Chemical engineering education

chmial eniern education

Chemical Engineers...

Analyze A Career
With The Number One
Opportunity Company
As you consider your future, consider joining Procter & Gamble,
where there are outstanding research and product development
openings for BS, MS and Ph.D. Chemical Engineers.
No other company can beat P&G's overall opportunity:
Immediate, meaningful responsibility with thorough,
customized on-the-job training and continuing education
opportunities.
Association with top technologists who know your
development is one of their major responsibilities.
Promotion only from within based on the results you
achieve.
Highly competitive salaries and benefits among the
very top in all of U.S. industry.
A heavy financial commitment to support your work.
This year P&G will spend nearly $700 million in research
and product development, among the largest R&PD
programs in all U.S. industry.
Wide recognition of P&G's superiority by industry surveys:
Fortune "Most Admired U.S. Corporations;"
Forbes "Most Innovative Consumer Products Company;"
Computerworld "Ten Best Places To Work;"
Black Enterprise "50 Best Places For Blacks To Work."
Involvement across a wide range of product categories,
including beauty care, beverage, detergent, fabric care,
food, healthcare, household care, paper, pharmaceutical,
industrial chemicals and food service and lodging products.
The challenge to participate in the progress of technological
innovations that have produced more than 15,000 patents
currently in force worldwide and nearly 120 of the world's
top consumer products.
U.S. technical centers in Baltimore, MD; Cincinnati, OH;
Memphis, TN; Norwich, NY; and Shelton, CT; and over 20
technical facilities in other major countries.

For more information on USA and international openings, send
your resume to:
F.O. Schulz, Jr.
The Procter & Gamble Co.
Ivorydale Technical Center (#7CEE)
June Street at Spring Grove Ave.
Cincinnati, OH 45217-1087
An Equal Opportunity Employer

EDITORIAL AND BUSINESS ADDRESS:
Chemical Engineering Education
Department of Chemical Engineering
University of Florida
Gainesville, FL 32611

EDITOR: Ray W. Fahien (904) 392-0857
ASSOCIATE EDITOR: T. J. Anderson
CONSULTING EDITOR: Mack Tyner
MANAGING EDITOR: Carole Yocum (904) 392-0861
PUBLICATIONS BOARD
* CHAIRMAN
E. Dendy Sloan, Jr.
Colorado School of Mines
*PAST CHAIRMEN
Gary Poehlein
Georgia Institute of Technology
Klaus Timmerhaus
University of Colorado
* MEMBERS*
Richard M. Felder
Jack R. Hopper
Donald R. Paul
James Fair
J. S. Dranoff
Frederick H. Shair
Alexis T. Bell
Angelo J. Perna
Stuart W. Churchill
Raymond Baddour
Charles Sleicher
Leslie W. Shemilt
Thomas W. Weber

STATEMENT OF OWNERSHIP. MANAGEMENT AND CIRCULATION

CaIB1CaL CLMEMl NGC EUCATIa I I 01 1 o| 01 01 1 9| I 0//.
Qoartarir 0 See atts cd rates
II0"" t= ... .. a a,!a' I-"" .. .t.
OIIUIC L E.1IIUUhI EDU.T10. ., lo 317, Ohlca l Ellrrlnit l . l.rtnt,
aIltl li of ilo da, Cainl lll, AaIchu, FlorId. 326110
QMCst Eiurlwr. n.rt o, tta~rlu. A oalaltr 9 ar ilnrlq Uacatloo.
11 DuPot Circl. rIsblo t. C 20m03

aSU ChaItl IqiLa DttlL. 11 DuO-, CIlrcl, Uuhlnloa. W-., X 200M
Iar V. F .him, Ca.l l il i. trli D.lper. ., .a 31. UI.L. of nForlid. GciU I.11. Fn
Carole C. lcua Chml l EIl. Dpt.. |to 317, iUta oft Florid. CGmIa1 lll FL 32611

[t.LC' *ot al 4 tOf1o O O** "-

---------217 -------- -- I
,. .:: i. .a+ a . ao :. o...a..

*:I~i'S'- _____ ---..mi*-----*62 --------A

1, m2 2m lua 2' -- -
&. I*M..a...4C.dA 3216 2021
116 59

| -OO** A, 4 4-
S...ilt .....H.... I cfl*-.Ci C. .a 0..0
M i2l. r HM .o-... f. E l A J '

Chemical Engineering Education

Volume XXV

Number 1

Winter 1991

DEPARTMENT

2 Louisiana State University: Development and History
James B. Cordiner, Jesse Coates

EDUCATOR

6 Jim Stice of the University of Texas

LABORATORY

10 A Membrane Gas Separation Experiment for the Undergraduate
Laboratory, Richard A. Davis, Orville C. Sandall

16 An Engineering Applications Laboratory for Chemical
Engineering Students, W.A. Davies, R.G.H. Prince, R.J. Aird

CLASSROOM

24 A "User-Friendly" Program for Vapor-Liquid Equilibrium,
Francisco A. Da Silva, Luis A. Bdez, Erich A. Miiller

28 Teaching Effective Oral Presentations as Part of the Senior
Design Course, E.L. Hanzevack, R.A. McKean

40 A Robust Alternate to Least Sum of Squares for Linear
Regression, G.P. Rangaiah

CURRICULUM

34 Use of a Modern Polymerization Pilot-Plant for Undergraduate
Control Projects, S.A. Mendoza-Bustos, A. Penlidis, W.R. Cluett

46 The Power of Spreadsheets in a Mass and Energy Balances
Course, Michael Misovich, Karyn Biasca

54 Use of PC Based Mathematics Software in the Undergraduate
Curriculum,
Joseph M. Slaughter, James N. Petersen, Richard L. Zollars

RANDOM THOUGHTS

22 Engineering Education Verses, Richard M. Felder

CLASS AND HOME PROBLEMS

50 Amundson's Matrix Method for Binary Distillation Revisited,
J.J.J. Chen

21 Positions Available
45 Book Review
53 Books Received

CHEMICAL ENGINEERING EDUCATION (ISSN 0009-2479) is published quarterly by the Chemical Engi-
neering Division, American Society for Engineering Education and is edited at the University of Florida. Cor-
respondence regarding editorial matter, circulation, and changes of address should be sent to CEE, Chemical
Engineering Department, University of Florida, Gainesville, FL 32611. Advertising material may be sent di-
rectly to E.O. Painter Printing Co., PO Box 877, DeLeon Springs, FL 32130. Copyright 1991 by the Chemical
Engineering Division, American Society for Engineering Education. The 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 informa-
tion on subscription costs and for back copy costs and availability. POSTMASTER: Send address changes to
CEE, Chem. Eng. Dept., University of Florida, Gainesville, FL 32611.

Winter 1990

Department

LOUISIANA STATE UNIVERSITY

Development and History

JAMES B. CORDINER, JESSE COATES
Louisiana State University
Baton Rouge, LA 70803

It is fair to say that the development of chemical
engineering at Louisiana State University began in
1893 with the arrival of Charles Edward Coates, a
professor of chemistry and chemist on the staff of
the Agricultural Experiment Station.
Coates' personal interest in sugar chemistry and
sugar engineering was to dominate the focus of the
department for the next third of a century, although
the topics of chemistry were expanded to include
theoretical, physical, electro, historical, and physio-
logical chemistry. He served for over forty years as a
professor of chemistry and Chairman of the Depart-
ment of Chemistry.11' He also served concurrently
with the Audubon Sugar School, as a professor until
1907 and as its dean from 1907 to 1931.
Copyright ChE Division, ASEE 1991

Coates was appointed dean of the newly-formed,
multi-disciplinary "College of Pure and Applied Sci-
ences" in 1931 and remained in that position until
his retirement in 1937. The LSU Catalog for 1934-
35 '( shows that Division I contained the depart-
ments of chemistry, chemical engineering, and sugar
engineering (Audubon Sugar School), while Division
II consisted of sugar agriculture, agricultural chem-
istry, and biochemistry and Division III was physics
and applied electronics.
The major thrust of this article is directed to-
ward chemical engineering, including its develop-
ment within the Audubon Sugar School, and the fol-
lowing section (largely abstracted from a 1917 publi-
cation by Charles E. Coates'13) is of special interest
in that respect.

AUDUBON SUGAR SCHOOL, 1893-1917[31
(The following are quotes from "An Experiment in

Chemical Engineering Education

the Education of Chemical Engineers. The Twenty-
Fifth Anniversary of the Audubon Sugar School, by
Charles E. Coates.)
The part which the chemist has played in modern
development we have known in a way for some years, but we
are appreciating now as never before, the vital and imperative
importance to our nation of a body of men who cannot only
discover chemical principles but can also apply them
industrially.
A little over a century ago, when sugar was first made
from beets, the root was low in sucrose and the process gave
a poor yield of an inferior grade of sugar with an almost
valueless molasses. . the chemist and engineer, working
together, slowly improved the processes until a good yield of
sugar was turned out, practically pure, and both the molasses
and all the other by-products became sources of profit and
not of loss. In consequence the net cost of beet sugar fell year
by year until it finally became a serious competitor of cane
sugar and, finally, it was offered at prices closely approaching
the cost of cane sugar production.
The sugar planters of Louisiana, .. seeing the increasing
gravity of the situation . in the late [eighteen] eighties,
called to Louisiana Dr. W. C. Stubbs and established... the
Sugar Experiment Station at Kenner, Louisiana, which was
subsequently moved to Audubon Park, on the outskirts of
New Orleans. ... But when the planters began to look for
chemists and engineers, they were simply not to be obtained.
S. In 1890, therefore . it was decided to establish, in
connection with Sugar Experiment Station, a school for the
training of experts in sugar work . opened in 1891 as the
Audubon Sugar School .... The school was successful from
the outset and, in a couple of years, more students were
applying for admission than could well be accommodated.
In the meantime the Sugar Experiment Station was taken
over by the State of Louisiana as part of the Louisiana State
University.... In 1908 its numerical importance was such
that it was reorganized as a college of the University.
From the first the writer [C. E. Coates] and his colleagues
were given a free hand by President [Thomas D.] Boyd in
formulating the course of study, and changes were made
year by year as experience or circumstances dictated. ... The
purpose of the school when first organized was to offer to the
citizens of Louisiana the opportunity to secure such training
as would qualify them to enter most advantageously the
sugar industry of the state.

The course as formulated in 1897, was four years in length.
S. .It soon became clear, however, that a satisfactory
foundation could not be given to high school graduates in
four years, so, in 1899, the course was made five years in
length. The first three years were spent on fundamentals-
chemistry, physics, mathematics through calculus,
economics, English, engineering sciences such as mechanics,
and thermodynamics. These courses were comparable to
those given in the chemical engineering departments of MIT,
Illinois, and Cornell. Specialization in the sugar industry
was reserved for the last two years. Courses in sugar house
control, sugar house machinery, mechanical engineering,
machine design, steam engineering and the like were offered.
The salient feature of this instruction was that it was
accomplished by a combination of classroom work and
practical instruction in the Audubon Sugar Factory and,
later, in various Louisiana sugar factories. So far as the
writer knows, this was the first five years' course in chemical

engineering ever offered in this country.

Was this chemical engineering? In the well-known
textbook by Badger and McCabe,'41 Elements of
Chemical Engineering, examples of the type of work
chemical engineers are concerned with are given.
Listed there we find the flow of fluids, flow of heat,

... it is certain that by 1908 chemical engineering
was firmly established at the university,
and that LSU was the birthplace of
chemical engineering in the south.

filtration, evaporation, crystallization, and extrac-
tion, among others, as chemical engineering opera-
tions. These are also key operations in the manufac-
turing of sugar. Moreover, some of the data taken in
the Audubon Sugar Factory in the early days are
still cited today in Perry's Chemical Engineer's Hand-
book."' Were the students in the Audubon Sugar
School being taught chemical engineering? The an-
swer is clearly in the affirmative. The foregoing facts
speak for themselves.

Whether the date be 1897 or 1908, it is certain
that by 1908 chemical engineering was firmly estab-
lished at the university, and that LSU was the birth-
place of chemical engineering education in the south.

THE DEPARTMENT FROM 1897-1908

Since the primary concern of this article is chemi-
cal engineering education at Louisiana State Uni-
versity, attention will be focused first on the year
1897 when the Audubon Sugar School became an
official part of LSU. It was operated as a private
corporation (with funds subscribed by the Louisiana
Planters' Association) with a course originally of two
years' duration. Additional details of its history were
published in an article by E. A. Fieger.161 Excerpts
from that article follow:

After a careful consideration of the chemical developments
which have occurred in Louisiana, it seemed appropriate to
present the history of one of the first chemical industries of
the state and to show how its introduction led to a series of
developments which had far-reaching effects. This industry
... was born during a period of agricultural adversity. It
developed and flourished, due to the application, diligence,
and patience of a small group of men who probably
unconsciously applied chemical principles to a crystallization
process . and caused an awakening-and its salvation
through the use of chemists and engineers. This is the story
of the sugar cane industry.
If history is correct, the first sugar cane was introduced
into Louisiana by the Jesuits in 1751, about thirty years
after the founding ofNew Orleans.

Winter 1990

TABLE 1
LSU Chairmen (pictured) and Faculty Members, 1893-present

Si
Paul M. Horton
Chairman 1937-55

Jesse Coates
Chairman 1942-43, 1955-67, 1969-70

Paul W. Murrill
Chairman 1967-69

LCarles '. Coates
Chairman 1893-1937

Charles Coates, Johns-Hopkins, 1893-1937
Chairman 1893-1937
Paul M. Horton, Columbia, 1919-58
Chairman 1937-1955
Arthur G. Keller, LSU, 1934-68
Jesse Coates, Michigan, 1936-73
Chairman 1942-43, 1955-1967, 1969-70
C. S. Carlson, Penn State, 1940-4?
Bernard S. Pressburg, LSU, 1941-42, 1945-83
Dale E. Von Rosenberg, MIT, 1957-63
James B. Cordiner, Washington, 1958-81
Frank R. Groves, Jr., Wisconsin, 1958 -
Adrian Johnson, Jr., Florida, 1960-62, 1968-
Paul W. Murrill, LSU, 1960-80
Chairman 1867-1969
Clayton D. Callihan, Michigan St., 1963-83
David B. Greenberg, LSU, 1961-74
Ralph W. Pike, Jr., Georgia Tech, 1964 -
John J. Seip, LSU, 1962-76
Jerome A. Planchard, Tulane, 1967-7?
Alexis Voorhies, Jr., Loyola (Hon), 1964-80
Roger Richardson, Iowa State, 1965-77
Richard C. Farmer, Georgia Tech, 1967-79
Cecil L. Smith, LSU, 1966-79
Albert H. Wehe, Jr., Texas, 1966-76
Edward McLaughlin, London, 1967-68, 1970-
Chairman 1979-1987
Bert Wilkins, Jr., Georgia Tech, 1968-80
Edgar C. Tacker, Florida, 1969-74
Philip A. Bryant, LSU, 1967-80
Armando B. Corripio, LSU, 1968-
Joseph Polack, MIT, 1970-88
Chairman 1970-1976

F1lm
Joseph Polack
Chairman 1970-76

Douglas P. Harrison
Chairman 1976-79

Edward McLaughlin
Chairman 1979-87

John R. Collier
Chairman 1988 -

Douglas P. Harrison Texas, 1971-
Chairman 1976-1979
Arthur M. Sterling, Washington, 1975-
Chairman 1987-88
Ramsey S. Chang, Stanford, 1975-79
Michael Frenklach, Hebrew Univ., 1979-85
Geoffrey L. Price, Rice, 1979-
David M. Wetzel, Delaware, 1979 -
Kerry M. Dooley, Delaware, 1983-
Louis J. Thibodeaux, LSU, 1984-

F. Carl Knopf, Purdue, 1980-
Richard G. Rice, Pensylvania, 1980 -
Danny D. Reible, Cal Tech, 1981 -
Don Ristroph, Pennsylvania, 1982-87
Conrad B. Smith, Houston, 1983-87
Gregory L. Griffin, Princeton, 1988 -
Martin A. Hjortso, Houston, 1988 -
Donald Freshwater, Birmingham, 1988 -
John R. Collier, Case Inst., 1988-
Chairman 1988-

I Chemical Engineering Education

Sugar engineering, as cited in the LSU Catalog
for the year 1902, was one of eight regular courses of
study leading to appropriate degrees. The course
was designed to train experts in the sugar industry
to fill good positions in the field. It included instruc-
tion in the agriculture, chemistry, and manufacture
of sugar. The students had full courses in the lecture
rooms and laboratories of LSU and then spent the
grinding season of their junior and senior years in
the field, the sugar house, and the laboratory of the
Sugar Experiment Station (originally located at
Audubon Park, New Orleans, but moved to Baton
Rouge in 1897). The chemical engineering curricu-
lum is first mentioned in the 1907 LSU Catalog.

