EXP-1800 extension board[6] (both from Keithley Instruments, Inc.) are used. The data-acquisition board has eight analog input channels. It receives and time-discretizes the incoming analog input signals. The expansion board allows one to expand each input channel of the data-acquisition board to eight input channels. Therefore, the data-acquisition system can support up to 64 analog input channels. It receives 12 analog signals from the eight temperature sensors, the two flow meters, and the two level sensors, and sends one analog signal to the control valve and two binary signals to the two heaters. The data-acquisition board can communicate with the central processing unit via interface software, such as Visual Basic, Visual C++ (Microsoft Corp.), and LabVIEW (National Instruments Corp.). The software then analyzes the data. A graphical user interface (GUI) is then used to present the data. In this setup, the data-acquisition application is developed by using Visual Basic as the

PAGE 43

W inter 2006 43Figure 3. Front-end interface for temperature control and data storage. F1 F1 T1 F2 T2 L2 T6 MeasurementAdjustable Variable L1 T5 He ater Tank 1 T4 T3 T8 Tank 2 T7 P2 P1 Figure 4. Adjustable and measured variables of the setup. fro nt end and C++ as the backbone. Visual Basic can be used to create a GUI easily, and C++ used to support Windows-based input/output operations. For this setup, frontend windows for temperature control, flow control, and data storage are developed. Two of the windows are shown in Figures 2 and 3. Data from the setup can be saved as Excel files and then be imported to Matlab (Mathworks, Inc.) easily.TYPICAL APPLICATIONS OF THE PILOT-SCALE SETUPW ith the flexibility to operate in various configurations, and its many sensors and actuators, the setup allows real-time study of a variety of process-systems engineering concepts. Figure 4 shows the variables that can be measured and/or adjusted in thi s setup. Below is a brief description of typical real-time studies that one can perform using the setup.Process ModelingGiven the online measurements, models including first-principles, empirical (black box), or hybrid (first-principles/empirical)[7] can be developed to describe water temperature and/or level in one or both tanks. In the case of empirical and hybrid modeling, the students can be taught model-parameter estimation as well. Hybrid model parameters include the resistances of the tank exit pipes as well as the overall heat transfer coefficients of the coiled copper tube banks. The model structure is obtained from mass and energy balances in the cases of first-principles and hybrid modeling, and f rom prior process knowledge (an assumption) in the case of empirical modeling. The empirical modeling can be off-line or online. In the latter case, one must use a model identification method.[7]Process Design AnalysisThe setup can be used to analyze the following process design aspects: Feasibility. Given desired steady-state values of temperature(s) and level(s), and nominal values of temperature and flowrate of the disturbance stream (inlet stream with no control valve), students are asked to evaluate theoretically and experimentally the feasibility of the design to operate at the desired steady state; that is, to check whether the design can provide heater power, water flowrate, and energy (through the heating/cooling coils) adequate to operate the process at the desired steady state.[8] For example, a desired water temperature below the city water temperature is definitely infeasible. Flexibility. Flexibility is feasibility in the presence of uncertainties such as disturbances and parameter uncertainties/variations. In this analysis, the students are asked to evaluate the feasibility of the design to operate at a given steady state when the temperature and flowrate of the disturbance stream vary within a given range.[8] Students can map theoretically the disturbance region in which the design is feasible and then verify the region experimentally.Process ControlThe setup can be used to carry out the following process control studies: Measurement Selection. Many control problems with one or more objectives can be posed, and students are

PAGE 44

44 Chemical Engineering EducationLaboratory systems experiments in an academic setting provide students with an invaluable opportunity to familiarize themselves with important practical issues (i.e., nonideality of industrial processes), such as process-model mismatch, measurement noise, inadequate number of measurements, digital measurements, actuator saturation, unmeasured disturbances, and process nonlinearityissues often neglected in computer simulations. then asked to list the measurements needed to achieve the control objective. These objectives include control of temperature and/or level in Tank 1, and/or control of temperature and/or level in Tank 2. For example, for control of temperature in Tank 1, at least, the temperature measurement T5 is needed. Control Configuration Selection. After choosing the necessary measurements, students can be asked to propose a set of manipulated inputs that can be used (adjusted) to realize the control objective(s). The controlled outputs should be controllable from the manipulated inputs. For example, temperatures in T anks 1 and 2 are controllable from heater power P1 and P2. The state and/or output controllability[7, 9] of the control configuration can be tested. Input-Output Pairing. For multi-input, multi-output (MIMO) control problems, students can be asked to pair the inputs and outputs of the selected control configuration so that completely decentralized control can be implemented. To evaluate the level of interactions among the process variables, students can use tools such as the relative gain array,[10]r elative orders,[11] and/or time delays to propose effective pairs.[12] Controller Selection. One can select a feedback or a feedback/feedforward control system depending on what control system is desired: completely decentralized (set of single-input, single-output, or SISO, controllers) or centralized (multivariable). For example, one can use the flow measurement F2 and the temperature measurement T2 (measurements of disturbance inputs), to add feedforward loop(s) to feedback control of temperature and/or level control in Tank 1. Furthermore, the controller can be: (1) a conventional controller, such as a proportional (P), a proportional-integral (PI), or a proportional-integralderivative (PID) controller; or (2) an advanced controller such as a model-based controller.[7, 10] The setup can be used to understand the limitations of decentralized control and implement decouplers in r educing the effect of interactions. Further, the modelbased controller can be analytical (such as an input-output linearizing controller) or numerical (such as a model-predictive controller). Whether conventional or not, adaptive features can be added to the controller.[7, 10] In real time, students can observe and compare the performance of different controllers, and evaluate the pros and cons of each.Parameter EstimationGiven the flowrate, level, and temperature measurements, students can estimate the heater powers and the heating/cooling coil-tank overall heat-transfer coefficients. In the case of the heater powers, since the heater powers are set by the computer, the values of the heater powers are known. This allows one to evaluate the accuracy of the estimated heater powers by comparing them to the actual values. A parameter estimator that can be implemented in real-time on this setup is described by Tatiraju and Soroush.[13]Fault Detection and IdentificationThe equipment can be used to demonstrate fault detection and identification. Sensor Fault Detection and Identification. The setup can be used to learn sensor fault detection and identification in real-time. Noise, drift, and/or bias are added to a sensor reading, and a sensor fault detection and identification method is then used to detect the fault in the sensor and identify the fault type (noise, drift, and/or bias). An example of such a study can be found in Mehranbod and Soroush,[14] in which sensors L1, F1, and F2, and a PI controller to control the liquid level in Tank 1, were considered. Process Fault Detection and Identification. Partial or complete failure of one of the process actuators is an example of a process fault. A process fault detection and identification method can be used to detect an actuator failure and the type of the failure. An intentional fault can be introduced in any of the actuators, and it can be detected and identified in real time by using a process fault detection and identification method.Instrument CalibrationThe setup has three actuators and 12 sensors. For each actuator, a calibration curve is obtained by finding the relation between the raw digital signal (that the computer sends to the data-acquisition board) and the actual value of the corresponding physical variable. For example, the control-valve cali-

PAGE 45

W inter 2006 45 012345 0 1000 2000 3000 4000 digital signalflow-rate (gpm) Figure 5. Calibration curve for the control valve.ing concepts in real time. Among these concepts are design feasibility, design flexibility, control configuration selection, parameter estimation, process and instrument fault detection and identification, controller design and implementation, instrument calibration, and process modeling. The setup can be used to provide graduate and undergraduate students with hands-on experience and to carry out research in process systems engineering.ACKNOWLEDGMENTSThe authors would like to thank Srinivas T atiraju, Neeraj Zambare, and Roberto Pena for their input into the project, and Dan Lau for his essential role in assembling the setup. The authors would also like to thank the Department of Chemical and Biological Engineering at Drexel University for supporting this project.REFERENCES1.Ang, S., and R.D. Braatz, "Experimental Projects for the Process Control Laboratory," Chem. Eng. Ed. 36 (3), 182 (2002) 2.Gatzke, E.P., R. Vadigepalli, E.S. Meadows, and F.J. Doyle III, "Experiences with an Experimental Project in a Graduate Control Course," Chem. Eng. Ed. 33 (4), 270 (1999) 3. Johansson, K.H., "The Quadruple-Tank Process: A Multivariable Laboratory Process with Adjustable Zero," IEEE Trans. Contr. Sys. Tech., 8 456 (2000) 4.Keithley Instruments, DAS-1700 Series User's Guide (1996) 5.Keithley Instruments, DAS-1700 Series Function Call Driver (1996) 6.Keithley Instruments, EXP-1800 User's Guide (1995) 7.Ogunnaike, B.A., and W.H. Ray, Process Dynamics, Modeling, and Control Oxford University Press, 1st Ed. (1994) 8.Grossmann, I.E., and M. Morari, "Operability, Resiliency, and Process Design Objectives for a Changing W orld," Proceedings of the 2nd Int. Conf. on Foundations of Computer-Aided Process Design Westerberg, A.W., and H.H.Chien, Eds., 931-1030 (1983) 9.Chen, C.-T., Linear System Theory and Design Holt, Rinehart, and Winston (1970) 10.Seborg, D.E., T.F. Edgar, and D.A. Mellichamp, Process Dynamics and Control 2nd Ed. (2003) 11 Daoutidis, P., and C. Kravaris, "Structural Evaluation of Control Configurations for Multivariable Nonlinear Processes," Chem. Eng. Sci ., 47 1091 (1992) 12.Holt, B.R., and M. Morari, "Design of Resilient Processing Plants-V: The Effect of Deadtime on Dynamic Resilience," Chem. Eng. Sci ., 40 1229 (1985) 13. T atiraju, S., and M. Soroush, "Parameter Estimator Design with Application to a Reactor," Ind. Eng. Chem. Research 37 (2), 455 (1998) 14.Mehranbod, N., and M. Soroush, "A Method of Sensor Fault Detection and Identification," J. of Process Contr ., 15 (3), 321 (2005) bration curve can be obtained by measuring the flowrate with the rotameter at different, constant, raw, digital signals set at the computer. For a sensor, a calibration curve is obtained by finding the relation between the raw digital signal that the computer receives from the data-acquisition board and the actual value of the corresponding physical variable. For example, an RTD is calibrated by placing it in beakers of water at different known temperatures and recording the value of the corresponding steady-state, raw, digital signal received by the computer. A typical calibration curve is presented in Figure 5. It shows how the flowrate of the water stream through the control valve depends on the raw digital signal.Calorimetric StudiesThe electrical heaters can be used to simulate heat of reactions. An exothermic reaction or set of exothermic reactions can be considered and simulated on the microcomputer, and the rate of heat production by the simulated reaction(s) is then sent to the heater to set the heater power to the calculated r ate of heat generation. Material and energy balances for the tanks, considered with the temperature and flowrate measurements, can then be used to estimate the power to the heaters; that is, the rate of heat production by the simulated reactions.CONCLUSIONSThis manuscript describes a low-maintenance, low-safety-risk, flexible, pilot-scale setup that can be used for training students and carrying out research in process systems engineering. It briefly states typical applications of the setup. Detailed specific sample applica tions of the setup together with real-time results will be presented in forthcoming paper(s). The setup allows one to study a variety of process-systems engineer-