THE DEPARTMENT FROM 1909 TO 1936

The roster of students in 1910121 includes several
in the Audubon Sugar School and two sophomores in
chemical engineering; that of 1912 lists possibly the
first graduate student; and that of 1913 lists seven
undergraduates.
Paul M. Horton is listed in the 1919 catalog as
Assistant in Chemistry; in 1925 as Assistant Profes-
sor of Chemistry; in 1927 as Associate Professor of
Chemistry; in 1935 as a Professor of Chemical Engi-
neering (all within the Department of Chemistry).
In 1936, chemistry and chemical engineering are
listed as separate departments within the College of
Pure and Applied Sciences (C. E. Coates, Dean). The
first PhD was awarded in 1935.
A special posthumous tribute to Dean Coates
was instituted in 1957 with the establishment of the
Charles E. Coates Memorial Award for outstanding
contributions to the professions of chemistry and
chemical engineering, the corresponding professional
society and the community. In addition to his many
other achievements, Dean Coates was a charter
member of the Louisiana-Mississippi chapter of the
AIChE and helped to organize the Louisiana section
of the ACS.

THE DEPARTMENT FROM 1937 TO 1956

In 1937 the catalog shows chemical engineering
as a separate department within the College of En-
gineering. Dr. Horton is listed as head, and Jesse
Coates and Arthur Keller as assistant professors.
Dr. Coates ran the department almost single-
handedly during the war years of 1942-45 since Dr.
Horton was on leave working on a high-priority proj-
ect and Dr. Keller was on leave for another assign-

The catalogs for this period indicate that Horton,
Coates, and Keller taught a tremendous
number and variety of chemical
engineering courses.

ment at LSU. Bernard Pressburg joined the faculty
as Assistant Professor in 1941, but was on military
leave from 1942 to 1945.
The catalogs for this period indicate that Horton,
Coates, and Keller taught a tremendous number
and variety of chemical engineering courses. In ad-
dition to the courses listed earlier, Horton conducted
considerable research in the pulp and paper field.

THE DEPARTMENT FROM 1957 TO 1969

The years 1957 to 1963 saw substantial increases
in the complexity of course offerings and in the
number of faculty. Dale Von Rosenberg joined the
faculty in 1957; James B. Cordiner and Frank R.
Groves in 1958; and Adrian E. Johnson in 1960.
Paul W. Murrill came to LSU as a graduate student
in 1960, received his PhD degree, and eventually be-
came department head in 1967. In 1969 he became
Vice-Chancellor, and shortly thereafter Chancellor,
of the Baton Rouge campus, but left in 1980 to be-
come Executive Vice President of Ethyl Corporation,
and then Chairman and Chief Executive Officer with
Gulf State Utilities Company.
Several individuals were permitted early retire-
ment from Exxon Corporation to come to LSU. They
included Alexis Voorhies, who came in 1964, and
Roger Richardson, who joined in 1965. Edward
McLaughlin, from Imperial College of London Uni-
versity, was a visiting professor at LSU for the 1967-
68 academic year, returned to London for two years,
and then joined the LSU faculty permanently.

THE DEPARTMENT FROM 1970 TO THE PRESENT
In 1970 Joseph A. Polack was granted early re-
tirement from Exxon Research and Development
Laboratories to become a professor and head of the
LSU department. He served in that capacity for the
next six years.
In 1976, Polack became Interim Director of the
Audubon Sugar Institute in addition to his duties as
head of the chemical engineering department, but
soon thereafter resigned as head to become the full-
time director of ASI, where he remained until his
retirement in 1988.
Continued on page 33.

Winter 1990

S44ducator

JIM STICE

of The University of Texas

A practical problem intrigued the young professor, Jim Stice: How can we
improve engineering teaching?
Just as most engineers would do, he began to create a
simple, logical solution to the problem. Unpredictably, how-
ever, the problem and its solution went on to consume twenty-
six years of his career.
A leading authority on engineering teaching effective-
ness today, his research began with a 1963 doctoral disser-
tation creating the first integrated approach to teaching
automatic control. It grew into one of the nation's foremost
centers of teaching effectiveness at the University of Texas-
Austin, and Jim served as its director for sixteen years
before returning to the classroom full time.
Jim had his first taste of teaching when, in his first in-
dustrial position, the technicians in his group asked him to
give short courses in mathematics and chemistry during the
noon hour. "I didn't really want to do it at first, but I felt I
should, and before long I found that I enjoyed that session "
more than anything else in the day. It never occurred to me
at that time, however, that I might eventually spend most of
my professional life as a teacher."
Transition of Stice the 11-year-old-boy (top) to
Jim was born in Fayetteville, Arkansas ... home of the Stice the full-fledged-professor.

Arkansas Razorbacks. He has
been a fan of both their football
and basketball teams all his life,
and he still remembers going to
the games and passing out pro-
grams when he was a Cub Scout,
and later when he was a Boy
Scout. "You got into the games
free and really only had to work
about half an hour before you ran
out of programs. It was the best
deal in town!"
After graduating from high
school, he enrolled at the Univer-
sity of Arkansas in the fall of 1945.
"Spider" Stice intended to go to
work in the chemical industry
Jim and his wife Patsy celebrate her graduation, when he graduated four years
Copyright ChE Dwlsion, ASEE 1991

Chemical Engineering Education

later, but only about a third of his class got offers,
and he wasn't one of them. So he opted instead
for graduate study at the Illinois Institute of Tech-
nology, partly because it was in the North and
partly because he wanted to experience life in the
big city. "But mostly, I went there because they
offered me an Armour Research Foundation Fellow-
ship," he adds.
While at IIT, Jim met another young student,
Patricia Stroner, who stole his heart and who later
became his wife.
After graduation he went into industry to seek
his fame and fortune, and worked for a time for
Visking Corporation, which was later bought by
Johnson & Johnson. His only other industrial job
was with the Thurston Chemical Company, which
later became a division of W.R. Grace and Co. Jim
contends that his employment with these two com-
panies had absolutely nothing to do with their sub-
sequent sale.
His industrial experience showed him that he
could function well as a practicing engineer, but at
the same time he found he was often bored with
what he was doing. Then, an early-morning phone
call changed his life. Dr. Maurice Barker, head of
the chemical engineering department at the Univer-
sity of Arkansas, was calling to explain that the de-
partment had lost a professor and needed a last-
minute replacement. He was hoping he could per-
suade Jim to take the job for a year, to help the de-
partment out of the hole it found itself in. He sweet-
ened the pot by suggesting that during that year Jim
could use the University Placement Office to look for
an industrial opening that appealed to him.
"Here I was, someone who had never considered
teaching as as career, becoming an assistant profes-
sor at the State University! Unbelievable. Then, even
though I had never worked so hard in my life, I
found that I really liked teaching, and I began con-
sidering it as a career track instead of industrial in-
volvement." He returned to IIT, got his PhD., and
returned to the University of Arkansas as associate
professor in 1962.
For the next fourteen years Jim did all the things
professors do to "get ahead," but he was always more
challenged by, and found more satisfaction in, teach-
ing than in research. Then in 1968, Johnny
McKetta, Dean of Engineering at the University of
Texas, offered him a job that would involve working
with faculty members in the College of Engineering
to help them improve their teaching skills. Jim ac-

cepted the position even though there was no other
program like it in the country and as a result there
were no precedents, no examples to follow. He was
on his own.
Jim says that half the time he didn't know what
he was doing, but that he certainly enjoyed doing it.
Evidently his efforts were successful since the UT

Stice says, "These people spend a great deal of
time and energy studying the elements of their
disciplines, but little or no time learning how
to communicate, to motivate, to convey
information, insights, and ideas..."

Faculty Senate later decided there should be a simi-
lar office to serve the entire campus and that Stice
should be the one to head it up. Thus, in 1973, the
new Center for Teaching Effectiveness came into
being, with Jim at its helm.
"I have always thought it strange that people
who are hired to teach in a college or university are
not expected to have any training or skills in teach-
ing," Stice says. "These people spend a great deal of
time and energy studying the elements of their disci-
plines, but little or no time learning how to commu-
nicate, to motivate, to convey information, insights,
and ideas to others.
"I have tried to attack the problem in two ways.
First, the University of Texas had an old rule that
all graduate teaching assistants were supposed to
take a course in 'supervised teaching.' I searched the
campus over, however, and found only one such
course out of sixty-eight departments! So I began
teaching a course in the chemical engineering de-
partment in the summer of 1972 and persuaded
several professors in other departments to sit in on
it so they could eventually offer a similar course in
their own discipline. Now we have such courses in
about thirty departments, and most of them seem to
be going well. If any of their students elect higher
education as a career, they will be a lot better pre-
pared than I was!
"Second, I tried to persuade various university
administrators to let us provide a similar experience
for new faculty members. No one showed any inter-
est until Dr. Peter Flawn arrived. He allowed us to
give it a try, and the program was quite successful.
We now have a three-day seminar for all new fac-
ulty, regardless of rank, at which our own people
give presentations on everything from writing in-
structional objectives to teaching creativity. The

Winter 1990

President foots the bill for two free lunches, coffee
breaks, an end-of-seminar Attitude Adjustment
Workshop, and all the handout materials. Cost is
around $3,000. It would be a bargain at five times
that price.
"We now have about 750 'graduates' of the
program. Even if they don't use all the ideas
that were presented to them, they are consider-
ably more sophisticated about what they are trying
to accomplish.
"In 1986, just before the beginning of the spring
semester, we began a similar (two-day) program for

For ... fourteen years Jim did all the
things professors do to "get ahead," but
he was always more challenged by, and found
more satisfaction in, teaching than in research.

experienced faculty. The response both surprised
and gratified us-attendance has been over 150 for
each of the past three years! That kind of response
demonstrates that many faculty members really do
care about teaching and that they will readily par-
ticipate in a program like this if it is available.
Additionally, many of the attendees offer to give a
presentation in the following session-there are more
offers than we can accommodate. It is a self-renewing
program, and it costs peanuts."
Stice's philosophy is simple-offer practical help
that can be implemented immediately. "There are a
lot of things we could use in our teaching if we knew
about them. It's just that no one ever told us about
them, and it's inefficient for us to discover them for
ourselves," he says, and adds, "Teaching is an an-
cient, honorable, and extremely important profes-
sion. It can be tremendously satisfying. However, we
are capable of doing it much better than we are now
doing it, and I think 'educating' the faculty is the
first step toward that goal."
Jim was one of the early engineering educators
who investigated the possibilities of using the digital
computer in computer-aided instruction. Somewhat
later he became aware of Fred Keller's work with
the Personalized System of Instruction (PSI), also
known as the Keller Plan. With the aid of a grant
from the Alfred P. Sloan Foundation, he and a group
of faculty members developed nineteen PSI classes
and compared student performance in them with
that of students in conventional classes in the same
subjects. They concluded that, in most cases, the PSI

students learned more and remembered it longer.
Another innovation was the Student Input Proj-
ect (SIP), funded by the Exxon Education Founda-
tion. It established that periodic meetings between
faculty members and designated members of a class,
throughout the course, resulted in better satisfac-
tion with the class by both students and instructor.
It also furnished the instructor with useful feedback
and allowed changes to be made in class organiza-
tion, structure, or procedures while the class was
still in progress. "This method was more useful than
the more conventional end-of-course evaluation where
suggestions were received too late to be incorporated
into the course."
In the 1980s, Jim became interested in efforts to
teach problem solving- particularly the work of
Lois Greenfield (at the University of Chicago), Don
Woods and his colleagues (at McMaster University),
and Art Whimbey and Jack Lochhead (at the Uni-
versity of Massachusetts). All these teachers used
pairs of students to discuss problem comprehension,
analysis of elements, formulation of a plan, its solu-
tion, and evaluation of the solution.
"When I was an undergraduate student," Jim
says, "I was a memorizer. I could do things that I
had done before, but when a teacher gave us a new
or different situation, I was stumped. This caused
some problems in my junior and senior years...but it
was potentially disastrous in graduate school where
the tests routinely dealt with things we had not
specifically covered in class. What they were trying
to do, of course, was teach us to think. (I viewed it,
however, as a dirty plot to flunk us out.) If it hadn't
been for one of my roommates, who undertook to
show me how to analyze, I may have become a victim
of that imagined plot. After several weeks I began to
see that there was a strategy to this business, and
my work started to improve. It was almost thirty
years later that I realized that my roommate had
been doing pairs-learning with me!"
Lately Jim has become interested in learning/
teaching styles. He says, "For a good many years
years after I started teaching, I guess I thought that
most of my students learned things the same way I
had learned them. But then I heard about the
Canfield profile, and later the Kolb learning-style
inventory, and a whole new way of looking at the
learning process opened up for me. I began to real-
ize that while some students could learn readily
from Professor X, others in the same class found him
puzzling, disorganized, and difficult: some students

Chemical Engineering Education

wanted details while others preferred a global ap-
proach; some loved everything about a course while
others were bored out of their skulls (and a few of
the latter changed majors as a result).
"Discovering the difference in learning styles has
made me think hard about the way I handle my own
courses. As a result, I have changed the way I do
some things in order to reach more of the students,
and the result has been greater satisfaction for both
the students and myself."
Together with Rich Felder and Rebecca Leonard
of North Carolina State University, Stice has a new
project: a three-day National Effective Teaching In-
stitute (NETI) for engineering and engineering tech-
nology instructors. It will be held just prior to the
1991 and 1992 annual conferences of ASEE, and its
goals are to

improve the teaching effectiveness of the individuals
participating, and
provide an outline for courses in college teaching for
graduate teaching assistants.