PAGE 46

46 Chemical Engineering EducationThe polymer production or transformation industries employ a very significant percentage of chemical engineers. This justifies the presence of a variety of polymer science and engineering subjects in the chemical engineering undergraduate curricula. Topics on solid polymer mechanics, in particular, are often quite useful for future engineers. They establish the basic tools for evaluating whether a material is appropriate for an intended use, or to tune its performance by acting on the synthesis/processing conditions. These subjects are present in general polymer textbooks ( e.g. References 1 and 2). Because ChE student laboratories are not traditionally equipped with the machinery used for mechanical testing, however, introductory courses on this subject often suffer from not having an appropriate applied component. Therefore, students don't gain a hands-on understanding of the phenomena involved. One of the most commonly used mechanical tests in industry is tensile testing, in which the stress exerted by the material is measured at a constant strain rate. In addition to providing direct measurements of relevant properties, the stressstrain curves constitute "fingerprints" of a material's mechanical characteristics. The stress-strain curves are also often used for quality control of either raw materials or final products. Most mechanical testing machines can perform several standardized tests (tensile, compression, flexural, etc.) by using appropriate accessories. Such machines are designed for heavy loads, however, and are significantly expensive. In addition, in order to use such a "heavy duty" machine, one has to prepare a test specimen that will exhibit a measurable stress-strain behavior. Furthermore, the specimens must be cut or cast into a standard shape and dimension, which might not be easy for many materials of interest. Standard-shape polymeric specimens of known composition and molecular weight are commercially available, but at a cost. There is also equipment available for low stress/strain measurementssuitable for testing specimens of smaller dimensions, such as thin filmsbut these are also high-priced (about $8,000 USD). On the other hand, polymeric materials are readily available in the form of everyday-use items. Even informal observation of the behavior of these materials under mechanical solicitation may give the attentive student a wide variety of illustrations for important concepts in solid-polymer mechan-MECHANICAL TESTING OFCOMMON-USE POLYMERIC MATERIALSWITH AN IN-HOUSE-BUILT APPARATUSCRISTIANA PEDROSA, JOAQUIM MENDES, FERNO D. MAGALHESUniversity of Porto 4200-465 Porto, Portugal Copyright ChE Division of ASEE 2006 ChElaboratory Cristiana Pedrosa graduated in chemical engineering from the Faculty of Engineering of the University of Porto (FEUP), Portugal, in 2004. She is currently working as a research assistant on permeation measurements on porous materials. Joaquim G. Mendes is an assistant professor in the mechanical engineering department at FEUP, Portugal. He graduated in mechanical engineering from FEUP in 1988 and obtained a post-graduate degree in automation and management of industrial processes in 1989. He received his M.Sc. in industrial computing and his Ph.D. in industrial automation from FEUP and the University of Minho, respectively. His research interests in clude sensors, data acquisition, remote labs, and virtual instrumentation. Fern‹o D. Magalh‹es is an assistant professor in the chemical engineering deparment at FEUP, Portugal. He graduated in chemical engineering from the same faculty and received his Ph.D. from the University of Massachusetts, Amherst, in 1997. Among other courses, he is currently teaching an introductory course on polymer science and technology. His main research interests involvein addition to polymeric materials applied to the wood and paint industriesmass transport and sorption in porous solids and membranes.

PAGE 47

W inter 2006 47ics. An inspiring example was provided by J. Walker in the "Amateur Scientist" column of Scientific American magazine,[3] in which the molecular phenomena involved in the stretching of a polyethylene film are illustrated and discussed in a simple and captivating fashion. Our challenge was to build, on a very tight budget, a tensile-testing machine that chemical engineering undergraduate students could use freely in the lab for testing polymeric specimens gathered from common-use objects selected and prepared by them. Some of the design criteria we adopted for this project were: Keep costs as low as possible, without compromising the quality of the machine's measurements (i.e., r easonable accuracy and reproducibility). Keep in mind that the specimens should be easy to obtain and prepare. Using thin films for testing seemed to be a good idea. Many materials (plastic or otherwise) are commonly available in that form and can be easily cut into standard shapes. This option implies designing the machine for small loads and strains. Build a compact setup, so as to allow portability and use inside temperature-controlled chambers (e.g., r efrigerators and ovens). Make it fully automated, allowing total control of its functionalities through a computer-based dataacquisition system. Make it operationally robust, safe, and intuitive, since students are supposed to operate it themselves.This paper describes the machine developed, as well as some of the experiments performed. This project has been successful in giving students a hands-on perspective on some key aspects of the mechanical behavior of polymeric materials.SETUP OF THE TENSILE-TESTING MACHINEThe design of the testing machine comprised four key components: (1) a set of two grips, which firmly hold both ends of the test specimen; (2) a motor, which pulls one of the grips at a constant pre-set speed; (3) a force sensor, which measures continuously the force exerted on the material; and (4) a displacement sensor, which measures the distance traveled by the moving grip during the test. Figure 1 shows the original sketch of the machine's layout, comprising these components. Further details are discussed below. We intended to mostly test polymeric materials, such as polyethylene, in the form of thin films. We looked at some of the standardized tests used in industry to have an idea of the sizes and shapes of the test specimens used. According to ASTM D882-02,[4] the plastic films being tested are cut into rectangular specimens (at least 150 mm in length and 5 to 25 mm in width). On the other hand, on ASTM D237098(2002),[5] which applies to organic coatings ( e.g. elastomeric paint films), the specimens are also rectangular in shape, but the length may be lower (at least 50 mm in length and 13 to 25 mm in width). We decided to use a short specimen length to minimize the maximum strain involved in the tests, and thus allow the use of a reasonably priced continuous-displacement transducer and keep the machine's size small. Therefore, we adopted the latter standard. (Note that we actually also tested paint films with this machine.) Some crude preliminary tests done w ith film strips cut from supermarket plastic bags gave us the basic information for the specifications for the load sensor, the motor, and the displacement transducer. To provide linear motion to a lead screw, we chose a permanent magnet-stepper motor that employs a rotor with an internal thread. One end of the screw pulls one of the grips. The other end is attached to the moving lead of the LVDTtype displacement transducer. This transducer was the most expensive component in the setup (about 35% of the total cost). Cheaper alternatives are available, but an LVDT offers high accuracy and reproducibility, as well as wear-free andFigure 1. A 3-D sketch of the tensile-testing machine, showing the major components: (1) grips; (2) motor; (3) force sensors; (4) displacement sensor.

PAGE 48

48 Chemical Engineering Education Figure 3. Schematic representation of digital and analog signal flow between the machine and the data-acquisition system. Figure 2. Photograph of the tensile-testing machine (showing the holders used for the three-point bending test).T ABLE 1T ensile-Testing Machine Components (Not Including Data-Acquisition Board and Computer) Price (pre-tax, USD)Stepper motor + lead screw Mclennan L92411-P2 Max. linear force = 88 N $263 Driver board for stepper motor Eurocard $48 Displacement transducer Solartron DC50 920128 Range = 75 mm $463 (LVDT) Force sensors (2) Honeywell FSG Max. load p/sensor = 15 N 2 $75 (piezoresistive) Power supplies (2) EMS B811 and Astec LPS23 12 V $60 + $53 Holding structure and grips Local workshop $250 (carbon steel + polyacetal) Other components $33 TOTAL $1320 Component Maker and Model Specificationsfrictionless operation. Two force sensors, of piezoresistive type, are attached to the end of the lead screw. These sensors can measure only compression loads. The pulling grip is attached to a transversal bar that sits on top of the force sensors, thus transferring the tensile load, as is shown in Figure 1. Table 1 lists the main components and their costs. Figure 2 shows a photo of the actual unit, in its final working form. The machine can measure loads up to 30 N. The maximum strain rate is 300 mm/min and the maximum linear displacement is approximately 140 mm, but this value can be increased by using a longer lead screw combined with longer lateral support bars. The machine is monitored and controlled with a desktop computer using a data-acquisition card (National Instruments PCI-6014). The program LabVIEW 7.1 (from National Instruments) was used to develop the software that fully controls the apparatus and analyzes the measured data. Figure 3 summarizes the information flow between the computer and the machine. This machine was first implemented in an introductory course on solid polymer mechanics, which is an optional part of our chemical engineering undergraduate program. The students perform the tests themselves, on materials that they have gathered according to the instructor's suggestions. They analyze the results both qualitatively and quantitatively, computing different parameters from the measured data. Some representative tests and results are described in the ensuing text. For the sake of conciseness, the discussion in this paper is kept on a qualitative level.TENSILE AND TEAR TESTING HDPE FILMSStudents are asked to prepare test specimens from plastic shopping bags obtained at their favorite supermarkets. These are commonly made of high-density polyethylene (HDPE). Students can easily identify the polymer by noting the recycling symbol that is typically printed on the bags. The specimens to be used on the stress-strain tests are cut into 60 x 15 mm rectangles and reinforced with adhesive tape at their extremities, covering 20 mm on each end (see Figure 4a). The grips hold the specimens by grabbing on these reinforced ends. Special care must be taken in cutting these specimens. The cut should be perfectly straight and without indentations; im-

PAGE 49

W inter 2006 49 adhesive tape indentation a) b) 60 mm 15 mm40 mm “L“specim en: cut along longitudinal direction “T” spec imen: cut along tr ansversal direction tr ansversal directio n l ongitudinal directio n Figure 4. Specimen cut from plastic bag films: a) for stress-strain tests and b) for tear-strength tests. Figure 5. Definition of the longitudinal and transversal directions on a plastic bag. Figure 6. T ypical stress-strain curves obtained for highdensity polyethylene films from a plastic shopping bag. The two specimens were cut along two perpendicular directions (see description of L and T specimens in Figure 4). The end point on each curve corresponds to rupture of the film. Operating temperature = 20 C; strain rate = 200 mm/min; original film thickness = 0.016 mm. Note that the stress indicated is the "conventional" or "eng ineering" stress, i.e., the measured load divided by the initial crosssection of the specimen. Decreases in cross-section along the test are disregarded. 0 10 20 30 40 50 60 70 050 100150200250300350400 El ongation (%)Stress (N/m2) L s pecimen T s pecimen (MPa) perfections may cause the films to tear prematurely instead of reaching the ultimate rupture point. Sharp scissors or a fresh razor blade should be used. Students are asked to cut the specimens in two different ways: longitudinally and transversally in relation to the direction of the bags' "vertical" position (see Figure 5). Each specimen is labeled with a s oft-tip marker so that the information on the specimen's cutting orientation is not lost. Students also prepare specimens for tear-strength tests, which consist of 40 x 40 mm squares with an initial indentation (10 mm long) at the center of one of the sides (see Figure 4b). These indentations are done so that the tearing will propagate as intended: along the longitudinal or transversal directions mentioned before. Before performing the stress-strain tests, students measure the thickness of each film, using a digital micrometer. This information is used to compute the initial cross-sectional area of the specimen, on which the loading stress will be based. T ypical values are in the order of 10-2 mm. Figure 6 shows representative results for two specimens cut along perpendicular directions as described before. The distinct behavior presented by the two is quite noticeable. The specimen cut along the bag's "longitudinal" direction (L specimen) shows a rapid increase in stress, followed by a short plateau and afterwards a gentler increase, up to rupture. On the other hand, the T specimen (cut along the transversal direction), after a similar initial raise goes through a very well-defined maximum in stress and then stabilizes on an essentially constant value, almost up to the final rupture. The stress for this specimen is always significantly lower than for the L specimen. Another fundamental difference can be observed when, prior to performing the tests, horizontal lines (perpendicular to the direction of elongation) are drawn with a soft-tip marker at different sections along the specimens. As the L specimen is elongated, one can see that the lines increase almost identically in thickness, up to rupture; this indicates that the material is being uniformly deformed. On the other hand, on the T specimen some lines become noticeably thicker as others remain almost unchanged; this indicates that the specimen is being stretched at the expense of deforming the material in limited regions. As elongation continues, the extent of the undrawn regions successively decreases until the entire specimen becomes uniformly str etched and rupture occurs. Some students recognize this as being an example of cold drawing a phenomenon discussed in previous classes. It occurs on some semicrystalline polymers, like HDPE. The stress maximum corresponds to the yield point and the onset of necking. But why is cold drawing not observed on the L specimen?

PAGE 50

50 Chemical Engineering Education direction of tear propagation 0 1 2 3 4 5 020406 080100 Displaceme nt ( mm)Force (N) tr ansversal tearing l ongitudinal tearing Figure 7. Schematic representation of a tear-strength test. Figure 8. T ypical tear-strength curves obtained for polyethylene films cut from a plastic shopping bag. Tearing propagates along perpendicular directions on the two specimens (see description of transversal and longitudinal directions on Figure 4); same operating conditions as in Figure 5. fibrils lame llae Machine direction Transv ersal di rection Figure 9. Row-nucleated morphology of film-blown HDPE. The fibrils oriented along the machine direction act as nuclei for the growth of the lamellae (chainfolded crystalline structures).This mechanical anisotropy is not usually expected by students, and they are encouraged to offer explanations. The tearstrength tests that are performed afterward supply extra material for the discussion. The test is performed as schematized in Figure 7. The force exerted by the material is measured as the ends of the specimen are pulled at a constant rate and the tear propagates. T ypical results from this test are shown in Figure 8. When the tearing propagates along the longitudinal direction, the force is essentially constant and relatively low; in the end one observes that tearing originated a straight and clean cut. On the other hand, a much higher force is necessary to tear the specimen along the transversal direction; visually one can see that the material is stretched and distorted during the test and the final cut shows permanent deformation of the material at the edges. After analyzing this second set of results, students often suggest that this anisotropy is associated to a particular molecular orientation of the polyethylene chains in the shopping bag. It becomes clear that this is the correct hypothesis after the instructor describes the manufacture process for HDPE bags, commonly known as blown-film extrusion This process is described in several processing handbooks.[6, 7] It involves submitting the polymer to a sequence of transformations: melting, extrusion, blowing, drawing, cutting, and sealing. A continuously extruded thin-polymer tube is inflated by a jet of air blown into it. The blowup ratio (defined as the ratio between the diameters of the expanded film bubble and the die) controls the molecular orientation along the transversal direction. This ratio is usually 2 to 4. On the other hand, the drawdown ratio (the ratio between the speeds of the film at the nip rolls and at the die exit) determines the longitudinal orientation (called machine direction). A balance between these two parameters governs the final orientation within the film. Partial crystallization occurs as the material is cooled, thus conserving the molecular orientation imposed. This flow-induced crystallization is actually a bit more complex than it would seem, due to the particular morphology that polymers exhibit upon crystallization. At a sufficiently high drawdown ratio, film-blown HDPE undergoes a so-called r ow-nucleated crystallization[8]: Extended molecular chains oriented along the machine direction form fibrillar structures that act as nuclei for the crystallization of the bulk material, in the form of radially grown lamellae. Figure 9 schematically illustrates these structures.

PAGE 51

W inter 2006 51 0 5 10 15 20 25 012 345 Defl exion (mm)Force (N) T = 20 C v = 100 mm/min T = 50 C v = 100 mm/min T = 80 C v = 100 mm/min T = 20 C v = 2 mm/min Figure 10. Schematic representation of a three-point bending test. The two holders at the extremities are connected to the machine's upper grip and move at a constant rate. The holder at the center is attached to the lower grip and remains stationary. The two holders at the extremities are 30 mm apart. The specimen length is about 50 mm. deflexion Figure 11. T ypical force vs. deflexion curves obtained for polystyrene coffee stirrers. The labels indicate the specimens' temperature in Celsius (T) and the deflexion rate (v). The end point on each curve corresponds to rupture of the material.These long fibrils with perpendicular growths are often appropriately described as shish kebabs. They are responsible for the high tensile strength of the material along the machine direction (corresponding to the response of the L specimen in Figure 6). Because, in the case of HDPE, there is no significant interconnection between the "kebabs" of adjacent fibrils, the tensile strength in the transversal direction is significantly lower. This transversal straining causes a noticeable yielding of the material, associated to local fibrillar reorientation toward the direction of the applied strain. The stress remains typically constant along this drawing process (see the curve for the T specimen in Figure 6). When the fibrillar rearrangement has extended throughout the entire material, the stress often rises slightly and rupture occurs shortly after. This cold-drawing phenomenon is characteristic of many semicrystalline polymers and is described in most polymer science textbooks. A recent paper by Zhang, et al,[9] provides an interesting discussion of the mechanical anisotropy and crystalline morphologies of different kinds of polyethylene-blown films. From Figure 6 we see that the tensile strength (defined as the maximum stress measured during the test) is about 2.5 times higher when the material is strained along the direction that coincides with the fibrillar orientation (which we named the longitudinal direction). It makes a lot of sense that plastic bags are assembled so that normal use implies applying the stress along this direction. The tear-strength test results (Figure 8) further confirm these findings. It is clear that it will be much easier to tear the material along the direction of the fibrils than along the transversal direction. The "cherry on top" for this set of experiments comes when the instructor suggests that students take a piece of the plastic bag and heat it above a flame lighter (holding it with pincers and taking care to not actually burn the material). Immediately they see that the film starts to crumple and shrink. The temperature rise causes the gradual melting of the crystalline regions, loosening the mobility of the polymeric chains (the melting temperature of polyethylene is about 140 C). This allows the originally extended chains to rearrange toward the more favorable coiled conformation. The result is the crumpling of the polyethylene film. Further heating would cause the total disappearance of the crystalline regions, resulting in a polymer melt.FLEXURAL TESTING OF PS BARSThe machine is limited to tensile testing of thin films of relatively soft materials. Glassy polymers cannot be tested, since they involve much higher tensile stresses. Nonetheless, we have adapted the machine to perform a different kind of test on such materials: a three-point bending test at constant deflexion rate. We used polystyrene (PS) coffee stirrers, collected from a nearby vending machine, as test specimens. Figure 10 schematizes how the test is implemented: by using hard-wire hooks to attach the specimen to the grips. PS is glassy at room temperature (its glass-transition temperature, Tg, is about 100 C). Figure 11 shows some of the results obtained for different temperatures and deflexion rates. Since it is faster to place the rigid PS specimen on the support hooks than it was to attach a film to the grips, these tests did not involve placing the machine in a temperature-controlled chamber. Rather, the specimens were stabilized in an oven at the desired temperature. Prior to testing, each specimen was quickly transferred to the machine. The entire procedure (including performing the test) took no more than 30 seconds. In the figure one can see that the maximum deflexion is relatively low, as expected from a glassy polymer. For a deflexion rate of 100 mm/min, as the specimen temperature approaches the glass-transition temperature, its softness is significantly increased. Students

PAGE 52

52 Chemical Engineering Educationare asked to visually inspect the broken specimens. They notice that the ones tested at higher temperature show a visibly higher extension of crazed material (crazing i.e., the appearance of semi-opaque transversal bands in the neighborhood of the break surface, is a localized molecular-orientation phenomenon that occurs when some glassy polymers are close to rupture). Note that the measured values of the deflexion at break are not reproducible and should not be considered. In the many tests performed, some discrepancies were obtained for this parameter. This was probably due to sample heterogeneity. The force-deflexion curves measured at lower deflexions were always quite reproducible. When the test is repeated at room temperature, but at a lower deflexion rate (2 mm/min), the material's stiffness decreases, coincidentally giving a curve similar to the one previously obtained at 50 C. This is a good illustration of time/temperature equivalence. In polymeric materials, molecular response is highly time and temperature dependent.CONCLUSIONThe in-house-built tensile-testing machine proved to be an economical tool for allowing students to test the mechanical behavior of different polymeric materials. The results can be analyzed both quantitatively and qualitatively. The fact that the test specimens can be obtained from everyday-use materials is not only an economic advantage but also an added factor of interest for students, since they can do the material selection and preparation themselves. The pedagogical benefits obtained from direct experimentation were confirmed by the interest and motivation shown by our students. Awareness that the machine was built in-house actually seemed to raise the students' curiosity toward comparing its components to the ones used in the commercial models. The results shown here are representative of some important aspects of the mechanical behavior of solid polymers, such as the influence of processing conditions and the effects of temperature or strain rate. Stud ents are encouraged to analyze their results and provide interpretations on a molecular level. Other materials used successfully with this machine include rubber bands and films of elastomeric wall coating (EWC). The latter constitute a type of paint that, due its elastomeric character, is able to protect cracked walls from rain damage, since the film stretches to keep the gaps covered. It is an enriching exercise to analyze the mechanical response of EWC films under conditions such as low temperatures, aging (UV degradation), or water exposure (plasticization). The tests performed with this device can be used either as an illustration of the concepts and phenomena previously discussed in class or, perhaps more interestingly, in a reversed approach. Indeed, the experimental observations are quite thought-provoking and motivate students to ponder and hypothesize on the reasons for the results obtainedthus paving the way for a structured discussion led by the instructor. It must be remarked that the tensile and tearing tests described here for polymeric films are actually similar to some of the standard industrial practices, both in terms of specimen dimensions and operation parameters. It can be noted, as an example, that a data sheet for blown film obtained from ExxonMobil's HDPE resin HTA 001HD[10] (recommended for shopping bags, among other uses) reports an elongation at break of about 380% and a tensile strength (stress at break) of 56 MPa (it does not exhibit anisotropy for the particular processing conditions employed); this result is quite consistent with the values obtained with our shopping-bag material (of unknown origin). In addition, a fairly good reproducibility is obtained with our in-house-built machine. The only problems are usually associated with discrepancies in the rupture points. For stressstrain tests, this is due to premature breaking caused by tears that initiated at imperfections in the specimen side cuttings, as mentioned before. These anomalous rupture behaviors can be easily detected visually and can be minimized by cutting the samples carefully. In the flexural tests, rupture discrepancies are probably associated to imperfections or inhomogeneities among samples. Naturally, the machine presents limitations when compared to the models used industrially, namely in terms of measurement accuracies and limitations on load and strain ranges.ACKNOWLEDGMENTThe authors would like to thank the Department of Chemical Engineering of the Faculty of Engineering of the University of Porto for the financial support provided for assembling this machine.REFERENCES1.Kumar, A., and R.K. Gupta, Fundamentals of Polymers McGrawHill International Editions, New York, 376 (1998) 2.Sperling, L.H., Introduction to Physical Polymer Science 3rd Ed., W iley-Interscience, New York, 477 (2001) 3.Walker, J., "The Amateur Scientist," Scientific American February 86 (1990) 4.ASTM D882-02, Standard Test Method for Tensile Properties of Thin Plastic Sheeting ASTM, Philadelphia 5.ASTM D2370-98(2002), Standard Test Method for Tensile Properties of Organic Coatings ASTM, Philadelphia 6.Rosato, D.V., Extruding PlasticA Practical Processing Handbook Springer-Verlag, New York, 315 (1998) 7. Crawford, R.J., Plastics Engineering 3rd Ed., Elsevier, Amsterdam, 265 (1998) 8.Kumar, A., and R.K. Gupta, Fundamentals of Polymers McGrawHill International Editions, New York, 340 (1998) 9.Zhang, X.M., S. Elkoun, and M.A. Huneault, "Oriented Structure and Anisotropy Properties of Polymer Blown Films: HDPE, LLDPE and LDPE," Polymer 45 217, (2004) 10.