Participants will not have to pay any fees for the
NETI-registration, coffee breaks, luncheons, and
all handout materials will be provided free of charge.
Participants' deans will have to nominate them and
agree to pay for their transportation, hotel, and mis-
cellaneous expenses. DuPont, Union Carbide, and
Dow Chemical have already signed up as sponsors
for the institute. More information and details about
applying will appear in Engineering Education prior
to the conference.
Jim has been a member of ASEE since 1962 and

"Teaching is an ancient,
honorable, and extremely
important profession. It
can be tremendously
satisfying....If I had it
all to do over again,
I would try to do
about the same things..."

is an active participant in both the Chemical Engi-
neering and the Educational and Research Methods
Divisions of that organization. He has held numer-
ous offices, including Chairman of the Chemical En-
gineering Division, and is currently Chairman of
Professional Interest Council (PIC) 1 and a member
of the ASEE Board of Directors. He has also been a
member of the American Institute of Chemical Engi-
neers for thirty years.
Stice was named T. Brockett Hudson Professor of
Chemical Engineering in 1985, and Bob R. Dorsey
Professor of Engineering in 1990. He may be the
only professor on the UT campus who holds a named
chair because of his teaching activities rather than
his research-certainly, he is the only one in the
College of Engineering. When asked about this, he
said, "It surely would have been easier to go the
conventional route and get research funding, sup-
port graduate students, write technical articles, and
all the rest of it. Colleagues, although they are will-
ing to let everyone do their own thing, still do not
value research and publication on questions pertain-
ing to engineering education as much as they value
regular research. So I have had to develop a thick
skin and stay pretty fast on my feet. But I really
believe administrators are willing to provide encour-
agement to people who are sincerely interested in
good teaching.
"Teaching can be tremendously satisfying. If I
had it all to do over again, I guess I would try to do
the same things again. I have had a lot of luck, have
met some really great people, and have had a bunch
of fun along the way. And the students make the
whole show worthwhile." [

Winter 1990

laboratory

A MEMBRANE GAS SEPARATION

EXPERIMENT FOR

THE UNDERGRADUATE LABORATORY

RICHARD A. DAVIS, ORVILLE C. SANDALL
University of California
Santa Barbara, CA 93106

Synthetic membranes have been the focus of much
attention recently because of their simple and
economical operation for separating gases. The Per-
mea Corporation offers a Prism separator package
as a laboratory-scale system for demonstrating
membrane gas separation. The apparatus consists of
four columns, with each column being two inches in
diameter and four feet long and filled with bundles
of hollow fibers. The system can be conveniently
used to separate oxygen from the air. We purchased
one of the units, have used it in our required senior
laboratory course for the past three years, and have
found it straightforward to operate.
For the first two years the suggested objective for
the students was to determine the effects of pressure
and feed-flow rate on the degree of separation. The
data analysis required to meet this objective involved
only overall mass balances. We felt that the experi-
ence was not satisfactory for senior-level chemical
engineering students, so we modified the apparatus
and changed the objectives of the experiment in or-
der to make the apparatus more suitable for our

Richard Davis received his BS degree in chemical
engineering from Brigham Young University in 1987.
He is currently a PhD candidate in the Department of
Chemical and Nuclear Engineering at the University
of California, Santa Barbara. His thesis research
pertains to gas separation by facilitated transport in
liquid membranes.

Orville Sandall is a professor in the Department of
Chemical and Nuclear Engineering at the University
of California, Santa Barbara. He is a graduate of the
University of Alberta (BSc and MSc) and the Univer-
sity of California, Berkeley (PhD). His teaching and
research interests are in the area of mass transfer

A Air Feed
B Pressure Gauge
C Oxygen Analyzer
D Flow Meter
E Permeate
F Non-Permeate
G Hollow Fibers

FIGURE 1. Flow diagram for laboratory scale
Prism' separator system.

laboratory. The plumbing was modified so that meas-
urements could be taken either on a single column
or on the original four-column arrangement. An
analysis was carried out and software was written
for the single-column data in order to determine the
transport properties needed to predict membrane
performance. A software package was also devel-
oped to use these parameters as determined from
the single-column data in order to predict the air
separation to be achieved with the four-column ar-
rangement and to compare with the observed data.
This paper describes the new objectives and pro-
cedures that were used to increase the technical
content of the experiment and to teach the students
about the fundamental mass-transfer characteris-
tics of membranes.

EXPERIMENTAL DESIGN

The experimental apparatus consists of four
Prism" separator columns arranged as shown in
Figure 1. Each column contains thousands of non-
porous, semipermeable membranes in the form of
Cpyright ChE Division, ASEE 1991

Chemical Engineering Education

... we modified the apparatus and changed the objectives of the experiment in order to make the
apparatus more suitable for our laboratory ... This paper describes the new objective and procedures
that were used to increase the technical content of the experiment and to teach the students
about the fundamental mass-transfer characteristics of membranes.

hollow fibers. The oxygen permeates through the
fiber walls and is collected in a manifold at the
bottom of the separator. The less-permeable nitro-
gen passes through the column and exits from the
top of the separator.

A column consists of a shell with a hollow-fiber
membrane tube-bundle potted at each end, similar
to a shell-and-tube heat exchanger."' A filtered,
compressed-air stream is fed to the bottom of the
first column. The high-pressure air stream, fed to
the bottom of the first column, flows through on the
shell side of the membrane tube-bundle. The pres-
sure is measured at both the feed and outlet of the
high pressure, non-permeate stream. The hollow-
fiber tube bundles are capped at the top so that the
permeate, or oxygen-rich, streams leave from the
hollow-fiber membranes at the bottom of each
column. The permeate streams are arranged in
parallel and exit through a common manifold. The
non-permeate streams are connected in series. Thus,
the first and third separators operate in counter-
current flow conditions, and the second and fourth
separators operate in a cocurrent flow pattern. Two
oxygen analyzers measure the percent of oxygen in
the exit non-permeate and the permeate streams.
The exit stream of the non-permeate side is con-
nected to a volumetric flow meter. The flow rate of
the feed and permeate streams can be calculated by
mass balances.

The original apparatus allowed for only a four-
column separation. The system was modified so that
measurements could also be made on a single col-
umn. In single-column operation, the conditions are
counter-current flow, and such a modification en-
ables the student to determine the important mem-
brane transport properties from measurements taken
on a single column.

The objectives of the experiment are:
1. To determine the separation factor as a function of
pressure and non-permeate feed flow rate for mass
transfer in a single separator. The separation factor is
defined as
([02]/[N2]) 02 enriched stream
([02]/ N2]) N2 enriched stream
and to compare this with the ideal separation factor
defined as the ratio of the permeabilities of the more-
permeable species (A = 0) to the less-permeable species
(B = N)

a A (2)
QB
2. To predict the exit oxygen concentration of the non-
permeate stream for the four-column setup based on
the analysis of the first separator. This should be
repeated for several pressures and flow rates and
compared with actual values from experiments.
3. To compare the degree of separation between cocurrent
and counter-current flow conditions.

THEORY

The membrane separators are modeled with the
assumption that air passes through the column with
no axial diffusion or mixing. It is also assumed that
the amount of gas permeating is small enough and
that the feed gas rate is low enough so there is no
axial pressure drop on either side of the membrane.
This is a good assumption for the apparatus de-
scribed here with the high-pressure feed passing
on the outside of the hollow-fiber tubes.'21 The
other critical assumptions are that the membrane
is homogeneous, that the gas permeabilities are
constant, and that there is no gas phase mass trans-
fer resistance.

The governing equations presented here are well-
known.'34' The equations are developed here for co-
current flow conditions (see Figure 2). The results
for a counter-current flow pattern are presented
after this derivation.

For cocurrent flow conditions, the flux of 02 from
the high-pressure side to the low-pressure side in a
volume element is described, using Fick's Law and
assuming ideal gas behavior, by

differential volume element

membrane

Feed
Gi.xi

closedend G', y, p

G.x,P

I z +dz

Permeate
Go', yo

---+

Non-permeate
Go, xo

FIGURE 2. Diagram of a separator with cocurrent
flow conditions.

Winter 1990

-d(Gx) =QA (Px-py)dA
The flux of N2 is described by
-d[G(l-x)] = -[P(1 x)-p(1-y)]dA
where the differential area is
dA= adV= andz
4

In these equations, QA and Q,, are the permea-
bilities of oxygen and nitrogen, respectively; 6 is the
membrane thickness; a is the interfacial membrane
area per unit volume of separator; and d is the inner
diameter of the column.
The equations are difficult to solve as they stand.
Also, the values for QA, QB, A, and 6 are unknown for
the apparatus. These problems are avoided by com-
bining the unknown parameters and making use of
a more convenient form developed by Walawender
and Stern.'3' These authors use the overall material
and species balances for 0 (Eqs. 6 and 7) together
with the flux equations (Eqs. 3 and 4) to formulate
the differential equations describing the concentra-
tion profiles for 02 in the non-permeate and perme-
ate streams.
Gi=G+G' (6)
Gi x, = Gx+G'y (7)
The combined form of Eqs. (3) through (7) is

G,_ d^ (1-x)- (Px-py)-x [9-)p(1-x)-p(1-y)]l
dA y-xi 8 \
(8)

G dy- (-y QA (Px-py)-yQBp-x)-P-)
dA x-x, )L 6 ) 1 6
(9)
The equations are further simplified by substituting
the ideal separation factor, a the dimensionless
differential height, z and r, the ratio of the high to
low pressures, to give
r = P (10)

K d x-x [(i x)a (rx-y)-x[r(1 x)-(1 y)]] (11)
dz y-x,)

K dy= (1-y) (rx-y)-y[r(1-x)-(1-y)] (12)
dz x-xi
where
K= G 2 G,2 (13)
QB h and2 QBphand

K and a become the key transport parameters that
describe the separation process.
The value of y at the closed end, which is needed
to integrate Eqs. (11) and (12), can be evaluated by
noting that G' = 0 at z = 0.'1-'1 The ratio of Eqs. (3)

and (4), with the appropriate substitutions of a and
(3) r, becomes
d(G'y) a (rx-y) (14)
d[G'(1-y)] r(1-x)-(1 y)
The left-hand side of Eq. (14) can be rearranged to
the following form:

d(G'y) y G'dy
+ (15)
d[G'(1-y)] 1-y (1-y)d[G'(1-y)]
At z = 0, the last term in Eq. (15) vanishes, and the
result is substituted into Eq. (14), yielding a quad-
ratic equation for y,
Y' a (rxi-yi) (
1-yi r(1-xi)-(1-yi)
Eq. (16) can be solved for y,

(c -1)(rxi+l)+r- [(a -1)(rx+1)+ri -4c rx,(c -1)
y ----' I ------ (17)
(2a -1)
Eq. (12) is indeterminate as z approaches 0, and
special consideration is required to evaluate the de-
rivative at z = 0, which is needed to start the nu-
merical integration. L'Hopital's rule'3'4' is used to
obtain this value

dy (x- Yi)rxa -yia )]
^dzl L (x-y)[(a -1)(2y -rxi-1)-r]]
K-
dx
dz )' ,
Integration of Eqs. (11) and (12) describes the sepa-
ration that will be achieved for cocurrent operation.
The flow pattern for counter-current conditions
is shown in Figure 3. The final form of the governing
equations is
K' =-d (x[1-x)a'(rx-y)-x[r(1-x)-(1-y)]] (19)
dz xy-xo

dzX-X,, J X.

Permeate
Go', yo

membrane

Feed
Gi, xi

differential volume element

G', closed end

G,x P

Non-permeate
Go, xo

z+dz

z-O

FIGURE 3. Diagram of a separator with counter-current
flow conditions.

Chemical Engineering Education

K'= G 4G, (21)
K'QB ph and Q Bph and
r 4
In this case, the value of y at the closed end is

a -1)(rx+1)+r- [(a -1)(rx,,o+l)+rl -4a rxo a -1)
i=----------/ -- \---------
(2a -1)
(22)
Eq. (20) at z = 0 becomes (22)

dy (xo-yi)r [ -y )] (23)1
ddz K { (xo -Yi)[( -1)(2yi-rxo-1)-r }

K- dx
dz ,
The system of non-linear initial-value differen-
tial equations is solved simultaneously by using
a fourth-order Runge-Kutta numerical scheme.
The required initial values are determined from
experiments.

METHOD OF SOLUTION
Initially, the student is required to evaluate a:
for the 0/N2 membrane system and K' as a function
of the non-permeate exit stream flow rate, using
data from a single column operating under counter-
current flow conditions. The evaluation of a and K'
is accomplished by making several measurements of
the non-permeate and permeate exit compositions
for a range of non-permeate flow rates and pres-
sures. The governing equations (Eqs. 19 and 20) are
integrated with the known inlet and exit conditions.
The only unknowns are a and K'. The solution re-
quires "shooting" for the known end conditions with
guesses for a and K' until the experimental end
conditions are met and the solution converges on the
desired values for a and K'. This is very similar to
solving a two-point boundary-value problem. A sys-
tematic procedure, based on a modified Newton
method, was developed to iterate on subsequent trials
for a and K' until acceptable convergence criteria
were satisfied. Generally, this method requires less
than ten iterations to achieve convergence. The stu-
dents can easily arrive at a good initial guess for a
based on their experimental data. A reasonable esti-
mate for K' is more problematic. It is possible to
estimate K'/G from data in the literature. However,
since this would require a considerable amount of
student time, we give the students an approximate
value to start the calculation (K'/Go = 4000 s/gmol).
The information for a and K' is used to make

predictions for separations in the four-column ar-
rangement and to model the separation for compari-
son between cocurrent and counter-current flow pat-
terns. Three programs are provided for the students
to use. The first program requires information from
experiments on the first column and solves for a
and K'. The other two programs solve the 0, concen-
tration profiles for either cocurrent or counter-cur-
rent flow conditions based on the initial conditions
specified by the user. The required input values for
all the programs are the pressure, the mole fraction
of 02, and the flow rate for the inlet high-pressure
stream. The program for counter-current flow calcu-
lates K' based on iterated computed results and mass
balances. Note that for counter-current flow condi-
tions, K' is a function of G., and that K for cocurrent
flow conditions is a function of G. This does not
create a problem for our design because the columns
are arranged so that each counter-current column is
followed by a cocurrent column. In this case, K = K'
from the previous column.
These programs, along with mass balances, are
used to make predictions for the separation that
occurs in the four-column arrangement. The pro-
grams are run on an IBM PC which is located in the
laboratory. Thus, the students can analyze their data
while the apparatus is running, and the analysis can
be used to set operating conditions. Listings of the
programs developed here in True BASIC" are avail-
able from the authors.