PAGE 53

W inter 2006 53"It would be beneficial to our understanding of particle technology if we were allowed a more hands-on approach rather than analyzing given data." Student feedbackPartnering With IndustryContinued from page 37 Instructor comment: I suspect the three students who rated this query as poor may have been reflecting on how valuable they felt their work was to Wacker and Air Products. This perception is expressed in the student comment #3 below. Three students provided personal comments of the industry project on the evaluation form: Comment #1: "I think the project would have gone better if we were able to run the equipment and take the samples ourselves." Comment #2: "It would be beneficial to our understanding of particle technology if we were allowed a more hands-on approach rather than analyzing given data."Instructor comment: I believe these two students were referring to collecting samples from the process, as all students were required to run the Beckman Coulter Counter. Comment #3: "I thought it was neat to see an actual application of particles, but I didn't feel we actually accomplished anything."Comments from Industry Participants on evaluation results: "I thought the feedback from the students was interesting and very candid. The students that rated the exercise as fair to poor shouldn't be viewed negatively, but rather that their engineering interests might lie in marketing, sales, or areas other than manufacturing." "Many students saw this project as a research study or make-work' study with no commercial application or contribution to a company's profit. When we started to connect the dots to commercial applications during the presentations and relate to benefits for the company, many students felt better about the project and started to appreciate their contributions."CONCLUSIONS AND RECOMMENDATIONSIt is obvious from the feedback that certain students were frustrated with the amount of contact they had with the process, and didn't perceive any benefit to the company from the projects. Benefits weren't discussed until the end of the projects, in the presentation phase, which, in retrospect, was too late. In the future, it would be better to introduce benefits earlier in the execution of the projects. This might be best accomplished by having the industry personnel visit the classroom and introduce projects themselves, including potential benefits for the company. The students, however, should also be made to realize that these projects are chosen partially for the benefit of the industry, but the main driving force is to provide the students with a real-world learning experience. T wo of the biggest challenges of this exercise were: (1) finding industry projects that could feasibly be completed by the students in the project time frame, and (2) finding three projects requiring a comparable quality of student experience. As is obvious from these three projects, one resulted in a better student experience than others. In Project 3, the students had more project participation since they were able to plan and run experiments using the Ro-Tap machine, as well as run the particle analyzer. Projects 1 and 2, on the other hand, were straightforward as far as obtaining samples, which were collected by industry personnel, and the students' only participation in data collection was running the particle analyzer on the samples. In the future, this deficiency could be overcome by suggesting students shadow the industry participants during procedures that they can't perform themselves due to safety and liability issues. Also, more pre-planning by the instructor to assure better equity of the project experience may be necessary (initiation of partnership occurred in July, with the course beginning in August). By the very nature of the projects being based on unanswered questions about the process, however, it would be impossible to predict project results and effects in this scenario. Overall, the majority of the students felt the industry project was beneficial to their careers and experience. The project accomplished the main goals of (1) exposing students to a real-life particle manufacturing process, (2) gaining handson experience r unning a state-of-the-art particle measuring device, and (3) applying the basic concepts presented in the course.ACKNOWLEDGMENTWe wish to acknowledge the support given by Josh Brien, W acker engineer, in assisting students with data collection.REFERENCES1.State of Kentucky Title V Permit No. V-99-057 for Wacker Polymer Systems Spray Dryer Plant at Calvert City. 2.Wacker Polymer Systems, VINNAPAS: "Redispersible Powders and Dispersions Product Brochure," Nr. 5838-5838 (USA) 04 (2001) 3.Wacker Polymer Systems, Air-Flow Model for Spray Dryer Process Burghausen, Germany (not formally published). 4.Rhodes, Martin, Introduction to Particle Technology John Wiley and Sons, West Sussex, England (1998)

PAGE 54

54 Chemical Engineering EducationThe laboratory course in process control constitutes an important component of an undergraduate chemical engineer's education because it provides hands-on training in the application of process control to real processes. T he laboratory course exposes the student to industrial process control hardware and the impact of measurement noise and unmeasured disturbances upon the control of real processes. In most university courses these laboratories are essentially linear single-input, single-output (SISO) unit operations. Until recently, the Department of Chemical and Petroleum Engineering at the University of Calgary was no exception. Yet such SISO control laboratories do not expose the student to the complexities of nonlinear or multi-input, multi-output (MIMO) processes. A few laboratories in the literature[1-4] have attempted to address these shortcomings. Rivera at Arizona State University[1] describes a salt-mixing laboratory that examines the concentration dynamics at different tank levels using system identification techniques in a first process dynamics and control course. Fisher and Shah at the University of Alberta[2]describe a complex three-tank-level plus temperature arrangement that allows MIMO processes and process nonlinearity to be studied at the senior undergraduate or first-year-graduate course level. Braatz, et al., at the University of Illinois[3, 4]describe a nonlinear but SISO pH neutralization process and a quadruple-tank apparatus that illustrates time-varying dynamics for a senior undergraduate process control course. In this paper we describe a relatively simple salt-mixing laboratory in the undergraduate chemical engineering process control course at the University of Calgary that allows students to study both MIMO behavior and nonlinearity.THE UNIVERSITY OF CALGARY'S PROCESS CONTROL COURSEThe University of Calgary requires process dynamics and control as part of the degree requirements for undergraduate students in chemical engineering, in a course that pioneered the hands-on, real-time (time domain) approach to teaching process dynamics and control.[5] Students in the class employ dynamic process simulation using a dynamic process simulator, such as HYSYS or Aspen Dynamics,[6] to model chemi-A NONLINEAR, MULTI-INPUT, MULTI-OUTPUTProcess Control Laboratory Experiment ChElaboratoryBRENT R. YOUNG, JAMES H. VAN DER LEE, AND WILLIAM Y. SVRCEKUniversity of Calgary Calgary, Alberta T2N 1N4, Canada Copyright ChE Division of ASEE 2006 James van der Lee is a software engineer with Virtual Materials Group, Inc., Calgary, Alberta, Canada. He received his B.Sc. degree in chemical and petroleum engineering from the University of Calgary in 1999 and successfully defe nded his Ph.D. thesis in 2004. He was instrumental in the design of the new laboratory while a graduate student. W illiam Svrcek is a full professor of chemical and petroleum engineering at the University of Calgary, Alberta, Canada. He received his B.Sc. (1962) and Ph.D. (1967) degrees in chemical engineering from the University of Alberta, Edmonton, Canada. Dr. Svrcek's teaching and research interests center on process control and design. He is a registered professional engineer in Alberta and Ontario and is actively involved in applied research. Brent Young is a senior lecturer of chemical and materials engineering at the University of Auckland, New Zealand, and an adjunct associate professor at the University of Calgary, Alberta, Canada. He received his B.E. (1986) and Ph.D. (1993) degrees in chemical and process engineering from the University of Canterbury, New Zealand. Dr. Young's teaching and research interests center on process control and design. He is a registered professional engineer and is actively involved in applied research.

PAGE 55

W inter 2006 55cal process plants and their control systems. The students then create "disturbances" in the plant, which may involve changes in feed composition, flow, system temperatures, and/or pressures. The simulator demonstrates in real time what the effects of these "disturbances" would be on the plant operation, and it allows the student to evaluate the strengths and weaknesses of a given process control scheme. The course is accompanied by a textbook written by the course instructors, A Real-Time Approach to Process Control .[7] The text has 10 chapters, each of which focuses on a given aspect of process dynamics and control, whether it be investigating the concepts of process gain, time constants, and deadtimes, studying control schemes for distillation columns, or examining plant-wide control. Associated with the chapters are eight workshops[8] that are to be completed by the student using a dy namic simulator. Each individual workshop explores the concepts explained in the associated chapter, allowing students to assign meaning to the words. Due to the electronic nature of the workshops, hands-on, real-time experiments on laboratory unit operations equipment were considered a necessity to further reinforce the practical approach of the textbook. As a consequence, there is a compulsory laboratory component to the course.LABORATORY OVERVIEWThe laboratory component of the process dynamics and control course includes two traditional experiments: (1) a three-tank cascade where simple process identification and level control are the objectives, and (2) a double-pipe heat exchanger with a variable deadtime leg which can be configured to investigate feedback, cascade, and feedforward control. While these experiments offer students the chance to ex perience the effects of process/measurement noise and unmeasured disturbances, the behavior of the experiments is essentially linear, and the control loop studied is SISO in structure.SALT-MIXING LAB EXPERIMENTThe salt-mixing lab experiment that incorporates nonlinearity and MIMO behavior was designed in 2002 for immediate introduction into the curricula. Figure 1 is a schematic of the laboratory process experiment. The following is a description of what occurs in the process: A concentrated salt solution is mixed and stored in a large holding tank that was sized to give a five-hour or more run time. This solution is pumped into the conical mixing tank, passing through a magnetic flow meter and flow-control valve, which are used to regulate flow via a flow-control loop. Fresh water is supplied via building utilities; the water passes through a magnetic flow meter and control valve that are used in a flow-control loop to regulate the fresh-water flowrate. Upon entering the mixing tank the freshand saltwater streams are blended using a stirrer. The conical section of the mixing tank provides a strong process Figure 1. A schematic of the Salt-Mixing Laboratory Process.

PAGE 56

56 Chemical Engineering Education Figure 2. A photograph of the Salt-Mixing Laboratory. Figure 3. A screen-shot from the DeltaV DCS. nonlinearity. The level in the mixing tank is measured using a differential pressure cell. The blended solution enters a pump, is pressurized, and then moves to a pipe segment that allows for one of three flow paths of larger tube diameter to be selected. This setup allows one of three deadtimes to be examined. The stream will then pass through a conductivity cell/transmitter, which is used as the input to the master conductivity control loop. This loop's output is a cascaded setpoint to the slave fresh-water flow controller. Before going to drain, the stream passes through a control valve that is manipulated in order to regulate the level in the mixing tank. The flowrate, level, and conductivity inputs are all fed to the DCS system, as are the fresh-water, saltwater and level-control-valve-manipulated variables for this MIMO system. The input and manipulated variables are used within the DCS system with predefined function blocks to create the appropriate control loops.Figure 2 shows the salt-mixing laboratory skid. The instrumentation, tank pumps, and additional parts were purchased from suppliers but the construction of the skid and commissioning of the equipment was completed in-house with the help of university support staff. This resulted in a compact unit that has capacity for expansion and is completely portable, allowing for more efficient use of laboratory space. Figure 3 is a screen shot from the Emerson DeltaV distributed-control system (DCS) that is used for process data acquisition, monitoring, and control in the laboratory. The advantage of using a DCS is that they are common to modern industrial installations; as such, undergraduate engineering students should be taught what a DCS looks like as well as be provided with experience in controlling processes using such graphical interfaces. Other laboratories in the literature[1, 9-11] have also realized this necessity and addressed it in different ways. Rivera, et al.,[1] also employed an industrial DCS (Honeywell, in that case), as did Skliar, et al., at the University of Utah[9] in a graduate course also open to seniors (Opto 22, in the latter work). The approach of Bequette, et al., at Rensselaer Polytechnic Institute[10] was perhaps the more typical use of Matlab/Simulink block diagrams as an interface to simulated experiments. Braatz, et al. ,[11] employed the Hewlett Packard V isual Engineering Environment (HPVEE) to construct their student-operator interfaces to have a similar look and feel to an industrial DCS.

PAGE 57

W inter 2006 57 Figure 4. A timedomain plot from the DCS demonstrating the system response to saltwater flowrate changes from 0 to 0.5 then to 1.0 L/min (plus a few more).Overall Mass Balance Equation (Assuming constant density and isothermal) Salt Species Balance Equation (Assuming constant density and isothermal)T ABLE 1Overall Mass and Species Balance Equations T ABLE 2System NonlinearitiesNonlinearity Nonlinear Characteristic Linearized Characteristic V olume change with level in the conical section Product flowrate change with the level due to the valve Multiplicative nonlinearity between the volume and the salt concentration VhVhhop== 1 32322tan.tan.. qq FKhF K h hproductvproduct v op== 2 The overall mass and species balance equations that describe the dynamics of the system are included in Table 1, and the system nonlinearities are delineated and linearized in T able 2 so that the nonlinearities are clearer to non-control experts who have been assigned to teach process control. Figure 4 gives a time-domain plot from the DCS showing system response to saltwater flowrate changes from 0 to 0.5 then to 1.0 L/min (plus a few more). The effective tank-time constant varies with the flow.LABORATORY TASKSMyriad tasks can be done with the aforementioned apparatus. The purpose of this laboratory portion of the course is to allow students the opportunity to evaluate a variety of control schemes. To initiate this with the mixing-tank experiment, students set a tank level and then perform three step tests, where each step test is either an increase or a decrease from a nominal value. Tuning parameters (PI) are then calculated from the resulting process-reaction curves, using the students' choice of method (Cohen-Coon, Ziegler-Nichols, or IMC open-loop rules). The calculated tuning parameters are then compared with the tuning parameters obtained using the DeltaV automated tuning program (DeltaV tune), and both sets are tested by making setpoint changes or disturbances in the saltwater flowrate. The "best" set of tuning parameters is then chosen based on visual observations of the system response, including time to steady state, for each set of tuning parameters. With the best FFF dV dtFreshWaterSaltwaterP+-=roduct FxFy dVy dtSaltwaterP.. -()roduct dVy dt V dy dt y dV dt()+

PAGE 58

58 Chemical Engineering EducationThe simulator demonstrates in real time what the effects of . "disturbances" would be on the plant operation, and it allows the student to evaluate the strengths and weaknesses of a given process control scheme. tuning parameters entered into the system, the level in the mixing tank is then changed significantly, for example from 65% to 35%, which would mean moving from the cylindrical (linear) to the conical (nonline ar) section of the mixing tank or vice versa. Setpoint change(s) are then made in order to allow students to examine the process response. The students are then asked to perform a full analysis of the process behavior in both open and closed loop, including comments on linearity, order of response, and possible better control strategies for the apparatus. As well, the students are given an additional open-ended problem: to calculate the amount of salt initially added to the storage tank. The information given to the students to complete these tasks includes printouts of process data ( e.g. flowrates, conductivity) and the initial height of water in the storage tank. Students are also able to measure the tank dimensions if they so desire.EVALUATIONAlong with an analysis of the process behavior, the students were asked to provide some general comments on the laboratory. Overall, the laboratory was found to provide good exposure to the latest process equipment, along with demonstrating different tuning methods (including those done using the built-in autotuner). Students were able to recognize the nonlinearity in the system and provide an explanation, as well as provide explanations for the changes in time constant and deadtime with different flowrates. System noise was well demonstrated in this laboratory and its effect on the graphical method for calculating tuning parameters was noted. As well, the effect of capacity was seen. Many students also attempted the open-ended problemto calculate the initial mass of saltand used a number of approaches in attempts to solve it. General student comments and laboratory reports indicated that students enjoyed working with the new laboratory experiment, and that it was helpful to see a real process that could provide them with a feel for what types of disturbances can be made in a plant. (Whereas, in the simulation workshops, unrealistic disturbances are quite possible and it is sometimes difficult to measure the actual time effect a disturbance would have.) Because it was a real process, the students did find the experiment was a little long, as it usually ran slightly in excess of four hours (the time period scheduled for the experiment). A smaller process could be considered, but long time constants are a reality of industrial plants and this is an important fact for students to realize that is often somewhat overlooked in their process control education. In general, it was felt that the laboratory was well received by students, and that it provided them with good exposure to state-of-the-art control hardware. The students were also exposed to instrumentation they had not seen before, such as magnetic flow meters and conductivity cells. The experiment also effectively displayed the difference between a simulation and a real process, in that it took up to 30 minutes to achieve steady state in closed loop, depending on the tuning parameters and the setpoint change made. Some ways in which this "down" time could be used more effectively include: Quizzes Lab discussions T utorial support Additional reading material Increased time to explain the apparatusThese options could be used to keep the students focused on the experiment since it is felt that what was actually going on in the process was often overlooked due to other distractions during the time lags. Despite this, students did seem to take note of some pitfalls that can be encountered when tuning controllers, such as the errors associated with the graphical methods and the importance of proper input design. The experiment also reaffirmed the value of a DCS in the teaching environment. Unit operations laboratories had previously had DCS systems integrated into them, but the DCS was not used in a control context and students did not need to make use of all of the data-collection and

PAGE 59

W inter 2006 59 General student comments and laboratory reports indicated that students enjoyed working with the new laboratory experiment, and that it was helpful to see a real process that could provide them with a feel for what types of disturbances can be made in a plant.handling capabilities of the system. This experiment also showed a practical application of cascade control as the fresh-water supply pressure was not regulatedtherefore changes in the water system would propagate through the system but would be quickly compensated for by the slave fresh-water flow-control loop that is manipulated by the master-conductivity control loop. It was felt that the bonus question worked well and that it should be made mandatory for future labs. It was also convenient for the teaching assistants that the lab could be run differently for each group by simply changing the initial salt concentration or flowrates. As well, this changeability provided the teaching assistants with an opportunity to learn more about process control. Overall, it was thought the lab performed very well and showed much promise as well as many other areas of potential use. For instance, it would be useful in a more advanced process control course where it could be used to demonstrate system identification and model predictive control in a practical setting.CONCLUSIONSThe introduction of this new lab was successful from the students' point of view. They enjoyed working with the latest process control instrumentation. They also gained a new appreciation of the problems associated with real plants, in the form of noise and unexpected disturbances. The comparison of conventional open-loop tuning methods and an automated tuning package was appreciated, as was the chance to show their creativity in the solution of the open-ended bonus question. From the instructors' point of view, the laboratory was considered successful. The only real concerns with the lab were based on the length of time it took to complete. This will be addressed in coming years with the introduction of quizzes and discussion while waiting for the process to reach steady state. Despite these concerns the lab provided an effective demonstration of a nonlinear and MIMO system. Most importantly, it was felt the students were better able to understand process behavior by being able to see many of the classroom concepts on an actual process. The department also gained a valuable tool for additional process control courses due to this lab's ability to have the control configuration changed, the ease in which it can be upgraded or modified, and its extensive data-collection and data-handling capabilities.ACKNOWLEDGMENTSFinancial support from the Calgary Engineering Endowment fund is gratefully acknowledged for the purchase of the laboratory hardware. Bernie Then is thanked for the assembly of and assistance in commissioning the laboratory hardware.REFERENCES1.Rivera, D.E., K.S. Jun, V.E. Sater, and M.K. Shetty, "Teaching Process Dynamics and Control Using an Industrial-Scale, Real-Time Computing Environment," Comp. Apps. in Eng. Ed. 4 (3), 191 (1996) 2.Badmus, O.O., D.G. Fisher, and S.L. Shah, "Real-time, Sensor-based Computing in the Laboratory," Chem. Eng. Ed. 30 (4), 280 (1996) 3.Braatz, R.D., and M.R. Johnson, "Process Control Laboratory Education Using a Graphical Operator Interface," Comp. Apps. in Eng. Ed. 151 (1998) 4.Rusli, E., S. Ang, and R.D. Braatz, "A Quadruple-Tank Process Control Experiment," Chem. Eng. Ed. 38 (3), 171 (2004) 5.Svrcek, W.Y., D.P. Mahoney, and B.R. Young, "A Real-Time Approach to Process Control EducationA Paradigm Shift," ASEE99 Conference Charlotte, NC, June (1999) 6.Aspen Dynamics and HYSYS, Products of AspenTech Inc., and subsidiaries, Boston (2002) 7.Svrcek, W.Y., D.P. Mahoney, and B.R. Young, A Real-time Approach to Process Control John Wiley and Sons Ltd., Chichester, UK (2000) 8.Young, B.R., D.P. Mahoney, and W.Y. Svrcek, "Real-Time Simulation Workshops for Undergraduate Process Control Education," Proceedings, ACE2000, 5th IFAC/IEEE Symposium on Advances in Control Education Nara, Gold Coast QLD, Australia, December (2000) 9.Skliar, M., J.W. Price, and C.A. Tyler, "Experimental Projects in Teaching Process Control," Chem. Eng. Ed. 32 (4), 254 (1998) 10.Bequette, B.W., K.D. Schott, V. Prasad, V. Natarajan, and R.R. Rao, "Case Study Projects in an Undergraduate Process Control Course," Chem. Eng. Ed. 32 (3), 214 (1998) 11 Ang, S., and R.D. Braatz, "Experimental Projects for the Process Control Laboratory," Chem. Eng. Ed. 36 (3) (2002)

PAGE 60

60 Chemical Engineering EducationShortly after starting as an assistant professor, I realized that quite a few of our students were unable to analyze laboratory data at a level consistent with that expected when I had worked in industry. Having been put in charge of the Florida Institute of Technology's introductory chemical engineering course and its materials science and engineering laboratory course, I decided that a strong emphasis on data analysis would be added to each of these courses in order to satisfy ABET's requirement regarding the ability of students to analyze data. Most departments emphasize spreadsheet calculations and plotting of data in Microsoft Excel as part of their introductory chemical engineering course. Experience in our department has shown that unless sufficient time is spent on data analysis instruction such that spreadsheet calculations, plotting, and curve fitting become second nature, such skills are either forgotten or are never learned properly. We have incorporated DataFit from Oakdale Engineering[1]throughout the entire curriculum at Florida Tech, beginning with CHE 1102, an eight-week, one-day-per-week, two-hour, one-credit-hour, second-semester Introduction to Chemical Engineering course in a hands-on computer classroom. The syllabus for CHE 1102 is shown in Table 1. The examplesDATA ANALYSIS MADE EASY WITH DATAFITJAMES R. BRENNERFlorida Institute of Technology Melbourne, FL 32901James R. Brenner received his B.S. degree from The University of Delaware and M.S. and Ph.D. degrees from The University of Michigan. After a postdoc at Argonne National Laboratory and industrial experience at W estinghouse Savannah River Company, he became an assistant professor of chemical engineering at Florida Institute of Technology. His research interests are in hydrogen purification and sensing, electronic noses, and nanoporous materials. ChEclass and home problems Copyright ChE Division of ASEE 2006 The object of this column is to enhance our readers' collections of interesting and novel problems in chemical engineering. Problems of the type that can be used to motivate the student by presenting a particular principle in class, or in a new light, or that can be assigned as a novel home problem, are requested, as well as those that are more traditional in nature and that elucidate difficult concepts. Manuscripts should not exceed 14 double-spaced pages and should be accompanied by the originals of any figures or photographs. Please submit them to Professor James O. Wilkes (e-mail: wilkes@umich.edu), Chemical Engineering Department, University of Michigan, Ann Arbor, MI 48109-2136.