SAMPLE CALCULATIONS

The data for the calculations presented here are
from actual student experiments. Sample data from
operating the single-column arrangement are listed
in Table 1. Exit 02 mole fractions and non-permeate
flow rates are reported for three feed pressures. A
plot of the experimental separation factor against
the feed flow rate in Figure 4 shows how pressure
has a large effect on the degree of separation inde-
pendent of the flow rate.
Next, the differential equations (Eqs. 19 and 20)
are solved for a and K', using the results from
Table 1. The calculated results for a and K' are also
presented in Table 1. It may be seen that a is a
constant independent of flow rate and pressure as
predicted by theory, and that K' is independent
of the non-permeate pressure, P; thus the results for
K' vs G for the three pressures are plotted together
in Figure 5. It may be seen in Figure 5 that, as
expected, K'is a linear function of flow rate. The

Winter 1990

results for K' as a function of G are correlated by an
equation of the form
K'= mGo (24)

where m is the proportionality constant from
Eq. (21) 45
m=- (25)
QB phand2
For this data, a least squares fit of the results yields
m = 3974 13 s/gmol. The average value for a
is 5.90 0.11. It is immediately evident that a is
a poor approximation for the actual separation
factor, a, when this is compared with the results in
Figure 4.

This information can also be used to compare
cocurrent with counter-current separation in
membranes operating under plug-flow conditions.
Non-permeate O2 mole fractions for these two
cases are plotted against the dimensionless column
length in Figure 6. The profiles are for P = 653 kPa;
p = 101 kPa; G& = 0.0117 gmol/s; and xi = 0.21.

Next, predictions are compared with experiments
for separation in the four-column arrangement.
The prediction requires several mass-balance
calculations and an understanding of the operat-
ing parameters. A sample calculation is given
for the prediction of a for four columns with
x,, = 0.21; GI, = 0.0355 gmol/s; P p = 552 kPa;
and T = 25 C. The numbered subscript refers to the
separator column.

The first column is in counter-current flow. The
differential equations for this condition are solved

Table 1
Data From a Single Column with Counter-Current Flow

P(kPa) Go x 102(gmol/s) Xo y a* K'

377 0.73 0.18 0.43 5.87 31.01
377 0.74 0.18 0.43 5.87 31.01
377 1.03 0.19 0.44 6.00 49.34
377 1.32 0.19 0.44 6.00 49.34
377 2.54 0.20 0.44 5.74 98.41

515 0.62 0.15 0.45 5.93 26.10
515 0.73 0.16 0.46 6.01 33.14
515 0.95 0.17 0.47 6.16 43.73
515 1.51 0.18 0.47 5.84 57.96
515 2.25 0.19 0.48 6.00 91.84

653 0.74 0.14 0.46 5.77 31.40
653 0.95 0.15 0.47 5.82 38.64
653 1.32 0.16 0.48 5.92 48.89
653 2.18 0.18 0.49 5.75 85.30
653 3.44 0.19 0.50 5.82 134.70

numerically with the initial conditions for the prob-
lem. The results give x,.o = 0.189 and y,.o = 0.502. The
equations for the mass balance are
0.0355= G1, +GCo

(0.21)(0.0355)= 0.189 G,.o +0.502 G;'.
These equations are solved simultaneously for
G1,' = G, =0.0331 and G'I,, = 0.00236 gmol/s. This
value multiplied by m gives K'I = K, = 131.6.

The next column is cocurrent flow and the initial
conditions are the results from the material bal-
ance around column 1. The numerical results are
x,o, = 0.170 and y2.o = 0.460. A mass balance per-
formed around column 2 for G2,o gives
0.0331 = G ,, +G',,

(0.189)(0.0331) 0.170 G2,o + 0.460 G2,
The solution yields G2. = G3, = 0.0309, and

0 377 kPa -
0 0 515 kPa -
A A 653 kPa
5.0-

45 -

G, x 102 (kgmol/s)
FIGURE 4. The experimental separation factor, a,
as a function of G, and P.

0.0 05 1 0 1.5 2.0 25 30 3.5
x 102 (kgmol/s)
FIGURE 5. K'as a function of G for a single column
in counter-current flow.

Chemical Engineering Education

G'2. =0.00221 gmol/s.
The above procedure is repeated for the next two
columns. The results are x3o = 0.151; yo = 0.429;
G3, = 0.0287; G'2o = 0.00213 kgmol/s; x4, = 0.133;
Y4o = 0.383; G4, = 0.0267; G'4, = 0.00201 gmol/s.
The calculation of the separation factor is not as
straightforward as that for the single column. The
average mole fraction of 02 in the outlet permeate
stream, y., is found by weighting the value of y from
each column with the permeate flow rate
4
i ynoG'no
yon (26)
Y G'.
n=l
This calculation gives y. = 0.446. The predicted sepa-
ration factor is a = 5.24, which compares very well
with the experimental value a = 5.26. Experimental
........ I ..... ... ,. ', .... I .... .. . I ........
S--- cocurrent
020 counter-current -

0.18 ,

x ... . . .. . .

0.0 0.2 0.4 0.6 0.8 1.0
Z
FIGURE 6. Comparison of mole fraction profiles for
co- and counter-current flow.

ap
FIGURE 7. Comparison of experimental results with
predictions for separation in four columns.

Winter 1990

data are listed in Table 2 for several trials using four
columns. The separation factors are compared with
predictions from the model for several pressures and
flow rates in Figure 7.

CONCLUSIONS

This membrane experiment provides the students
with experience in fundamental engineering skills
such as mass balances, modeling, and using the
computer as a research tool. They are also exposed
to a new separation method that employs membranes.
Without the analysis presented here, the students
are only able to carry out performance tests of the
apparatus. A simple modification of the apparatus
and implementation of the numerical procedure de-
veloped here permits the students to determine the
appropriate transport properties of the membrane
separator. Knowledge of these properties allows in-
tegration of the design equations to predict separa-
tor performance.
Our experience has been that it is too much to
expect undergraduates to derive, on their own, the
numerical techniques presented here. However, we
find that when they are given a handout on the
equation derivations, together with an explanation
of the numerical procedure, they are able to define
meaningful experiments in order to determine the
important transport properties and are also able to
predict separation performance.
Some interesting questions for the students to
consider are: Why are the experimental separation
factors less than the ideal separation factors? Why is
the separation factor an increasing function of non-
permeate pressure and a decreasing function of gas
flow rate? How can the individual units be arranged
to maximize the separation?
Continued on page 21.

TABLE 2
Data From Four-Column Experiments

P(kPa) Gi x 102 (gmol/s) Xo

y ap

377 2.36 0.16 0.40 3.50 3.50
377 2.77 0.17 0.41 3.39 3.42
377 3.76 0.18 0.42 3.30 3.31
515 3.30 0.15 0.43 4.27 4.31
515 4.95 0.17 0.45 3.99 4.10
515 5.82 0.18 0.46 3.88 4.00

3.55 0.13
4.66 0.15
6.30 0.16

0.44 5.26 5.24
0.44 4.45 4.40
0.47 4.66 4.64

laboratory

AN ENGINEERING APPLICATIONS

LABORATORY FOR CHEMICAL

ENGINEERING STUDENTS

W. A. DAVIES, R. G. H. PRINCE, R. J. AIRD1
University of Sydney
Sydney, New South Wales, Australia 2006

The department of chemical engineering at the
University of Sydney has recently commissioned
a new laboratory for first-year students. In a break
from traditional introductory undergraduate practi-
cal work which was confined largely to chemistry
and physics laboratories, this new venture puts un-
dergraduates face-to-face with an authentic process
engineering plant during their first weeks at the
university.
In a closely-supervised environment, student
groups are confronted with rigs built from full-sized
industrial machinery and equipment. They must
draw a flow sheet, dismantle and draw key compo-
nents, reassemble the parts, operate the rig, and
interpret the run data.
The laboratory, which completed its inaugural
semester in the first half of 1989, received immedi-
ate approval from the students, who felt that the
experience identified them as engineers from the
outset of the course. This was a most gratifying
response, especially since one of our major goals was
to integrate the practical and the theoretical aspects
of engineering and to do so in an interesting and
relevant way.
In this article we will describe both the physical
features of the laboratory and the nature of the
course built around it.

LABORATORY HARDWARE

There are eleven rigs, of which eight are near-
duplicate pairs. The rigs are built around key com-

SLoughborough University of Technology,
Loughborough, Leicestershire, England LE11 3TU

ponents comprised of process pumps, control valves,
steam traps, shell-and-tube heat exchangers, a plate
and frame filter press, a pressure-relief valve, and a
parallel-plate heat exchanger. Each rig performs a
simulated process. The pumps recirculate water from
a tank through a network of valves and flow meters;
the plate heat exchangers heat a viscous process
fluid using steam and then cool it using water; the
pressure relief valve lets air out of a holding tank
when it is set above a certain pressure. Each rig is
equipped with measuring instruments which are
appropriate to the task, such as pressure gauges,
flow meters, temperature gauges, and motor-speed
indicators.

Wayne A. Davies received his BSc and PhD at the
University of Sydney. After several years in biomedi-
cal research he returned to mainstream engineering
with strong interests in biological process engineering,
computerized control, and mineral processing. He is a
consulting engineer on a wide range of problems for
industry and government, mainly in nsk management
and waste disposal, and teaches part-time in the de-
partment at the University of Sydney.

Rolf G.H. Prince received his BE degree in chemical
engineering in New Zealand and his PhD at the Uni-
versity of Sydney. He was the Foundation Professor
of Chemical Engineering at the University of Queen-
sland and has been head of the Sydney department
since 1969. His research interests include distillation,
process modeling, fuel alcohol, and the use of expert
systems in design.

Robert J. Aird received an honours degree in me-
chanical engineering from the University of Durham
(UK). He has worked for the UK Atomic Energy Au-
S'i thority and for Canadian Westinghouse, as well as
S.holding a long-term appointment at the Loughbor-
Sough chemical engineering department, where he
has interests in plant reliability and energy monitor-
ing. He currently works for Brush Electrical Machines
S Ltd., a division of Hawker Siddeley.
Copyright ChE Division, ASEE 1991

Chemical Engineering Education

In a closely-supervised environment, student groups are confronted with rigs built from full-sized
industrial machinery and equipment. They must draw a flow sheet, dismantle and draw key
components, reassemble the parts, operate the rig, and interpret the run data.

LABORATORY PROCEDURE

In the first stage of the work, students must
become familiar with the rig, its construction, and
its function. Students are provided with notes which
give a general introduction to the rig and go on to
describe, in detail, the exact method of dismantling
the key component. They are instructed to study
each rig carefully, referring as needed to literature
which is permanently on display in a notice-case
outside the laboratory. All valves are uniquely num-
bered, and a key in the notes describes each by a
functional name. Students must draw a flow sheet,
using accepted process symbols, and label it appro-
priately. During this time they are quizzed by the
demonstrators to determine their level of compe-
tence to proceed to the next stage.
In the second stage students dismantle a key
component of the rig and then return it satisfactorily
to service. To do this safely, it is necessary to shut
down the rig and to isolate it from all sources of
energy. Following standard industrial safety prac-
tices, students must submit a "permit-to-work" form,
describing the work they intend to do, to their super-
visor. If all is in order, the supervisor signs the form
and proceeds to lock off the electricity, steam, or
compressed-air supply as appropriate.
During dismantling, students learn the correct
names of machinery parts and of the tools used.
They also get a feeling for the design and the materi-
als used, as well as for the logical order of doing
work on a piece of equipment. Having disassembled
the component, the students then make a drawing of
its key features. These may vary from intricate ex-
ploded views of the pumps or safety-relief valve, say,
to simpler drawings of the internals of a heat ex-
changer. They then reassemble the unit. If the su-
pervisors agree that it has been returned to an op-
erational state, they sign the "return-to-service" sec-
tion of the permit and remove the lock from the
energy supply.
In the third stage, students start up the process
and check to see that it is operating satisfactorily
(hoping they will not find any malfunction which
would involve a time-consuming dismantling job).
Upon startup, the students ensure that all process
parameters (such as pressures, temperatures, and

flowrates) are in their correct ranges and that no
leaks are evident. They must then vary an impor-
tant parameter, such as the flowrate or valve posi-
tion, and observe the effect on a measured value
somewhere in the system. Simple calculations are
done (such as heat and mass balances) and relevant
data are plotted.
In the final stage students present their findings,
which are written in a workbook. Their report
contains the flowsheet and exploded drawings of
key components of the rig, together with answers to
set questions in the notes and answers to spontane-
ous questions asked by the demonstrators. The en-
tire laboratory session is rigidly confined to a three-
hour period, and all writing must be done in that
time. At the end of the session books are handed in
for grading.

EQUIPMENT AND PROCEDURE DETAILS
Control Valve *

Each of the two rigs include a pumped recircula-
tion loop which conveys water from a header tank,
through a control valve and flowmeter, then back to
the tank. The control valves are isolatable with
manual gate valves so that they can be dismantled
without having to empty water from the entire sys-
tem. Students must dismantle the valve to display
the plug and seat.
The two valves dismantle in different ways, and
the two student groups compare valve construction
with each other-especially the shape of the plug,
which is sketched as a record. Upon reassembly, the
relationship between the valve plug and seating and
the position of the index mark become evident as the
students attempt to readjust the valve's stem posi-
tion. If all is in order there will be no leaks when the
pump is turned on, and the valve will close off al-
most completely or open to full capacity at appropri-
ate settings of air pressure to the actuator. Both rigs
have 500 L header tanks, and the pumps deliver up
to 150 L/min through 50 mm pipes and fittings.
Process Pump *

These rigs are similar to the above except that
the focus is on the pump itself, and therefore manual

Winter 1990

butterfly or gate valves replace the control valves.
The two, somewhat larger, centrifugal pumps are
quite different from each other. One is designed for
clean liquids and is a vertical mount design with
shrouded impeller, mechanical seal, and seal flush-
ing line. The other, a pump for mineral slurries, has
an open, rubber-lined centrifugal impeller seal. Stu-
dents dismantle the pumps as far as the sealing
mechanisms which they draw in exploded diagrams.
After reassembly, the pumps must operate without
leaks, undue vibration, or other malfunction. A stan-
dard flowrate versus pressure curve is drawn for
each rig, and students from the two groups compare
the results, together with impeller diameters and ro-

Students measuring spacings of plates on the
plate heat exchanger.

national speeds. Students are also asked to describe
the principle of the centrifugal pump and the con-
cepts of "priming and cavitation."
Pressure Relief Valve *

The rig consists of a pair of similar 100 L
compressed-air tanks connected by 25 mm piping
with a ball valve. One tank is connected directly to
the high-pressure main and can be filled with air to
about 600 kPa. The second tank is equipped with a
relief valve set just above the tank's nominal work-
ing pressure of 300 kPa. Students depressurize both
tanks and dismantle the valve to display the disc,
nozzle, and blowdown rings. After reassembly, the
valve is recalibrated on a test stand equipped with a
precision pressure gauge. It is then adjusted to give
the correct "cold-set" pressure and is leak-tested be-
fore being returned to the main air tanks. To demon-
strate the valve operation, students open the ball
valve when the first tank if full of air (600 kPa) and
the second is approximately half-full. They note the
pressure when the valve first lifts (with a loud exha-
lation of air) and when it abruptly shuts again. After

this they repeat the demonstration with a bursting
disc and housing replacing the relief valve. Ear pro-
tection is used for this exercise, and a wire cage
prevents any fragments from escaping. Students
write short notes comparing the operation of the two
types of safety equipment and describe the terms
"accumulation" and "blowdown" as applied to the
relief valve.
Shell-and-Tube Condenser *
Two essentially similar shell-and-tube exchang-
ers are equipped with condensing water flow meters
and gauges for the inlet and outlet temperatures.
Steam condenses on the shell side.
Students remove the head-pieces from the ex-
changers, exposing the internal view of the tube
bundles. They note the construction material, the
flow paths, and the tube diameters and lengths.
They also make notes on the quality of the gasket
and observe the correct order of tightening the many
nuts on the circular head-piece. After reassembly,
the rig is tested for leaks, and if all is well steam is
put to the shell. At steady state, heat fluxes (as de-
termined by flowrate and temperature rise of cooling
water) are compared to the rate of condensate pro-
duction. This heat balance shows the high efficiency
of heat transfer from steam to water. Students then
turn the cooling water flowrate down to a trickle and
show that after a time the cooling water can be
boiled, emerging as steam itself.
Plate Heat Exchanger *
The rig consists of a circuit in which a viscous
process fluid, molasses, is pumped from a tank via
a steam-heated plate exchanger, then to a water-
cooled plate exchanger, and finally back to the tank.
A progressive-cavity pump, fitted with a variable-
frequency speed controller, allows stepless and
consistent flowrate changes. Students dismantle
the plates of the steam-heated exchanger, observ-
ing the alternating pattern of gaskets which direct
the flows of the respective fluids. After drawing
the arrangement, they reassemble the exchanger,
observing the correct sequence for tightening bolts-
a key principle in the correct operation of this type
of exchanger.
After reassembly, with cooling water flowing and
molasses being pumped, steam is applied to the ex-
changer. If any leaks are evident, a time-consuming
dismantling job must be done. Students finally de-
termine the total duty of the exchangers, noting that
the steam-heated exchanger has a far greater heat
transfer coefficient than the water-cooled one.