PAGE 61

W inter 2006 61ticle. Students in CHE 1102 cover basic statistics during the first week of the course and get constant reinforcement of these concepts through the use of DataFit.[1] The second half of CHE 1 102 consists of problems that require Polymathor Excelbased solutions to either sets of l inear and nonlinear algebraic equations or numeric integration, as suggested by Clough.[2]All Excel and DataFit files are available at .SOLVING PROBLEMS WITH DATAFITProblem 1. Calibration of a Pressure Transducer Following the introduction to basic statistics, the first problem that I assign students is the calibration of a 0-250 psig Span Instruments' NTT-204 (now Millipore) pressure transducer against a 0-1000 psia Paroscientific pressure transmitter. In addition to being useful for teaching students how to make plots with error bars and determine the difference between absolute and gauge pressures, it provides a relatively simple problem for studying linear regression with DataFit. The repeatability and lack of drift of Paroscientific pressure transmitters is even su perior to that of a deadweight tester that was calibrated at NIST.[3] The repeatability of the quartz oscillator that the Paroscientific pressure transmitters use is certainly within the quoted 0.01% of full-scale precision ( i.e. 0.1 psia fixed error for a 1000psia transmitter). Span Instruments' pressure transducers output a signal that ranges from 4-20 milliamps to within 0.08 milliamps. After having the students prepare a plot of the data shown in Figure 1, including error bars, I ask the stu-Figure 1 Calibration of a Span Instruments' pressure transducer against an NISTtraceable Paroscientific pressure transmitter. 0 50 100 150 200 250 300 050 100150200250300 Transmitter pressure (psia)Transducer pressure (psig) y = (.998 + .005)*x + (-15.2 + .8) chosen, shown in parentheses, are selected so as to be consistent with concepts that students learn concurrently in other courses. DataFit also has become commonly used in our Physical Chemistry Lab and Materials Science and Engineering Lab courses, as well as in several courses in other engineering departments. Our experience at Florida Tech is that students retain data analysis concepts best when such concepts are formally taught to them in this short course and then periodically reinforced throughout their academic careers. Several examples covered in weeks three through eight will be discussed here. An introduction to basic statistics is included in nearly all introductory ChE courses and will not be discussed in this ar-T ABLE 1 Data Analysis Curriculum1)Statistics and Confidence Intervals 2)Introduction to Plotting and Calculations in Excel 3)y = ax + b Fitting in DataFit (Pressure Transducer Calibration) 4)y = ax Requires User-Defined Models (Hygrometer Calibration) 5)Semi-Log Functions (First-Order Rate Laws Felder and Rousseau 2.34) 6)Plotting and Curve Fitting of Power-Law Functions (Crystal Growth Felder and Rousseau 2.37) 7)Nonlinear Functions (Vapor Pressures) 8)Curve-Fitting in 3-D (Rate Laws W ith Two Reactants) Experience in our department has shown that unless sufficient time is spent on data analysis instruction such that spreadsheet calculations, plotting, and curve fitting become second nature, such skills are either forgotten or are never learned properly.

PAGE 62

62 Chemical Engineering Educationdents to copy and paste the data into DataFit, click on the Solve Regression option, click on OK, select the y = ax + b option, and let DataFit do the work for them. By clicking on Results Detailed, the Fit Information output is obtained (Table 2). Included in the output are the residual sum of squares (RSS), which is the sum of the squares of the differences between the calculated values of Y, the pressure in psig as determined by the pressure transducer, and the corresponding experimental values. A lso evaluated are the commonly seen R2 correlation parameter as well as several moreadvanced goodness-of-fit parameters. Most importantly, the 68%, 90%, 95%, and 99% confidence intervals are conveniently tabulated. This is an excellent opportunity to reinforce basic statistics, most notably the Gaussian distribution, which is typically taught at the beginning of CHE 1102.Problem 2. Calibration of a HygrometerThe second problem that I assign is Problem 2.32 from Felder and Rousseau's textbook.[4] This problem involves the correlation of a signal from a hygrometer versus the mass fraction of water in the inlet stream to the hygrometer. For this problem, first ask students to do the y = ax + b fit as described in the previous section. The 95% confidence intervals on the slope, a, and the intercept, b, are as follows: a = 470 20; b = 0 2, at the appropriate number of significant figures (proper use of significant figures is an extremely difficult concept to get students to consistently apply). Then ask them whether the intercept, b, is mathematically significant ( i.e. nonzero within the 95% confidence interval). They should answer that b is not mathematically significant at the 95% confidence level. Out of a sample of 100 students asked over the last five years as part of an in-class exercise, only 50% have answered correctly to this question; 25% of students replied "don't know." This is a surprisingly difficult concept to master that requires consistent reinforcement throughout CHE 1102. Yet, of the same sample of students, 98% replied correctly to a similar question during hourly and final exams. Once the students have realized that b is unnecessary, it is time to teach them how to create a user-defined model in DataFit, as y = ax is not one of the built-in models (one of DataFit's few shortcomings). This can be done by returning to DataFit's main menu and clicking on the Define User Model option under the Solve menu. The user defines a Model ID, (which I defined as "Linear, no intercept," in this case). The user also inputs the Model Definition, in this case Y = a*x. Mathematical functions in DataFit, such as multiplication and exponentiation, work in the same way as Excel. In many cases, including this one and all cases where the fitting is of a linear function, initial estimates are unneces-Sum of Residuals = 4.08562073062058E-14 A verage Residual = 3.14278517740044E-15 Residual Sum of Squares (Absolute) = 5.20799168906741 Residual Sum of Squares (Relative) = 5.20799168906741 Standard Error of the Estimate = 0.688079784556427 Coefficient of Multiple Determination (R^2) = 0.99994096 Proportion of Variance Explained = 99.994096% Adjusted coefficient of multiple determination (Ra^2) = 0.9999355927 Durbin-Watson statistic = 2.88469613789683 T ABLE 2Fit Information for Pressure Transducer CalibrationDataFit version 6.1.10 Results from project "F:\brenner\datafit\pcalib.dft" Equation ID: a*x+b Number of observations = 13 Number of missing observations = 0 Solver type: Nonlinear Nonlinear iteration limit = 2000 Diverging nonlinear iteration limit =10 Number of nonlinear iterations performed = 1 Residual tolerance = 0.0000000001 Regression Variable Results VariableValueStandard Error t-ratioProb(t) a0.9980017790.002312177 431.6287071 0 b-15.17627790.359917526-42.1659876 0 68% Confidence Intervals VariableValue68% (+/-)Lower LimitUpper Limit a0.9980017790.0024081320.9955936481.000409911 b -15.17627790.374854103-15.551132-14.8014238 90% Confidence Intervals VariableValue90% (+/-)Lower LimitUpper Limit a0.9980017790.004152438 0.9938493421.002154217 b-15.17627790.646375884-15.8226538-14.529902 95% Confidence Intervals VariableValue95% (+/-)Lower LimitUpper Limit a0.9980017790.005089101 0.9929126791.00309088 b-15.17627790.792178474-15.9684564-14.3840995 99% Confidence Intervals VariableValue99% (+/-)Lower LimitUpper Limit a0.998001779 0.007181158 0.9908206221.005182937 b-15.17627791.117831851-16.2941098-14.0584461 V ariance Analysis SourceDFSum of SquareMean Square F Ratio Prob(F) Regression188206.0227888206.02278 186303.3408 0 Error115.2079916890.47345379 Total1288211.23077

PAGE 63

W inter 2006 63 Figure 2. ClausiusClapeyron plot for water vapor pressures.[4]sary, but they become critical when doing some nonlinear fitting. The default values of each of the curve-fit parameters are unity in all cases. I look at this as one of DataFit's very few design flaws. When one goes through a Taylor series expansion, terms involving higher-order parameters are supposed to be corrections to the previous terms, meaning that the product of the curve-fit coefficient multiplying a highorder term and that higher-order term ( i.e. d*x3) should be less than those of previous terms. Without some exceptional physical justification, it would be difficult to throw out constant, linear, or quadratic terms and keep a cubic term. After manually assigning initial estimates and/or constraints on the curve-fit coefficients, clicking OK, clicking Solve Regression, and OK again, the user will need to locate his or her user-defined model in the list of models. After locating your recently defined model, click on Solve, click OK, and then click on Results Detailed to return to the Fit Information screen once again. The models are ranked by the RSS, and so the Fit Information that pops up first is the one with the lowest RSS, not the one for the most recent fit. By clicking on the uppermost dialog box to locate the user-defined model, one will get the Fit Information associated with the user-defined model, "Linear, no intercept." Interestingly, scrolling down to the 95% confidence interval shows that the confidence interval for the one-parameter model (a = 473 8) is narrower than the slope from the two-parameter model (a = 470 20).Problem 3. Fitting Water Vapor Pressures to the Clausius-Clapeyron and Antoine EquationsFitting water vapor-pressure data to the Clausius-Clapeyron equation is challenging for underclassmen, but usually can be done successfully if the previous examples have been worked out in class or for homework. This problem, along with the follow-up fitting of the same data to the Antoine equation, typically is either the final in-class or homework problem that students are asked to solve during CHE 1102. Data for the vapor pressure of water is tabulated in Appendix B.3 of Felder and Rousseau.[4] The Clausius-Clapeyron equation is as follows, and requires conversion of temperatures into Kelvin: log()101 PA B T =At this point in the course, the students know that they should plot pressure on a logarithmic scale on the y-axis and reciprocal temperature on the x-axis. Students are asked to plot 1000/T so that the values on the x-axis are between a more aesthetically pleasing 0 and 10, to estimate the slope (B) and the intercept (A) graphically, to use DataFit to determine A and B, and finally to superimpose the curve fit (the solid line) on top of the experimental points (Figure 2). The Clausius-Clapeyron equation is a reasonably good fit 1 10 100 1000 2.62.72.82.93.03.13.23.33.43.53.63.73.8 1000/Temperature (K-1) Vapor Pressure of Water (mm Hg)

PAGE 64

64 Chemical Engineering EducationT ABLE 3 Clausius-Clapeyron Constants for Vapor Pressure of Water from 0 to 60 oCClausius-ClapeyronLinear Fit ofNonlinear Fit of ConstantsLinearized Dataof Raw Data A 9.091 + 0.0049.003 + 0.004 B 2301 + 1 2274 + 1 T ABLE 4 Antoine Curve Fitting of Vapor Pressure of Water from 0 to 60 oCConstants250 iterationsProper Convergence Literature Data[4, 5] A 6.95 + 0.088.124 + 0.002 8.10765 B 1180 + 401759.8 + 0.6 1750.286 C 186 + 4235.8 + 0.1 235.000 T ABLE 5Clausius-Clapeyron Equation Parameters*Molecule AL BL AN BNCarbon Dioxide7.58 + 0.02 865 + 4 7.58 + 0.01 864 + 3 Ethane7.37 + 0.05 837 + 9 7.127 + 0.008 785 + 2 Propane7.71 + 0.081130 + 14 7.191 + 0.0071128 + 3 Isobutane7.69 + 0.071274 + 16 7.198 + 0.007 996 + 2 Butane7.61 + 0.061306 + 7 7.256 + 0.0091193 + 4 *Pressures in mm Hg and temperatures in KelvinLLogarithm of pressure taken firstNLogarithm of pressure not taken first of the vapor pressure of water data from 0 to 60 C, but one can see that there is a systematic deviation from linearity at low temperature and pressure. By graphically extrapolating a straight line through the portion of the data that appears to be linear, one can estimate the slope (-B) as -2200 and the intercept (A) as 109 from Figure 2. Interestingly, there are slight differences in the DataFit estimates of the curve fit parameters, depending on whether the logarithm of the pressure data and the inversion of the temperature data are taken before curve fitting in DataFit or not (Table 3). In the case where the data are not so linearized before entry into DataFit and then a nonlinear model is generated in DataFit, the points at low vapor pressures are de-emphasized relative to the other points. If one tries to fit the Antoine equation for water vapor pressures either below 60 C or above 60 C, in either case if one does not manually change the default parameter guesses of unity, DataFit's "solution" will require more iterations than the default number of iterations, which is 250. log()102 PA B TC =+() This problem can be changed using Edit Preferences. I have changed the default number of iterations permanently to 2,000. The problem with using the results for A and B from the Clausius-Clapeyron equation as initial guesses for A and B for the Antoine equation fit is that the Antoine equation requires temperatures to be in degrees Celsius instead of in Kelvin. In fact, if one uses the Clausius-Clapeyron equation constants to fit the water-vapor pressures above 60 C and lets DataFit set the default value of C to 1, then even after having made the appropriate conversion of the data from Kelvin into Celsius, DataFit will erroneously return a "successful" result after only one iteration that contains errors larger than the values of the parameters themselves. The Antoine equation cannot be solved for temperature ranges in which the denominator, (T+C), switches from negative to positive over the range of temperatures. If one uses the values of A and B from the ClausiusClapeyron equation and an initial guess for C of 273.15, then the Antoine equation does converge properly to the answers below in Table 4 in the "Proper Convergence" column. This discrepancy proved a difficult challenge for even the best students. At a minimum, the number of significant figures reported for Antoine equation constants in the literature[4, 5]is grossly overstated, and, for some molecules, is just not quite right (see Table 5).ASSESSMENTIn the first class exposed to this curriculum, 17 of 20 students successfully completed both the Clausius-Clapeyron and Antoine problems. Two of the three students who failed to make a proper plot and a proper fit in DataFit attended class less than one-third of the time, and the other student, although in good attendance, turned in less than half of the homework assignments and had significant language problems. The past four years of classes have had similar results. A similar problem, for butane vapor pressures, has been assigned to sophomores and graduate students, using data from the NIST Chemistry WebBook.[12]