Chemical Engineering Education

Steam Trap *
The process consists of a simple coil of copper
pipe through which steam is passed to heat water in
a 100 L tank. The condensate is passed to either of a
pair of alternative steam traps: 1) a mechanical
float-type trap, or 2) a thermostatic-type trap. Stu-
dents first dismantle both traps and sketch the in-
ternal workings, noting the mode of operation and
the materials of construction as well as removing
and cleaning the strainer. After reassembly, they
apply steam to the process and observe the rate of
temperature rise, comparing the two types of steam
traps for efficiency.
Heat balances are done by measuring the rate of
condensate production compared to the rate of tem-
perature-rise to show: 1) the steam is not 100% dry-
saturated, and 2) that some of the heat is lost to the
environment. Students are asked to describe the
best steam trap for a range of particular purposes.
Plate and Frame Filter *

A full-sized industrial filter is set up to separate
a slurry of PVC resin in water. Students dismantle
the filter rig and observe and sketch the internal
arrangement of plates, frames, and filter cloths. They
calculate the volumetric capacity of the unit for fil-
ter cake, and then prepare the process feed slurry in
a tank with the appropriate amount of solids. Dur-
ing the filtration run, they measure the pressure
drop across the filter and compare it to the rate of
filtrate production. The students are asked to de-
scribe the significance of their data and to answer
questions about constant-rate and constant-pressure
batch filtration.

DISCUSSION

After just one semester, the laboratory has
achieved a number of important educational goals.
Because novel concepts are treated from the outset
of the course, they make a strong impression on the
students, leaving (we hope) an enduring notion of
desirable engineering practice.
Students are exposed to a genuine chemical plant
that performs an identifiable chemical process. Sup-
pressing their initial anxiety, they applied them-
selves enthusiastically to the new tasks, learning
the names of plant items, reading instruments, open-
ing and closing valves, and recognizing the correct
sequence of events for any operation. They get an
idea of the scale of a chemical plant and the differ-

The plate-and-frame filter press stands in as a
workl)ench/desk prior I a run.

ence between it and the bench-type laboratory work
that they encountered in high school.
We emphasized industrial safety procedures.
Before the work began, we presented a detailed talk
on safety in the workplace, focusing on the hazards
in a "hostile and unfamiliar" environment. We in-
sisted on appropriate dress for the laboratory, con-
sisting of an approved hard-hat, a long-sleeved boiler
suit, steel-capped safety shoes, and safety glasses
when required. We also employed a "permit-to-work"
system prior to stripping a rig, and we constantly re-
inforced proper respect for sources of energy (steam,
electricity, and compressed air). At least 25% of the
assessment was on safety awareness and practice.
Accidents as a result of poor practices would not only
give the individual pain, but would also cause marks
to be lost.
We developed logical and systematic investiga-
tion methods. In writing up laboratory notebooks,
students were trained to describe a rig and its per-
formance in unambiguous language. This exercised
their ability to learn the names and functions of un-
familiar tools and equipment, to use precise techni-
cal descriptions, to become familiar with engineer-
ing units, and to state clearly what they observed.
Students quickly became aware of the difference be-
tween an engineering report and that of a more
traditional science laboratory.
A generous level of supervision helped greatly in
running the sessions successfully. Students appreci-
ated the immediate availability of demonstrators,
and they responded by spending extra preparation
time in drawing flowsheets and becoming familiar
with the rigs. This allowed the prepared students to
be unconcerned about the time remaining and to

Winter 1990

concentrate on understanding the equipment and
the data.

Information transfer was handled by illuminated
display boards and printed notes. Background read-
ing relevant to each rig was noted permanently on
display boards outside of the laboratory. These dis-
play boards were a constant reminder that informa-
tion was always near at hand. To help guide stu-
dents through the manual activities, they were given
laboratory notes describing the experimental proce-
dure in precise detail. This approach was greatly ap-
preciated since most of the students had never used
tools seriously before. Permanent copies of the notes
for each rig were kept in the laboratory, and the stu-
dents could purchase a set if they so desired.
The department at Sydney now has a high pro-
portion of women students (30-40%), and when they
first arrive many have the impression that they are
disadvantaged with respect to men, whom they see
as inherently more capable with machinery. How-
ever, it has been our experience that there is no ob-
servable difference between the performances of men
and women when it comes to their facility with tools
and equipment. All of the students enter the second
semester with a much greater feeling of equality and
confidence.
A novel aspect of the course was building the rigs
themselves. Seeing the importance of the venture,
our friends in industry agreed to construct rigs ac-
cording to our specifications, but first we had to
convince them that this would not be financially un-
attractive. Economies were realized by using "re-
tired" equipment when possible, and by using ap-
prentice labor augmented by some supervision and
design by senior students. Donations of this type
were welcomed as tax deductions. We approached
all of our industrial donors simultaneously, which
allowed us to commission the laboratory in just over
one year from its conception.
We operated the laboratory in much the same
way in 1990, but we commissioned two new experi-
ments: 1) investigation of the performance of a simple
level control rig using a pneumatic controller, and 2)
investigation of the behaviour of an air compressor
under variations in supply and delivery pressures.
Both experiments involved a dismantling task fol-
lowed by reassembly and operation

CONCLUSIONS

Over the last decade there has been a growing

realization that engineering education has under-
gone an expansion of theoretical exercises at the ex-
pense of practical experience. Australian and Brit-
ish reports on the engineering profession together
with the Institution of Chemical Engineering degree
requirements,1" reinforced the realization that there
is a need for a field of training, called Engineering
Applications, which would seek to correct the imbal-
ance. We feel that this new laboratory makes signifi-
cant progress in that respect.
In setting up the laboratory, one of our major
intentions was to give students a close look at the
profession that they have chosen. As a result, one or
two students may make an early decision that chemi-
cal engineering is not for them, something that they
might have taken several years to realize otherwise.
The students who do remain now identify them-
selves as engineers with a developing set of theoreti-
cal and practical skills that distinguish them as
worthy professionals. They also volunteer that while
the experience was hard work, it was also fun.

ACKNOWLEDGEMENTS

The laboratory has been made possible by the
generous support of: Alfa-Laval, Australian Paper
Manufacturers, BHP Steel International Group,
Caltex Oil (Australia), Catoleum, The Commonwealth
Industrial Gases, CRA Advanced Technical Devel-
opment, Crosby Valve and Engineering, CSR, Dow
Corning Australia, ICI Australia Operations, Phos-
phate Technology, and Shell Refining (Australia).
Much of the inspiration and development of the
laboratory was due to a six-month visit to Sydney by
R.J. Aird, with the support of the British Council.
Our thanks also to the laboratory and technical
staff of the department for their special effort in
commissioning the laboratory. It has been named
the "Chemical Engineering Foundation Laboratory
for Engineering Applications" in recognition of the
strong support of the Foundation in planning and
liaison with our industrial donors, most of whom are
members of the Foundation.

REFERENCES
1. Finniston, M., Chairman, "Engineering Our Future," Inquiry
into the Engineering Profession, p 94, Her Majesty's Station-
ery Office, London (1980)
2. Institution of Chemical Engineers, Accreditation of Degree
Courses, App. 2, London (1985)
3. Williams, B., Chairman, Review of the Discipline of Engineer-
ing, 1, 48, AGPS, Canberra (1988) 1

Chemical Engineering Education

GAS SEPARATION EXPERIMENT
Continued from page 15.
An evaluation survey was conducted of all the
students participating in this experiment during the
fall quarter of 1989. They were asked to evaluate
different aspects of the modified membrane experi-
ment, such as clarity of the handout on equation
derivation, appropriateness of objectives, and ability
to analyze data using the computer programs that
were provided. The responses indicated that they
liked the experiment, and there was a general feel-
ing of satisfaction with their laboratory experience.
As instructors, we were pleased with the outcome of
our efforts to enhance the technical aspects of this
experiment.

ACKNOWLEDGEMENT
This work was sponsored by a Teaching Assis-
tant Instructional Improvement Grant funded by
the UCSB Office of Instructional Development.

NOMENCLATURE
A = unit membrane interfacial area (m2)
a = interfacial membrane area per unit volume of
separator (m 1)
d = inner diameter of separator column (m)
G = molar flow rate of gas in the non-permeate stream
(gmol!s)
G' = permeate stream molar flow rate (gmol/s)
h = separator column height (m)
K = dimensionless transport parameter defined in
Eq. (12)
K' = dimensionless transport parameter defined in
Eq. (20)
m = K' correlation coefficient
P = absolute pressure in the non-permeate stream (kPa)
p = absolute pressure in the permeate stream (kPa)
Q = permeabilities (kgmol/m.kPa.s)
T = temperature ( C)
V = unit volume of separator column (m3)
x = mole fraction of 02 in the non-permeate stream
y = mole fraction of 02 in the permeate stream
z = distance along length of separator column (m)
z = z/h, dimensionless column length
Greek Symbols
a = separation factor
a' = ideal separation factor, QA/QB
8 = membrane thickness (m)

POSITIONS AVAILABLE
Use CEE's reasonable rates to advertise.
Minimum rate, 1/8 page $80;
each additional column inch $25.

VILLANOVA UNIVERSITY
DEPARTMENT OF CHEMICAL ENGINEERING
Two (2) full-time, tenure track positions at the Assistant or
Associate Professor level available commencing September
1991. Candidates must have an earned Ph.D. in Chemical
Engineering by August 1991. Previous industrial and/or teach-
ing experience will be desirable for all candidates, and will be
required for consideration at the Associate Professor level. All
candidates must demonstrate oral and written communica-
tion skills, and interest in undergraduate and graduate teach-
ing. Preference will be given to those having experience and
teaching interests in one or more of the following areas: proc-
ess equipment design; chemical reaction/reactor engineering;
chemical equilibria/solution thermodynamics. Each success-
ful candidate will be expected to develop an active research
program compatible with the faculty member's other obliga-
tions to the University. Research areas are open, but the
Search Committee will evaluate the appropriateness of each
applicant's research interests within the context of University
resources. Applicants should submit a discussion of teaching
and research interests, a curriculum vita, and the names of
three references, postmarked before March 1, 1991, to: Profes-
sor Vito Punzi, Search Committee Chairman, Department of
Chemical Engineering, Villanova University, Villanova, PA
19085. The University is a fully-accredited institution with a
strong emphasis on teaching. The Chemical Engineering De-
partment offers programs leading to the B.Ch.E. and M.Ch.E.
degrees. Villanova University is an Augustinian-related Ro-
man Catholic institution and is an AA/EO Employer. Women
and minorities are especially encouraged to apply.

Subscripts

A = oxygen
B = nitrogen
i = inlet
n = column number
o = exit

REFERENCES
1. Chern, R.T., W.J. Koros, and P.S. Fedkiw, "Simulation of a
Hollow-Fiber Gas Separator: The Effects of Process and
Design Variables," Ind. Eng. Chem. Process Des. Dev., 24,
1015(1985)
2. Pan, C.Y., and H.W. Habgood, "Gas Separation by Permea-
tion. Part II: Effect of Permeate Pressure Drop and Choice
of Permeate Pressure," Can. J. Chem. Eng., 56, 210 (1978)
3. Walawender, W.P., and S. A. Stern, "Analysis of Membrane
Separation Parameters. II: Counter-Current and Cocurrent
Flow in a Single Permeation Stage," Sep. Sci., 7(5), 553
(1972)
4. Pan, C.Y., and H.W. Habgood, "An Analysis of the Single-
Stage Gaseous Permeation Process," Ind. Chem. Funds.,
13(4), 323 (1974)
5. Hwang, S.T., and K. Kammermeyer, Membranes in Separa-
tions, John Wiley & Sons, New York (1975) 0

Winter 1990

Random Thoughts...

ENGINEERING EDUCATION

VERSES

RICHARD M. FIELDER
North Carolina State University
Raleigh, NC 27695

LIMIT CYCLE

It must be a canon of natural law
or a reflex reaction like jerking of knees,
That whenever a company's profits are down
They proclaim an across-the-board hiring freeze.

Seniors looking for work find the door has been
barred.
The red carpets that last year were there in full
view
Have been rolled up and mothballed and stored
out of sight,
And the letters all read, "Don't call us, we'll
call you. "'

For the next several years on each campus you'll
find
Would-be engineers singing their frustrated blues,
And the word gets to seniors in high school that if
You select engineering you'll pay some high dues.

Now these kids are no dummies, they soon get
the drift
And to business and law school they roll in a wave.
Engineering enrollments go down like a stone
And the deans struggle vainly their budgets to save.

It's a national crisis! Blue ribbon commissions
Spend years and big bucks in a terrible fright,
And proclaim that the old engineering profession
Is sick unto dying, with no hope in sight.

Then of course there occurs a dramatic reversal
That cuts short the agonized moaning and tears,
For it seems that the companies had a good quarter
And all of a sudden they need engineers!

The call soon goes out, all the freezes are lifted,
Red carpets are hauled out and vacuumed with care,
But recruiters soon find to their shock and surprise
That the students they're trying to get aren't there.

Now the circus begins, it's the law of the jungle,
If your pulse can be measured they want you
right then.
Up go salaries, perks, and enrollments once more
Till the whole silly cycle starts over again.

So what can we learn from this sad sordid story?
The moral in just a few words I'll disclose:
If our industry wants to stay healthy it might think
Of lifting its gaze past the end of its nose.

WANNA MAKE ABET?

Who cares if their coming inspection
Puts our jobs and prestige on the line?
Since our courses are called engineering,
What we teach has to count as design.

Chemical Engineering Education

OH NO, IT'S THE E WORD!

Thermo mavens rant about it.
Gibbs and Sandler and Van Ness
Give you formulas and prose
About this thing that's known as S.
It increases with disorder,
It's a property of state,
It is zero for some crystals,
It's the universe's fate.
It can be dq, dU
and sometimes dH over T,
It's a measure of extent of
Irreversibility.
It accounts for work that's lost
In engines, blenders, pipes and flues,

Due to friction or to mixing.
It's a game we always lose.

Its the reason we succumb
To Deaths inexorable crunch,
It equals k In omega,
Itis why others no free lunch.
You can lookitup intables,
Findit ona Molliergraph,
Butcanthey telyou whatthehel itrealyis?
Dontmakemelaugh!
Ther mustbsom waytoexplan
intrmsweallculd easlyndrstnd,
Ifntprhps weshuldsmplysayitsoneof
Thsthngsthtcnttbputntrmsthtlrtjaaqstjwxrmztbl...

TEACHER TRAINING WORKSHOP (SHORT FORM)1

If you're anxious for to shine in the academic line
As a man2 of wisdom rare,
You must cultivate the Dean and bring in lots
of green
And publish everywhere.
You must go to learned meetings and exchange
flamboyant greetings
With the heroes in your field.
Your rise will be dramatic and your peer
reviews ecstatic
And your reputation sealed.
And everyone will say,
As the plaudits come your way,
If this young man got a PYI
Not to mention an ERC,
Why, what a very paragon of scholarship
This bright young man must be.
Be nimble on your feet too when a VIP you meet
who
In your field enjoys respect.
Though you think that in his work he is a
thoroughgoing turkey
You must smile and genuflect.
For some day he'll have a vote on a proposal that
you float,
As toward major grants you steer,
And you wouldn't want him to cut you down in

his review,
Like you did to him last year.
And everyone will say,
As you wend your upward way,
If this young man has friends in court
At every funding agency,
Why, what a stunning academic superstar
This superstar must be.

As your star goes on ascendin' you must guard
against a tendency
To ease up on the pace.
Work at night and on the weekend
Lest in time you face a bleak end
In the crucial tenure chase.
Teach your class, serve on committees, go give
talks in far-off cities,
Put equipment out on bids,
And maybe every week or two a reasonable thing
to do
Is visit with your wife and kids.
And everyone will say,
As you turn prematurely gray,
If this young man works a hundred-hour week
Which is far too much for me,
Why, we had better find him a Distinguished Chair
So he doesn't jump to M.I.T.

Winter 1990

1 For those who lack the Patience for the long version. (Apologies to W.S. Gilbert.)
2 Or woman.

classroom

A "USER-FRIENDLY" PROGRAM

FOR VAPOR-LIQUID EQUILIBRIUM

FRANCISCO A. DA SILVA, Luis A. BAEZ,
AND ERICH A. MULLER
Universidad Simon Bolivar
Caracas 1086-A Venezuela

T hermodynamics, and particularly vapor-liquid
phase equilibrium, has been (and most likely
will be for many years to come) a "difficult" subject
for many students and researchers who are not ac-
customed to the subtle details of the discipline. Until
recently, solution of the more complicated phase
equilibria was restricted to research groups with
appropriate computing facilities. But the computer
age has now made the evaluation of these problems
possible for students who have access to personal
computers. The teaching of equilibrium thermody-
namics has therefore evolved from the statement of
models, the study of graphs, and the solution of only
very ideal situations to the solving of relatively com-
plex models by the students themselves.
In a single course it is not possible to ask the
student to master (1) the thermodynamic theory, (2)
the details concerning the programming of many in-
tricate models, and (3) the direct application of the
different models (i.e., recognizing when certain mod-
els are or are not useful).
The first step can normally be accomplished by
classical teaching methods, while the second is usu-
ally not an explicit curricular objective and may be
left to homework or similar assignments. As a re-
sult, little time is left for the last objective (which is
probably the most typical problem that an active
chemical engineer will encounter). Thus, students
and instructors are asked to sacrifice, because of
time considerations, one of the most practical as-
pects of the science.
The logical course of action is to provide the stu-
dents with a prepared program, let them learn by
using it, and save them the annoying experience of
Copyright ChE I
24

having to deal with the actual programming of the
models.

There are many programs available for the cal-
culation of the various aspects of equilibrium th-
ermodynamics. The programs we found usually fell
into one or more of the following categories:
SSimple programs, written by students or by research
groups, which illustrate or solve a very narrow aspect or
situation and usually have limited scope (e.g., a program
that can calculate UNIQUAC constants from VLE points).
These programs are frequently hostile to the user, do not
detect errors in input data, are slow-running, and/or are
not accompanied by a (much needed) on-line help file or
manual. Rarely do these programs include a graphic
interface since the usual programming language is
FORTRAN.
At the other end of the spectrum are the high-tech

Francisco A. Da Silva F. graduated Summa Cum
Laude in chemical engineering from Simon Bolivar
University in 1990 and is currently enrolled in the
graduate program at the same university. His expert
tise in the area of computer science is well known on
campus, especially due to the creation of several
utility programs and games.

Luis A. Baez L. graduated Cum Laude in chemical
engineering from Simon Bolivar University in 1989
and is currently a graduate student at Cornell Uni-
versity. An active sportsman, he enjoys high-moun-
tain exploration. He has also visited the Venezuelan
and Amazonian jungles on several occasions.

Erich A. Miller graduated Cum Laude in chemical
engineering from Simon Bolivar University (1986),
MSc Honor Mention in chemical engineering (1987),
and is an assistant professor in the Department of
Thermodynamics and Transport Phenomena at the
same university. He has written several technical
I papers in the area of equilibnum thermodynamics,
Which is his main research interest.
Dwismn, ASEE 1991

Chemical Engineering Education

Our basic objective was to produce a package that would require minimal computer skills to operate and
thus eliminate the use of manuals and other types of outside help. A high priority was the development
of a program that the user could master without training in computer-handling or programming.

simulation programs, recently made available for state-
of-the-art microcomputers. Even though they are very
powerful tools for the process engineer, they can be
frustrating for a newcomer to use and usually require
long hours of studying technical manuals. Moreover,
these programs are cost-prohibitive for most students
and faculty members.
SA third category is teaching programs, among which are
those developed at Cornell University "that present 3D
binary VLE diagrams and their 2D projections, and those
developed at Iowa State University that represent 3D
PvT surfaces for pure components and binary
mixtures.' The programs from Cornell run on specially
pre-established components or systems, and all stress
the visualization of the geometry (PvT, PTxy surfaces).

In the area of equilibrium thermodynamics there
is need for a program which can accomplish phase-
equilibrium calculations using various models, which
can permit the visualization of these equilibria by
phase diagrams, which has an easy-to-use and com-
plete database, and which is accessible to students
(a user-friendly teaching tool).

Our research group has developed such a com-
puter-aided package, called "Ekilib," which promises
to fulfill all of the above requirements.

DETAILS OF THE PROGRAM

Our basic objective was to produce a package
that would require minimal computer skills to
operate and thus eliminate the use of manuals
and other types of outside help. A high priority
was the development of a program that the user
could master without training in computer-handling
or programming.
A Macintosh computer was selected since it is
widely used and has excellent graphic capabilities.
The program was developed to run on a Macintosh
512K and will run on any similar or superior ma-
chine without the need of special hardware. The
complete package will fit on a single 800K diskette.
The main program recognizes color monitors and
new processors on the most recent Macintosh ma-
chines and thus exploits the full possibilities and
speeds of these new systems. All graphs, tables, and
other output can be printed on a standard dot-ma-
trix printer and can be seen in color if the black
ribbon is replaced by a color ribbon. The program is
not available in IBM-compatible versions.

The program has three distinct parts. The first is
a pure component data base designed to hold data
for up to 2000 components. The following data can
be stored for each substance: critical properties (Tc,
Pc, Zc), acentric factor, solubility parameter, normal
boiling temperature, molecular weight, empirical
formula, dipole moment, liquid molar volume, an
Antoine-type vapor pressure expression, and an ideal
gas constant pressure specific heat equation, along
with other constants. A data base with over 600
components is supplied with the program which can
be consulted substance-by-substance, then printed
out, or changed, or modified by the user. Figure 1
shows a typical window in which the properties of a
substance may be displayed. The data may be printed
out or transferred to other programs (e.g., a word
processor). Similar windows permit the addition or
modification of data.

The second part of the program is the VLE calcu-
lation section. Single compounds and binary and
multicomponent mixtures are supported. A wide
variety of models for the liquid and vapor phases are
presented. The models were chosen for this program
(1) on the basis of popularity and common use (e.g.,
Peng-Robinson, Redlich-Kwong-Soave, Wilson); (2)
on the basis of teaching fundamentals and compari-
son (e.g., ideal gas, ideal solution, Margules); and (3)
to incorporate some state-of-the-art cubic equations
of state (EOS) (Van der Waals-Adachi-Watson, et

SI-il Data Bank
Consult
Name: | ...1 ta Bank...
Name: t-1
ACETONE e.t
Molecular Formula: C3H60
M.W Pc (barl Tc (KI Zc w m n (lOebyes]
58.08 47.0 508.1 0.232 0.3040 0.79510 0.22050 2.9000
Normal boiling point temperature: Tb- 329.2 11K
Hildebrand's solubility parameter: a- 9.566 (cal/cm3)'1/2
Specific molar volume of the liquid : UI- 73.5000 (cm3/gmoll
Ideal gas specific heat (Cp) in [kJ/kmoll, T in IKl:
Cpr 6 30o1000+0 +2.6050Om -1T -1 253000e-4T2 2 038000e-STr3 +0 m00000e+OTt4
Uapor pressure by Rntoine-type equation. T in iK]:
In(Pv/Pc) 6.244412 2975.953/(T + -34.5226)

Figure 1. Window that permits the user to display the
properties of a pure substance. Similar windows allow the
user to modify data or to expand the data bank.

Winter 1990

al.,'5' Polar RKS,161 Schmidt-Wenzel,'71 PT-USB.1' The
program stresses cubic equations since they can pro-
vide good predictions without empirical constants.

It should be pointed out that some of the regu-
larly-used models have been omitted (e.g., Virial
equation, UNIQUAC, Unifac, etc.) for two basic rea-
sons: (1) space limitations, and (2) so as not to con-
fuse a "newcomer" in thermodynamics with an over-
whelming array of choices (which would undermine
the program's primary objective of user-friendliness).

One of the most innovative and practical aspects
of the algorithms used is that the user can select
independent models for the liquid and vapor
phases; e.g., one might specify Peng-Robinson for
the liquid phase and ideal gas for the vapor phase; or
Wilson could be chosen as the liquid model, and
Patel-Teja for the vapor phase. This permits
the user to discover which combination provides
better convergence or approximation to a given solu-
tion. Figure 2 shows the window that permits selec-
tion of the model to be used. Models that cannot be
selected (insufficient data, inapplicable, etc.) shift to
a gray tone.

Once the model is selected, changes can be made
in the reference points (in the calculation of enthal-
pies, etc.), in the minimum acceptable tolerances,
and in the maximum number of iterations and start-
ing values. If this is not done, built-in default values
are used. Depending upon the number of compo-
nents in the mixture, different options appear in the
pull-down menus. For single-component systems,
thermodynamic properties can be computed (en-
thalpy, entropy, availability, vaporized fraction) and
a P-T diagram generated. For a multicomponent
system, in addition to the aforementioned, equilib-
rium calculations (bubble T or P, dew P or T, isother-
mal or adiabatic flashes) can be performed. In addi-
tion to these, binary systems offer a possibility for
construction of binary diagrams: x-y, H-x-y, P-x, and
T-x. Figures 3, 4, and 5 present the graphics screen
output for several such constructions.

All the results and graphs can be printed on
paper, transferred to other programs, or saved to the
current disk in the same fashion that a word proces-
sor saves a file. The files that are generated contain
all the information about the substances involved
and can be transferred between users, even if they
don't have the same data base.

For those models that require binary interaction
parameters (Aij in the liquid models such as Wilson,

4 File Edit Calculations Binaries
Equilibrium Model Selection
Liquid Phase: Vapor Phase:
0 Ideal Solution 0 Ideal Gas
0 UdW-Rdachi-Watson ChUdW-Rdachi-Watson
0 Redlich-Kwong-Soaue 0 Redllch-Kwong-Soave
O R-K-S (Polar) 0 R-K-S (Polar)
0 Peng-Robinson 0 Peng-Robinson
0 Schmldt-Wenzel 0 Schmidt-Wenzel
0 Patel-TeJa 0 Patel-Teja
0 Patel-TeJa-USB 0 Patel-TeJa-USB
OSatchad Hildebrand .....
O IlJith Kij
@ Margules E]I__________
OUan Lear
O Wilson ,
O lWith I. huo snpillPr Q 2 i J
C IUlt th Polntinq

Liqui [ Cancel eadl i

Figure 2. Window that allows the selection of models for
the vapor and liquid phases. Items in gray cannot be
chosen due to insufficient data in the data bank
or to inapplicability.

6 File lditi Select Calculations Binaries
Ethanol-Water

0.8

os
06

0.4

0.2
0.0 Top :423 1?5 IK]
0.0 0.2 0 4 0.60 0 1 0

G ds: Redl ich-Knong-Soave
Liquid: Uilson

Figure 3. Composition (x-y) diagram for the azeotropic
mixture ethanol-water. Wilson parameters were
calculated using the experimental data stored
in the binary system data bank.

SFile Ildi Select Calculations Binaries
mm HeHane-Octane

Figure 4. Enthalpy-concentration diagram for the mixture
hexane-octane generated from the Peng-Robinson EOS.
The reference state may be modified by the user at will.