PAGE 65

W inter 2006 65Figure 3. Antoine fit of butane vapor pressure data clearly shows bias against low vapor pressure points.All but one of 12 sampled students who came to Florida Tech from other countries for ChE graduate school sought me out for help. None of the eight students that went to Florida Tech for both bachelor's and master's degrees needed help. Ninety percent of sophomore students who took CHE 1102 as freshmen were also able to solve the butane problem successfully. W ith the default guesses, DataFit failed to converge because it cannot handle the denominator changing from negative to positive, depending on temperature. When the second term exceeds A, the solution also diverges. Under some sets of initial estimates, DataFit "converges" to a flat line! When the initial estimates are reasonably close to what DataFit reports as the correct answer (A = 7.44 0.04; B = 1330 30; C = 294 4), the solution converges to what is shown in Figure 3. Even this is clearly incorrect as the low vapor pressure data is de-emphasized, because the magnitude of the error in such a small quantity is dwarfed by a small percentage error in the high vapor pressure points This kind of error is not unique to DataFit. I have seen it in Polymath curve fits as well.CONCLUSIONSOf the international graduate students asked to fit vaporpressure data for the previous problem, none had previous exposure to either Polymath or DataFit. While each of them also learned how to use Polymath in graduate school, 11 of the 12 polled said that they found DataFit easier to use. The reason that I downloaded DataFit in the first place was not because of its excellent curve-fitting capabilities, but because when I first started using it in industry in 1998, DataFit was the only program that did proper 3-D scientific plotting for less than $500. In 1999, when Florida Tech bought a site license for DataFit version 6.1, it cost only $750 for the entire campus (albeit a relatively small campus), whereas a single copy cost $100. Moreover, the site license allowed for students and faculty to use DataFit at home as long as they were doing academic work. A comprehensive set of solutions to similar problems can be found at .REFERENCES1.Gilmore, J., DataFit, v 6.1 Oakdale Engineering, 23 Tomey Road, Oakdale, PA 15071, (724) 693-0320, sales@curvefitting.com, 2.Clough, D., "Spreadsheets Across the Curriculum," ASEE Summer School for ChE Faculty, July (2001) 3.Brenner, J.R., and E.F. Dyer, Westinghouse Savannah River Company, unpublished results, December (1997) 4.Felder, R.M., and R.W. Rousseau, Elementary Principles of Chemical Processes John Wiley and Sons, 3rd Ed., New York (2000) 5.Dean, J.A., Lange's Handbook of Chemistry McGraw-Hill Companies, Inc., 14th Ed., New York (1992) 6.National Institute of Standards, NIST Chemistry WebBook,

PAGE 66

66 Chemical Engineering EducationSince 1999, Chemical Engineering undergraduate students have had the opportunity to participate in the Chem-E-Car Competition at the regional and national level under the direction of the American Institute of Chemical Engineers (AIChE). The competition was initiated by AIChE members to (1) provide an opportunity for students to participate in a team competition at the national level, (2) encourage professional society interaction, and (3) increase the awareness of chemical engineering in the public.[1] Examples of national competitions in other engineering disciplines include the concrete canoe race (civil engineering), minibaja race (mechanical engineering), and International AIAA/ ONR Design, Build, Fly contest (aerospace engineering). The Chem-E-Car competition involves the design and construction of a chemically powered car that has to travel a specified distance (50-100 ft) while carrying a certain amount of water (0-500 ml). The car must fit into a box no larger than 40 cm 30 cm 18 cm and the team must be composed of members from at least two undergraduate classes. Additional rules are applicable to the competition.[1] The objectives of the competition are applicable to numerous ABET educational outcomes including "an ability to design a system, component, or process to meet desired needs," "an ability to function on multidisciplinary teams," and "an ability to communicate effectively."[2]ENGINEERING ANALYSIS IN THE CHEM-E-CAR COMPETITIONRANDY S. LEWIS, ALIAKBAR MOSHFEGHIAN, AND SUNDARARAJAN V. MADIHALLYOklahoma State University Stillwater, OK 74078 Copyright ChE Division of ASEE 2006 Randy S. Lewis is a professor of chemical engineering at Brigham Young University and an adjunct professor of chemical engineering at Oklahoma State University. He received his B.S. and Ph.D. degrees in chemical engineering from Brigham Young University and Massachusetts Institute of Technology, respectively. He currently serves as past-chair of the Career and Education Operating Council of AIChE. His research interests include biomaterials development and the utilization of renewable resources for the production of chemicals. Aliakbar Moshfeghian graduated with a B.S. in chemical engineering from Oklahoma State University in 2003. He placed second in the 2003 AIChE National Student Paper Competition after winning the AIChE Mid-America Student Paper Competition. Currently he is completing a master's degree in chemical engineering at Oklahoma State University. Sundararajan V. Madihally is an assistant professor of chemical engineering at Oklahoma State University. He received his B.E. from Bangalore University and his Ph.D. from Wayne State University, both in chemical engineering. He held a research fellow position at Massachusetts General Hospital/Harvard Medical School/Shriners Hospital for Children. His research interests include tissue regeneration and the development of therapies for traumatic conditions. ChEclassroom

PAGE 67

W inter 2006 67To promote the competition among Oklahoma State University (OSU) chemical engineering students and to provide an additional design experience in the undergraduate curriculum, the competition was implemented in 2000-2001 as part of a spring sophomore course (Introduction to Chemical Process Engineering) and a spring junior course (Chemical Reaction Engineering). The juniors initially worked on designing the cars and were eventually joined by the sophomores who primarily helped with the calibration, poster, and safety aspects. The teams (six-eight students) currently compete in the middle of the spring semester for the opportunity to represent OSU at the AIChE Regional Chem-E-Car Competition. The evolution of the competition at OSU was recently presented.[3]In brief, funding for the OSU competition was initially provided by the department, but ChevronPhillips now provides funding for equipment costs, T-shirts, the awards banquet, and travel to regional and national competitions. Further, ChevronPhillips personnel provide extensive safety reviews on students' reports. Liquid effluent was allowed to discharge from the car in 2001, only water was allowed in 2002, and no liquid discharge has been allowed since 2003. In 2003, an additional fall junior course (Thermodynamics) was included in the competition to enable the students to spend more time working on their cars. As part of the integrated sophomore and junior team, the students are required to write a safety and environmental report, provide a detailed sketch of the car, build a prototype, provide preliminary and final calibrations, provide an engineering analysis, give a poster presentation, and participate in the department competition. The engineering analysis is performed solely by the junior students, although they have traditionally provided a vague analysis such as using empirical equations, providing detailed equations without any solutions, and identifying fundamental equations that may not be applicable. Engineering analysis is not required at the national competition and is often not applied. Rather, students rely on calibration data and trial and error to predict the distance traveled by their cars. To demonstrate and encourage the use of detailed engineering analysis among the students in predicting the distance traveled by a car, a model was developed for a car (previously used in the competition) in which pressure generated by a chemical reaction resulted in car movement via the discharge of water. This work presents the model for predicting the travel distance based on the initial pressure and various car parameters. Although discharged water must now be contained such that the model may not be applicable to current car designs, this work provides an example of how students can effectively apply engineering analysis. An advantage to engineering analysis is that it allows students an opportunity to determine the effects of design components ( e.g., vessel size, car weight, liquid volume, nozzle size, for this example) on the distance the car travels.MATERIALS AND METHODS Car Design and Experimental Runs The car, shown in Figure 1, was designed and built by Ali Moshfeghian, Christ Schulte, and Kyle Sharon (junior chemical engineering students at the time) and was used in the 2002 competition at OSU. The key car param-Figure 1. Picture (A) and diagram (B) of the Chem-E-Car. The left chamber was used to generate gas from a sodium bicarbonate (NaHCO3) and vinegar reaction. The right chamber contained water that was forced from the chamber following the removal of the clamp. The expelled water propelled the car forward. The parameters and values are shown in Table 1. The Chem-E-Car competition involves the design and construction of a chemically powered car that has to travel a specified distance while carrying a certain amount of water.

PAGE 68

68 Chemical Engineering Educationeters are provided in Table 1. The left chamber, shown in Figure 1B, was initially filled with 125 ml of a saturated aqueous solution of sodium bicarbonate (NaHCO3). Glacial acetic acid (vinegar) was then added to the solution, causing a chemical reaction to form CO2 that increased the chamber pressure. The acetic acid was added according to the amounts shown in Table 1 and, when necessary, additional water was added so that the acetic acid/water addition equaled 10 ml. Although acetic acid and sodium bicarbonate were used to generate the gas pressure, any pressure-generating chemical Run #Acetic Acid Initial gas Adjusted initial Distance (ml)pressure (atm) pressure (atm) (feet) 1 2.5 4.39 2.78 4.3 2 5.0 8.14 4.76 19.3 3 7.5 11.20 6.37 32.6 4 10.0 13.24 7.43 41.7 Parameter Ac A0 C g hi h0 m mcar ngas pgas p0 R T vcar Vgas Vliq Vtot v0 xcar rliq mk Description Area of water chamber Nozzle area Head loss coefficient Gravitational constant W ater height above nozzle Nozzle height Mass of water Mass of car Moles of gas Adjusted initial pressure Atmospheric pressure Gas constant T emperature Car velocity Gas volume W ater volume Gas and initial water volume W ater velocity Car distance W ater density Friction coefficient Value 11.4 0.087 0 to 0.2 9.81 374 2470 See above 1.0 82.06 298 0 390 375 765 Eq. (8) 0 0.997 0.07 Units cm2 cm2 unitless cm/s2 g g atm atm cm3atm/molK K cm/s cm3 cm3 cm3 cm/s feet g/cm3 unitless Note Constant Constant Constant Initial value Initial value Initial value Constant Constant Constant Initial value Initial value Initial value Constant Variable Initial value Constant Figure 4 T ABLE 1Parameters and Values Used in Engineering Analysis