Chemical Engineering Education

- W000 T. otUr 300.00 [IK

S72000

48DDO

240000
12000
0.2 0.4 0.0 0.8 1.
-12000

File ldit Select Calculations Binaries
Methanol-Ollsopropyl ether

S344.0
SPressure. 0.970 [barl
342.3
340.5
338.9
337.2

333 8
332 1
330.4
328.7
327.0
0.0 0.2 0 4 0.6 0.8 I 0
X

Figure 5. Temperature-composition diagram for the
mixture methanol-diisopropyl ether calculated
using Wilson-RKS and experimental data

TABLE 1
Technical information and capabilities of the program "Ekilib"

DATA BANKS: More than 600 pure components and over
100 binary systems including hydrocarbons, polar and
associating compounds. Electrolytes are not covered.
The user may expand or modify the data using built-in
text editors, import data files, or define his own pseudo
components.
VAPOR-PHASE MODELS: Ideal, RKS, PR, van der Waals-
Adachi-Watson, polar RKS, Schmidt-Wenzel, Patel-Teja,
PTUSB, all with optional use of kij.
LIQUID-PHASE MODELS: The same models as the vapor
phase, plus Scatchard-Hildebrand, Margules (only
binary), Van Laar, and Wilson, with options of Poynting
correction or the use of the Chao-Seader correlation.
CALCULATIONS: Thermodynamic properties: vaporized
fraction, enthalpy, entropy, availability and fugacity
coefficients. Equilibrium problems: dew P or T, bubble
P or T, isothermal or adiabatic flash. Binary interaction
parameters: for liquid models (Aij) and cubic EOS (kij).
GRAPHICS: P-T, x-y, P-x-y-, T-x-y, and H-x-y diagrams for
binary systems, and P-T for pure substances and
multicomponent mixtures. They may be printed on
paper in large size (8x10 in) and in color if the
appropriate ribbon is used.
UNIT SYSTEM: Standard default are SI units. The user
may select from a variety of units for pressure,
temperature, and energy.
COMPUTER HARDWARE: Apple Macintosh 512K
computer or superior. Supports newest processors and
color monitors. Outputs to standard Imagewriter II.
Entire package fits on a single 800K diskette.
HELP: A simple manual is provided and some basic
examples are given. The program is totally user-friendly
and warns the user of common mistakes.

Margules, and van Laar; kij for cubic EOS), the
program provides a third part in which the calcula-
tion of these constants from binary experimental
data can be performed. Aij is found by best-fit re-
gression and kij from the Paunovic, et al.,191 criterion
using experimental VLE data in the form of iso-
therms or isobars. This data can be supplied directly
by the user through a built-in editor, imported from
other machines (via ASCII files), or taken from over
100 binary-system data files provided with the pro-
gram. Thus, a user can calculate binary interaction
parameters from experimental data and use the re-
sults in other calculations.

All algorithms used in the program have been
tested extensively in other simpler programs, and
they have proved to be the best in overall speed and
convergence. Rarely does any flash calculation take
more than one minute, and some require only sec-
onds. Calculation of graphs and binary interaction
parameters are also very time-efficient. Greater de-
tails of these procedures and algorithms are given
elsewhere.'" Table 1 shows a brief summary of the
program and its capabilities.

USER INTERACTION

The program has an extensive user-interaction
interface. It is implemented by pull-down menus
which present different options according to the en-
vironment or situation. The most attractive aspect
of the program is probably the fact that absolutely
no knowledge of computing is needed, and no strict
format or rules must be followed. Most common er-
rors are not allowed to occur (e.g., in specifying a
feed for a flash, mole fractions whose sum exceeds
unity simply cannot be typed; the user is warned
with an audible signal, and a message appears on
the screen indicating the error).

At a teaching level, this program is a handy tool.
It is a well-known fact that one learns by doing.
Once the theory has been studied and a few
examples have been worked out by hand it seems
unnecessary to make students waste time and
effort in programming multiple algorithms. It
proves more efficient to let them study by them-
selves using this program.

The program is being used experimentally in
several courses at our university. It is a fact that
after a short time spent in dealing with the program
(typically a few hours) students discover facts that
would merely be empty words when given in class.
Continued on page 32.

Winter 1990

classroom

TEACHING EFFECTIVE

ORAL PRESENTATIONS AS PART OF THE

SENIOR DESIGN COURSE

E. L. HANZEVACK, R. A. McKEAN1
University of South Carolina
Columbia, SC 29208

Heavy emphasis on the technical aspects of their
education encourages many engineering students to
believe that once they graduate, their jobs will be
principally involved with collecting and analyzing
information. But, as practicing engineers know, tech-
nical expertise means little if an engineer cannot
effectively communicate it. Engineers must be effec-
tive communicators in order to discuss their work,
present their findings, and propose a course of ac-
tion. Essentially, if engineers cannot inform others
of what they have done, they might as well not have
done it.
Engineering faculty recognize the engineer's need
for effective communication skills. In the senior de-
sign and economics course, students traditionally
are asked to prepare a major paper and to present
their findings orally. However, due to time constraints
(or perhaps to the instructors' lack of prepared ma-

Emil L. Hanzevack has been an associate professor
in chemical engineering at the University of South
Carolina since 1983. He teaches the senior design
course and process control, and he does research in
computer applications to chemical engineering. He
was previously with Exxon Research and Engineer-
ing where he was responsible for generating and
administering R&D programs, which required writing
and editing many reports and making technical pres-
entations to many levels of audience.

Rob Adams McKean is president of Chart Communi-
cations, a consulting firm that specializes in helping
people speak and write more clearly and persua-
sively. Chart Communications provides on-site train-
ing for technical and nontechnical staff, in group set-
tings or individually. As a consultant to both industry
and government, he has led over three hundred train-
ing seminars.

'President, Chart Communications, 174 Hillside Street, Boston,
MA 02120

triall, students are often told to give a presentation
but are given little or no guidance on how to do so
effectively. Comments received from students indi-
cate that, in addition to natural nervousness when
speaking publicly, they are not even sure what crite-
ria define a good presentation.
But even though engineering faculty recognize
the need for training in communications, the argu-
ment is frequently put forth that there is already too
much technical material to be covered in the curricu-
lum. The pervasive feeling is that these "softer" ar-
eas are better left to the English department (for
writing courses) and to the communications depart-
ment (for public speaking courses).
While freshman composition is normally manda-
tory for all students, speech class is usually elective
for engineering students, and we find that students
readily avoid taking it. Furthermore, the kind of oral
presentation a practicing engineer is likely to be
called upon to make is somewhat different from a
typical public speech. We certainly encourage stu-
dents to make room in their schedules for speech
class, but we just as firmly believe the engineering
curriculum needs to provide opportunity for students
to learn, practice, and master effective communica-
tion skills.
Indeed, we suggest that it is possible within the
context of the senior design and economics course to
include, with little additional time, a brief lecture on
oral presentations accompanied by written guide-
lines. The discussion that follows sets forth sugges-
tions on how to organize a presentation, deliver it,
and prepare graphic aids to accompany it. In addi-
tion, we have condensed the discussion and pre-
sented the same suggestions in Table 1 as second-
person lecture notes to facilitate faculty who might
wish to incorporate them into their own lecture notes
or to copy them and use them directly as a handout
in their class.
Copyright ChE Division, ASEE 1991

Chemical Engineering Education

Equally important, we provide rating sheets to
students after their presentations, with a quantita-
tive score in each of several categories, as well as
qualitative notes and suggestions. Students seem to
appreciate these guidelines and the feedback. A
sample of the rating sheet is shown in Table 2.

SUGGESTIONS FOR ORGANIZING AND
DELIVERING A SUCCESSFUL PRESENTATION

Organizing a Talk

A successful presentation is almost always
a carefully-organized presentation. Organization is
the process of designing an intelligent, useful, co-
herent, engaging program, whether it is ten minutes
long or five days long. Out students are advised to
think in terms of what they can reasonable do in
their allotted time and adapt their material to suit
the assignment.

A common mistake that inexperienced present-
ers make is failure to control the time. Typically, the
poorest talks are not the ones that are too short, but
the ones that are too long. These are the talks that
seem to digress and wander at will, that seem dis-
connected and endless. The old axiom still holds: tell
us what you are going to tell us, tell us, tell us what
you have told us, and sit down.

We advise our students to ensure that their talks
answer three main questions:
What was done or is to be done?
How was it done or how will it be done?
What was or will be the significance?

An essential part of organizing a successful talk
is composing an outline. Since students will only be
able to cover the highlights of their written papers,
we advise them to think in terms of a summary. The
same factors, conclusions, and recommendations that
they use in a summary usually work in a presenta-
tion or briefing. Students are urged to start off
with the big picture (overall concepts) and to then
discuss the points most critical to understanding
their message.
A presentation needs to stand alone. That is, it
needs to be self-contained and make sense in and of
itself. We urge students not to reference their report
(not to report on their report) any more than neces-
sary. We orge them to resist the temptation to turn
us into secondhand listeners. For example, we coach
them to avoid sentences that begin, "In my report,
I...," or, "As I mention in chapter two...." Students

are urged to use "you" in their presentations and to
talk to those who are sitting before them right then
and there.

Targeting and Involving the Audience

Presenters know more about their topic than
anyone else in the audience. What is obvious to them
might not be obvious to others. We advise students
that targeting their audience means tailoring their
remarks to their immediate listeners. Presenters
should think about an audience's technical back-
ground (gained through both formal means and ex-
perience), its familiarity with the subject, its atti-
tudes, and its informational needs.

An audience will be swiftly bored by a one-way
conversation. To have a two-way conversation, pre-
senters must personally connect with the audience.
They should try to involve the audience in their
performance. We coach students to look at, talk to,
and give the audience something to examine. State
conclusions and action items explicitly. Don't as-
sume that any aspects are obvious, and don't force
the audience to guess them.

Traditionally, audio-visual aids (charts, over-

TABLE 2
Presentation Feedback Form

Name ofPresenter
Title of Presentation
Date of Presentation
My Name

Content & Organization Noteworthy Can be
Improved
Core message
Clear objectives
Overall structure
Visible logic
Targeted at audience
Audience motivation
and involvement
Presentation
Confidence,
enthusiasm,
forcefulness,
conviction
Controlled pace
Natural finish
Voice quality (clear, calm,
and understandable)
Gestures (natural,
intuitive, relaxed)
Body/facial signals
Frequent eye contact
Visual Aids
Interesting
Relevant
Easy to read

Winter 1990

TABLE 1
Suggestions for a Successful Presentation

ORGANIZING YOUR TALK
a Plan your presentation carefully.
Set an agenda and stick to it. A common mistake inexperienced presenters
make is failure to control the time. Organization is the process of designing
an intelligent, useful, coherent, engaging program, whether it's ten min-
utes long or five days long. Think in terms of what you can reasonably do
in your allotted time, and adapt your material to suit your assignment.
Typically, the poorest talks are not the ones that are too short, but the ones
that are too long. These are the talks that seem to digress and wander at
will, that seem disconnected and endless. The old axiom still holds: Tell us
what you are going to tell us, tell us, then tell us what you have told us.
Then sit down.

a Ensure that your talk answers three main questions:
What was done or is to be done?
How was it done or how will it be done?
What was or will be the significance.

a Compose an outline.
Since you will only be able to cover the highlights of your written paper, it is
useful to think in terms of a summary. The same factors, conclusions, and
recommendations that you use in a summary usually work in a presenta-
tion or briefing. Start off with the big picture (overall concepts) and then
discuss the points most critical to understanding your message.
Remember, a presentation needs to stand alone. That is, it needs to be
self-contained and to make sense in and of itself. Don't reference your
report any more than necessary. Don't report on your report. Resist the
temptation to turn us into secondhand listeners. For example, avoid
sentences that begin, "In my report, I .. .," or, "As I mention in chapter two
.." Use "you" in your presentation-talk to us who are sitting before you
here and now.

Target and involve your audience.
You know more about your topic than anyone else in the audience. Re-
member, what is obvious to you might not be obvious to others.
Targeting your audience means tailoring your remarks to us. Think about
our technical backgrounds (gained both through formal means and through
experience), our familiarity with your subject, our attitudes, and our infor-
mation needs.
Likewise, we listeners are going to be swiftly bored by one-way conversa-
tions. To have a two-way conversation you must personally connect with
us. Involve us in your performance: look at us, talk to us, give us some-
thing to examine.
Traditionally, audio-visual aids (charts, overheads, slides) have been the
most common means of involving us in your talk and helping us take in
your message. Make sure your graphic aids are interesting, relevant, and
easy to read (see last section). Above all, however, don't expect graphics
to carry your talk and automatically turn it into a scintillating two-way con-
versation.

n Give extra thought to your opening and closing sentences.
Openings and closings are the most remembered portions of your presen-
tation.
The opening should clearly demonstrate your command of the topic and
the situation, your purpose, and your organizational plan.
The closing should be fitting and final. Never end with a shrug or with a
weak statement like, "This ends my talk." Think of a snappy closing
sentence so that there is no doubt in our minds that you have finished.

o State conclusions and action items explicitly. Don't assume these
aspects are obvious, and don't force us to guess them.

o Remember above all while you are organizing, your purpose is to
inform, to request, or to persuade-never just to impress.

PRESENTING YOUR TALK
Dress for the part of speaker or invited guest. Make sure you will be
comfortable in and confident of the clothes you have selected.

a You may use light notes or an outline, but don't read your talk (never bring
the full text of your talk with you to the lectern).

a Walk briskly to the lectern: stand quietly for a few seconds before begin-
ning, take a deep breath, make eye contact with us, then begin on a firm
clear note.

n Have your opening sentences committed to memory. Presenters nor-
mally find that once they get started their initial nervousness wears off and
they lose self-consciousness. By knowing your opening comments cold,
you can guard against drawing a blank or getting off to a weak start.

n Accept a certain degree of nervousness as an inevitable part of being in
front of people.
Rather than fighting the nervousness or despairing over it, use it as a
stimulant. Some nervousness means you're keyed up, excited, ready to
connect with five, ten, or a thousand people.
Remember, your audience is made up of real human beings, much like
yourself. We have come to see and to hear you. Lift your face to us and
speak clearly, decisively, feeling your voice carry to the far comers of the
room. Your voice is your greatest tool and your greatest ally. As you gain
confidence, it will deepen and steady and come from a quiet center of
power within you.

If your nervousness Is chronic and seriously threatens the success of your
talk, then you need to take specific steps to control it.
First, analyze your nervousness. Pinpoint what's producing it. Are you
inadequately prepared (rarely, overprepared)? Are you not in touch with
your central message? Are you unhappy with your personal appearance?
Once you have established the root cause of your nervousness, you can
begin to work on neutralizing it. You may find extra practice or deep-
breathing exercises help. It may be that all you need is more experience in
front of groups. Whatever the cause of your extreme nervousness, if you
sincerely want to overcome it, you almost assuredly can.

a Last, but not least: SMILE! Along with a confident voice, a sincere smile
calms and cheers us.

PREPARING VISUAL AIDS
a Plan on using graphics (overheads or slides) to accompany your talk. As a
rule of thumb, don't plan to stay on any particular graphic more than two to
three minutes. If you do, your audience will grow bored.

If you have finished discussing a graphic but don't have another to
immediately replace it, turn the overhead projector off. That will bring our
focus back to you-where it belongs.

a Never photocopy books or typewritten material. They are illegible to most
of the audience and give the impression you quickly threw the material
together.

n Plan on a maximum of 6-10 lines per overhead. An outline form with
bullets is best. Remember, a graphic should present a distilled form of
your comments (in the same way the oral presentation is a distillation of
your written report).

a Plan your graphics as a harmonious suite. Design a format that presents
your information attractively, and stick to that format throughout all your
graphics.

a Use numbers sparingly. We will remember two or three key numbers, but if
you give us too many we will end up remembering few, if any.

n Other props are optional (samples, models, etc.). They can be good if they
are interesting and relevant, but too many will distract from the talk.