PAGE 69

W inter 2006 69 viP0,v0 Vliqhih0 ngasPgasVgasA0 Figure 2. Diagram of model used for the engineering analysis. The model represents the chamber that contains the water. The parameters and values are shown in Table 1. reaction would be sufficient for operation of the car. It is important to note that the generated pressure should not exceed the pressure limits of the materials to prevent material failure. A clamp was used to keep the gas pressure in the left chamber until the pressure reached equilibrium. The right chamber was filled with 400 ml of water (Vliq). Following pressure equilibration, the clamp was removed and the rear nozzle opened. The pressure above the water (Pgas), related to the moles of gas (ngas), forced the water to exit the rear nozzle (cross sectional area of A0) at a given velocity (v0). P0 represents the atmospheric pressure. The separator was added to minimize foam, generated in the left chamber, from entering the water chamber. The exiting velocity produced a thrust that moved the car forward. When the water ran out, the car rolled to a stop and the distance the car traveled was measured. As noted in Table 1, four experiments were performed. The experiments were performed on a smooth brick surface. Model Development A model was developed to predict the distance the car would travel based upon the initial pressure above the water. The model was used as a comparison with the experimentally measured distance. Figure 2 shows a representation of the water chamber that was used for the model. Vliq is the water volume, h0 is the height of the nozzle (assigned a value of zero), hi is the height of the water above the nozzle, vi is the surface velocity of the water at hi, A0 is the nozzle cross-sectional area, v0is the water velocity leaving the chamber, and P0 is the pressure of the surrounding atmosphere. Pgas, Vgas, and ngas represent the pressure, volume, and moles of the gas above the water, respectively. A material balance on the total mass of the car (mcar), which is equivalent to a constant mass plus the mass of the water in the chamber (m), shows that the mass changes with time according to dm dt dm dt vAcar liq==-r001 () where rliq is the liquid density. The right term represents the mass flowrate of water leaving the water chamber (and the car). Since m = Vliq rliq and if rliq is assumed constant, the water material balance shows how Vliq changes with time according to dV dt vA dV dtliqgas=-=-002 () The change in Vgas is also shown with time in Eq. (2) since any water volume decrease results in the same increase in the gas volume ( i.e., the total volume, Vtot, is constant and equal to Vgas+Vliq). To assess how the gas pressure (Pgas) changes with time, the ideal gas law was assumed where PgasVgas= Pgas(Vtot Vliq)=ngasRT Since Vtot is constant and ngasRT is constant as the water is leaving the nozzle (assuming negligible temperature change and no new gas is generated once the experiment starts), the time derivative of the ideal gas law gives dP dt P VV dV dt vA VV Pgasgas totliq liq totliq gas= -()= -()003 () Eq. (2) was substituted into the middle term of Equation 3 to obtain the term on the right. The velocity of the car with time was predicted from a momentum balance on the car. The momentum balance states that the change of momentum (mass of the car, mcar, times the velocity of the car, vcar) is equal to the sum of the forces acting upon the car: dmv dt m dv dt v dm dt vAmgcarcar car car car car liqkcar()=+=rm0 2 04 () The first term on the far right side of Eq. (4) represents the thrust force that pushes the car forward.[4] Only thrust occurring when water leaves the chamber was considered. Once the water runs out, residual gas pressure greater than atmospheric pressure will cause some thrust but the thrust is likely negligible since the gas density is small compared to liquid. The second term on the far right side represents the friction force between the car and the ground, with mk as the friction coefficient.[4] The negative sign signifies a force that decreases the car velocity. The drag force between air and the car was neglected. Substitution of Eq. (1) into Eq. (4) gives dv dt vAvv m gcar liqcar car k= +()r m0005 () Once the water runs out of the chamber, the first term on t he right side is zero and the velocity of the car will decrease as a result of friction until the car stops. The distance (xcar) at which the car stops was predicted from the definition of velocity, dx dt vcar car= () 6 To predict the velocity of water leaving the car (v0) for ap-

PAGE 70

70 Chemical Engineering Education 0 10 20 30 40 50 60 2.784.766.377.43 C=0 C=0.1 C=0.2 Initial Pgas(atm)Feet Figure 3. Measured (dashed lines) and predicted distance (bars with mk = 0.069) traveled by the car as a function of initial gas pressure (Pgas) above the water. The error bars show the predicted range with 0.066 £ mk £ 0.072.plication in Eqs. (1)-(5), the mechanical energy balance,[5]with the inclusion of frictional head loss due to the exit nozzle, was utilized such that C vvvghh PPi i i liq2 1 2 70 2 0 22 0 0+-()=-()+ -()r () The subscripts i and 0 refer to the values at the gas-liquid interface and the nozzle exit, respectively. C is the head loss constant. Since Pi= Pgas, (hih0) = Vliq/Ac (where Ac is the crosssectional area of the water chamber), and if vi << v0 (the liquid velocity leaving the chamber is much faster than the velocity of the water surface at the gas-liquid interface) then v C g V A PPliq c gas liq 0 02 1 8 = +()+ -() r () Eqs. (7) and (8) are only valid when water is present in the chamber. Thus, once the water completely runs out of the chamber Eqs. (7) and (8) no longer apply and v0 is zero in Eqs. (1)-(5). Eqs. (1)-(3), (5), and (6) [with the definition of Eq. (8)] were numerically integrated using Polymath[6] to obtain values of the integrated parameters as a function of time. When the model results showed that the water ran out (Vliq = 0), v0was set to zero for reasons stated above. At this point, only Eqs. (5) and (6) were numerically integrated. The initial values for solving the model were mcar=2470 g, Vliq=375 cm3, vcar=0, and xcar=0. For the water volume, the volume initially added to the chamber was 400 ml. Since 25 ml was below the nozzle and did not leave the chamber, the initial water volume was modeled with a value of 375 ml. The values of Pgas for the four experimental runs are shown in T able 1. Since the initial gas pressure (Pinit), as shown in Table 1, was measured prior to opening the clamp, the adjusted initial pressure was determined by Pgas=(205/390)*Pinit+(185/ 390)*1atm. The adjustment was based on the assumption that the pressure above the water chamber (with a volume of 185 ml) was 1 atm and that the initially measured pressure (with a volume of 205 ml) equilibrated (in the total volume of 390 ml) after the clamp was opened and prior to the opening of the rear nozzle. A value of C=0.1 is consistent with fluid leaving a large reservoir and entering a small rounded-edge entrance ( i.e. similar to liquid leaving the chamber and entering the nozzle).[5] T able 1 summarizes the model parameters with their associated values. Unit consistency was ensured when solving the equations.

PAGE 71

W inter 2006 71 y = 14.42x R 2 = 0.9899 0 5 10 15 20 25 30 35 0 0.51.01.52.02.5 V2 ca r,0/2g (feet)x (feet) Figure 4. Friction coefficient analysis as described by Eq. (11). The inverse of the slope represents the friction coefficient. The distance traveled (xf ) is shown as a function of the initial car velocity ( vcar,0 ).f Friction Factor Analysis The friction coefficient ( mk ) shown in Eq. (5) was needed for solving the system of differential equations. The coefficient is dependent upon the type of surface and the type of wheels contacting the surface. Thus, the coefficient can vary and must be measured for each surface upon which a car is tested. For this work, the friction coefficient was measured by pushing the car by hand, measuring the initial car velocity (vcar,0), and then measuring the final distance (xf) at which the car stopped from the point at which the initial velocity was measured. The initial velocity was measured a short distance from where the car was pushed to ensure that the car was decelerating during the analysis. A ruler was placed at the initial velocity measuring point while a video camera recorded the time for the car to travel a given distance of the ruler (5-13 inches). An average initial velocity was obtained by dividing the distance by the time. Since there was no thrust between the initial velocity point and when the car stopped, Eq. (5) states that dvcar/dt = mk g. Integration of Eqs. (5) and (6) gives dvgdtvvgtcar v car v car k carcark t , ()0 0 09= =mm dxvgtdtxvt gtcar x f carokfcarf k f t f 0 0 2 02 10=-() =-,, () m m Since vcar=0 at tf (the time for the car to travel the entire distance), tf = vcar,0/( mk g) according to Eq. (9). Substitution into Eq. (10) gives x v gf k caro= 1 2 112m,() Thus, a plot of xf versus vgcar,0 22 gives an inverse slope of the friction coefficient.RESULTS AND DISCUSSION Experimental Runs The distances the car traveled during the four experiments are shown in Figure 3 with the dashed lines. The furthest distance traveled was 41.7 feet at an adjusted pressure of 7.43 atm as shown in Table 1. The traveled distance increased with initial pressure as expected. Fr iction Factor The results of the friction factor experiments are shown in Figure 4. Six experiments were performed such that the distance traveled varied between 10 and 30 feet. The wide range of distances allowed for a more complete analysis of the friction coefficient. The plot of xf versus vgcar,0 22 yielded a straight line, which is in agreement with Eq. (11). Regression analysis resulted in an inverse slope of mk = 0.069 0.003 (95% confidence) for the friction coefficient. Model Predictions and Comparison Figure 3 shows the model predictions based on mk =0.069 and a head loss coefficient (C) ranging from 0 to 0.2. The error bars show the range of model predictions when mk ranges from 0.066 to 0.072 (the

PAGE 72

72 Chemical Engineering Education As part of the integrated sophomore and junior team, the students are required to write a safety and environmental report, provide a detailed sketch of the car, build a prototype, provide preliminary and final calibrations, provide an engineering analysis, give a poster presentation, and participate in the department competition.95% confidence interval). As shown in Figure 3, the model predictions were in good agreement with the experimental results when C=0.1. With C=0, the model predictions were much higher than experimental measurements for the three highest initial pressures. However, C=0 is unreasonable since head loss occurs as a result of the nozzle. Model predictions with C=0.2 are lower than experimental measurements for the three highest initial pressures. The model predictions with a range of C values are shown to demonstrate the effect of C on model predictions. W ith C=0.1, the predictions had a difference of 1.0%, 7.3%, and 8.6% from experimental values at initial pressures of 4.76, 6.37, and 7.43 atm, respectively. It must be remembered, however, that the only fitted parameter in the model was the friction coefficient, and the coefficient was measured via a different experiment than the experiment for which the model was used. All other parameters were car dimensions, the initial starting pressure, or the value of C. Thus, considering all of the model assumptions, the model did a reasonable job in predicting the traveled distance. There are several possibilities as to why the model had some disagreement. The first possibility was that the initial starting pressure was lower than the adjusted initial pressure used in the model. In the future, the measurement of the initial pressure following the removal of the clamp would be beneficial. A second possibility was a potential gas leak, such that the contributing pressure to the thrust of the car would be lower. No noticeable gas leaks were observed when running the car, however. A third possibility is a change in the value of the friction coefficient, mk during the course of experiments due to wind conditions and axle friction (since mk was a function of the experimental conditions). No noticeable wind changes were observed and the distances utilized in the evaluation of mk were similar to the experimental runs. The effects of changing mk however, are noticeable by the error bars in Figure 3. The validity of assumptions is an area that could be further explored. W ith the successful demonstration of the model predictions with the experimental results, the impact of car parameters on the traveled distance can be explored. For instance, the effects of varying the rear nozzle diameter, water volume, initial pressure, or friction coefficient (representing an increase or decrease in friction due to changing the type of wheels or the type of surface on which the car travels) can be assessed with regard to distance traveled. This type of exercise allows a student to have a better understanding of how engineering design can affect the function of the car, without the need for numerous experimental designs.CONCLUSIONSThis work describes the effective utilization of engineering principles in a model to predict the distance traveled by a Chem-E-Car using the acetic acid/baking soda reaction. Although the model is specific for one type of car-propulsion system, this work demonstrates how engineering analysis is applicable to the Chem-E-Car competition. One could extend the engineering analysis to include calculation of the theoretical pressure build-up in the reactor, and correlate the theoretical pressure to the experimentally observed pressure in the chamber. Similar analysis could be performed for hydrogen peroxide-catalase reaction systems that generate pressure. Engineering analysis is also applicable to other Chem-E-Car models, such as the iodide clock reaction used to stop a car via breaking an electronic circuitry. For example, the kinetics of the reaction could be incorporated with the momentum equation to predict the time at which the reaction stops the circuitry and the distance at which the car stops. In conclusion, engineering analysis concepts introduced through the Chem-E-Car competition not only provide an opportunity to reinforce theoretical concepts but also provide a tool for the design of the cars.REFERENCES1. (2005) 2.Criteria for Accrediting Engineering Programs, 2003-2004 Accreditation Cycle, Engineering Accreditation Commission; Baltimore, p. 19 3.Madihally, S.V., and R.S. Lewis, "Evolution of the Chem-E-Car Competition at Oklahoma State University," Session 1413d, Am. Soc. of Eng. Ed. Annual Meeting, Salt Lake City, UT (2004) 4.Halliday, D., and R. Resnick, Fundamentals of Physics 2nd Ed., John W iley and Sons, New York (1981) 5.Bober, W., and R.A. Kenyon, Fluid Mechanics 1st Ed., John Wiley and Sons, New York (1980) 6. (2005)