Chemical Engineering Education

We recommend that students learn to accept a certain degree of nervousness
as an inevitable part of being in front of people. Rather than fight the nervousness or
despair over it, we suggest that they use it as a stimulant. Some nervousness means a presenter
is keyed up, excited, ready to connect with the audience.

heads, slides) have been the most common means to
involve an audience in a talk and and help them
understand the presenter's message. Students should
make sure their graphic aids are interesting, rele-
vant, and easy to read. However, students should
not expect graphics to carry their talk and automati-
cally turn it into a scintillating two-way conversa-
tion. That must come from the presenter.
Openings and Closings

Presenters should give extra thought to their
opening and closing sentences. Openings and clos-
ings are the most remembered parts of a presenta-
tion. They are also the most critical time for both
presenter and audience. Openings should clearly
demonstrate the presenters' command of the topic,
their purpose, and their organizational plan. Clos-
ings should be fitting and final. We counsel students
never to end with a shrug or with a weak statement
such as, "This ends my talk." They are urged to
think of a snappy closing sentence so that there is no
doubt that they have finished.
Presenting the Talk

Students may use light notes or an outline, but
we do not permit them to read their talks. In fact,
we advise them to never bring the full text of their
talk to the lectern. We coach students to walk briskly
to the lectern and stand quietly for a few seconds
before beginning, then take a deep breath, make
eye contact with the audience, and begin on a firm
clear note.
Presenters should dress for the part of speaker or
invited guest. We advise students to make sure they
will be comfortable in and confident of the clothes
they have selected.

Overcoming Nervousness
Presenters normally find that once they get
started, their initial nervousness wears off and they
lose self-consciousness. By knowing their opening
comments cold, students can guard against drawing
a blank or getting off on a weak start.
We recommend that students learn to accept a
certain degree of nervousness as an inevitable part
of being in front of people. Rather than fight the
nervousness or despair over it, we suggest that they
Winter 1990

use it as a stimulant. Some nervousness means a
presenter is keyed up, excited, ready to connect with
the audience. We tell students to remember that the
audience is made of real human beings, much like
themselves, who have come to see and hear the
presentation. We urge them to speak clearly and
decisively, feeling their voice carry to the far corners
of the room. We encourage our students to believe
that as they gain confidence, their voices will deepen
and steady.
If a student's nervousness is chronic and seri-
ously threatens the success of the talk, then the
student needs to take specific steps to overcome it.
We counsel such a student to analyze the nervous-
ness in order to pinpoint the cause: perhaps the stu-
dent is inadequately prepared; perhaps the student
is not in touch with the audience; perhaps the stu-
dent does not fully understand the central message
or does not wholly believe in it.
Once a student has established the root-cause of
the nervousness, he or she can begin to neutralize it.
Students may find that extra practice or deep-breath-
ing exercises help. Or it may be that all the student
needs in more experience in front of a group.
Last, but not least-we tell our students to smile.
Along with a confident voice, a sincere smile calms
and cheers the audience.

Preparing Visual Aids
We tell students to use graphics (overheads or
slides). A rule-of-thumb we share with them is to
change the graphic every two to three minutes. This
forces the presenter to break down complex textual
graphics into simpler ones and keeps the audience
from growing bored. If students have finished dis-
cussing a graphic but don't have another to immedi-
ately replace it, we advise them to turn off the over-
head projector. This brings the audience's focus back
to the presenter, where it belongs.
We recommend that students plan on a maxi-
mum of six to ten lines per overhead. An outline
form with bullets is best. A graphic should present a
distilled form of their comments (in the same way
the oral presentation is a distillation of the written
report). Presenters should never photocopy books or
typewritten material. It is generally illegible to most
of the audience and gives the impression that the

presenter threw the material together quickly. Pre-
senters should use numbers sparingly. The audience
will remember two or three key numbers, but when
too many are given they will be forgotten.

In addition, we urge students to plan their graph-
ics as a harmonious suite. They should de- sign a
format that presents their information attractively,
and then stick to that format throughout all the
graphics.

Other props (samples, models, etc.) are optional.
They can be good if they are interesting and rele-
vant. But too many will distract from the talk itself.

GRADING AND OFFERING FEEDBACK

Giving the students feedback soon after their
presentation is equally important to giving advice
and pointers beforehand. The oral presentation is
worth ten percent of the semester grade, so students
must take it seriously. The talks are graded in four
categories: organization, presentation, visual aids,
and answering questions. Each category is graded
on a numerical scale, and handwritten comments,
explanations, or suggestions are written in the mar-
gins (see Table 2).

Organization (fifty percent) includes: does the
talk start with an outline and end with a summary;
is there a logical flow of ideas or a rambling mono-
logue; has thought been given to the level of under-
standing of the audience; is the talk aimed at in-
forming, or merely at trying to impress?

Presentation (forty percent) includes: is the
speaker enthused about the topic, or merely droning
through it; is the speaking clear or mumbled; has
the speaker practiced enough to give the talk from
the vuegraphs or light reference to notes, as opposed
to reading most of it; is the talk too long (past the
target time)?
Visual aids (ten percent) includes: were the vue-
graphs done with some care, or thrown together at
the last minute; do they contain a reasonable level of
detail; are they legible?

Questions are asked by the instructor and other
students after the talk, and the speaker must give
evidence of having done enough research on the
chosen topic to expand on specific points mentioned
in the talk or related to the subject. Often the
instructor asks questions aimed at forcing the speaker
to think about and summarize the main point of
the talk, or to recognize how this topic fits into a

broader context.

Most students are not especially happy when
this assignment is given to them in class. As the day
of their presentation nears, many of them become
apprehensive and nervous. However, unsolicited
comments after the talks are over or after the se-
mester is completed, are almost universally positive.
Discussions with students after they have completed
the course or (especially) after they have been work-
ing in industry for a while indicate that they realize
(sometimes enthusiastically, sometimes grudgingly)
that training in oral technical presentations is a
valuable tool for a new engineer.

REFERENCES
1. Campbell, John A., "An Overview of Speech Preparation."
Part of the Series MODICOM: Modules in Speech Communi-
cation. Chicago Science Research Associates, Inc., Chi-
cago, IL (1976)
2. Detz, Joan, How to Write and Give a Speech, St. Martin's
Press, New York (1984)
3. McCroskey, James C., An Introduction to Rhetorical Com-
mnunication, Prentice-Hall Book Company, Englewood
Cliffs, NJ (1972)
4. Oshorn, Michael, Speaking in Public, Houghton Mifflin
Company, Boston, MA (1982)
5. Morrisey, George L., and Thomas L. Sechrest, Effectice
Business and Technical Presentations, Addison-Wesley
Publishing Company, Reading, MA (1987)
6. Peoples, David, Presentations Plus, John Wiley & Sons,
New York (1988)
7. Smith, Terry C., Making Successful Presentations: A Self-
Teaching Guide, John Wiley & Sons, New York (1984)
8. Wydro, Kenneth, Think on Your Feet, Prentice-Hall Press,
Englewood Cliff, NJ (1988) -1

USER-FRIENDLY PROGRAM
Continued from page 27.

Some examples of the conclusions obtained by simple
exercises are:
SResults obtained with different cubic EOS do not vary
significantly with one another. This can be observed by
giving a student a multicomponent mixture-for example,
a hydrocarbon condensate of known bubble point
pressure-and letting him arrange the equations in order
from best to worst. A comparison of the errors of each
equation leads rapidly to the conclusion.
SThe use ofkij can radically modify equilibrium properties
and the shapes and curvatures of the various diagrams.
A very nice example is observed with CO,-hydrocarbon
systems. The use of a kij equal to zero will usually
produce a poor fit. We encourage the students to
randomly try to find the best kij and observe the changes
in the diagrams. Later they may use the program to
calculate the optimum parameter and compare their
results.

Chemical Engineering Education

Two-constant liquid models, such as Wilson, are useless
unless experimental data are known to regress the binary
constants. Once these Aij are calculated, the fit for
polar substances is noticeably better than with cubic
EOS, especially at low pressures. Azoetropic systems
(e.g., ethanol-water) are specially suitable to show this.
Using an EOS for both phases can improve the
convergence of high-pressure equilibria.

These ideas are obvious to a researcher in
the area who has obtained this knowledge after
many cases of trial and error. In this sense, it
is stimulating to see students learn from first-hand
information.

The program is not limited to thermodynamics
courses alone, but can also be used as a tool in sepa-
ration processes and unit operations. The x-y and
H-x-y diagrams produced by the printer are the size
of a letter-page and suitable for simple MacCabe-
Thiele and Ponchon-Savarit calculations. The flash
calculations are also very useful due to the fact that
most models involved give good results at commer-
cial pressure and temperature ranges. At our uni-
versity the program is also used in higher-level
courses as a method for obtaining fast and accurate
equilibrium information. On a research level, the
program should prove to be a handy and simple-to-
use tool.

SUMMARY

In this report we have presented "Ekilib," a user-
friendly program suitable for teaching and research
in the area of multicomponent vapor-liquid equilib-
rium. This program has proven to be an exceptional
medium for providing students with first-hand ex-
perience in these types of calculations and for rais-
ing their levels of comprehension far beyond that
usually obtained by conventional teaching methods.
The Ekilib program is available to faculty members
and students for a nominal fee.

REFERENCES

1. Charos, G.N., P. Clancy, K.E. Gubbins, and C.D. Naik,
Fluid Phase Equilib., 23, 59 (1985)
2. Naik, C.D.. P. Clancy, and K.E. Gubbins, Chem. Eng. Ed..
19,7(1985)
3. Jolls, K.R., G.P. Willers, and L.D. Jensen, Educom. 4, 19
(1977)
4. Jolls, K.R., J. Burnett, and J.T. Haseman, Chem. Eng. Ed.,
17,112(1983)
5. Watson, P., M. Cascella, D. May, S. Salterno, and D. Tassios,
Fluid Phase Equilib., 27, 35 (1986)
6. Soave, G., Chem. Eng. Sci., 35,1725 (1980)
7. Schmidt, G., and H. Wenzel, Chem. Eng. Sci., 35, 1503
(1980)

8. Baez, L.A., and F.A. Da Silva, "Development of an Applied
Thermodynamics Computer Package," Eng. Thesis, Univer-
sidad Sim6n Bolivar, Caracas (in Spanish) (1989)
9. Paunovic, R., S. Jovanovic, and A. Mihajlov, Fluid Phase
Equilib., 6,141 (1981) 1

DEPARTMENT: LSU
Continued from page 5.

Douglas P. Harrison served as department head
for the next three years until he chose to return to
the teaching ranks. Additional historical informa-
tion for this period can be found in a 1979 issue of
Chemical Engineering Educationl .

Harrison was followed by Edward McLaughlin,
who served as department head until 1987 when he
resigned to become Dean of the LSU College of Engi-
neering. Faculty additions during this time included
Donald C. Freshwater, Kerry M. Dooley, Michael Y.
Frenklach, Gregory L. Griffin, Martin A. Hjortso, F.
Carl Knopf, Geoffrey L. Price, Danny D. Reible,
Richard G. Rice, Don L. Ristroph, Conrad B. Smith,
and David M. Wetzel.

Arthur M. Sterling joined the faculty in 1975 and
in 1987 consented to act as interim department head
until a permanent head could be found. This was
achieved in 1988 with the arrival of John R. Collier
from Ohio University.

Table 1 lists chemical engineering faculty, with
school of highest degree and dates of service at LSU.
Department heads are emphasized by boldfaced type.

REFERENCES
1. Traynham, James G., "Creating the Environment: A History
of the Louisiana State University Chemistry Department,"
Louisiana Academy of Sciences, Proceedings, 51, (1988)
2. LSU General College. Published by Louisiana State Univer-
sity. Baton Rouge, LA 70803 (The appropriate year will be
included at the reference point.)
3. Coates. Charles E., "An Experiment in the Education of Chemi-
cal Engineers: The Twenty-Fifth Anniversary ofthe Audubon
Sugar School," J. oflndust. and Engg. Chem., 9(4), 379 (1917)
4. Badger, W.L., and W.L. McCabe, Elements of Chemical
Engineering, McGraw-Hill Book Company, New York, NY
5. Perry, John H., Chemical Engineers'Handbook, McGraw-Hill
Book Company, New York, NY
6. Fieger, E.A., "History of Chemistry in Louisiana: The Develop-
ment of the Sugar Cane Industry," J. of Choem. Ed., 19, 303
(1942)
7. American Men (and Women) of'Science: Biographical Diction-
ary, edited by Jacques Cattell., The Science Press, New York,
NY (several editions)
8. Sterling, Arthur M., and Douglas P. Harrison, "Chemical En-
gineering at LSU," Chem. Eng. Ed., 13, 54 (1979) 1

Winter 1990

curriculum

USE OF A

MODERN POLYMERIZATION PILOT-PLANT

FOR UNDERGRADUATE CONTROL

PROJECTS

S. A. MENDOZA-BUSTOS, A. PENLIDIS,
AND W. R. CLUETT'
University of Waterloo
Waterloo, Ontario, Canada N2L 3G1

ost chemical engineering processes exhibit con-
siderable deviations from ideality since com-
plex physico-chemical phenomena are involved which
are difficult to describe quantitatively. It is not un-
usual, therefore, that the popular approaches to the
design of automatic control systems may result in
controllers that fall short of their design specifica-
tions and that a significant amount of time must be
spent in subsequent implementation and fine-tun-
ing stages.
Many of these "unmodeled" phenomena, which
are not likely to be observed in a bench-scale labora-
tory, may readily manifest themselves in an envi-
ronment that more closely resembles an industrial-
scale operation, such as a pilot-plant set-up. Testing
control algorithms in a pilot-plant rather than only
at the simulation level can help the control engineer
to at least attempt to take into consideration these
phenomena and to address important implementa-
tion issues before a control system is attempted on-
line on the particular application. A substantial re-
duction of the implementation time, on-line tuning
of the controller, and a more suitable control system
design may then be possible.

Sergio A. Mendoza-Bustos received his BEng
(1980) from the Technological Institute of Monterrey,
Mexico, and his MASc (1987) in electrical engineer-
ing from the University of Waterloo, Canada. He is
presently completing another MASc in chemical en-
gineering at Waterloo. His research is in the field of
process control.

1 University of Toronto, Toronto, Ontario, Canada M5S 1A4

To illustrate some of the general statements made
above, the most representative steps of a series of
undergraduate control projects are presented in this
paper. These control projects are part of the senior
undergraduate design course. They can be done on
an individual or group (2-3 students) basis and their
duration can be between four and eight months.
In these projects, the undergraduates are en-
couraged to experiment and even to redesign a flex-
ible reactor set-up in a modern computer-controlled
polymerization pilot-plant. Our objective at the out-
set was to "challenge" the students with something
more than a simulation exercise. For the purpose of
this paper and for the sake of brevity, three popular
control strategies, namely, a PID controller, a Smith
predictor, and a Dahlin controller (for details see
Stephanopoulos'1 ) are tuned, evaluated, and applied
for temperature control of a polymerization pilot-
plant reactor. In addition, a commercially available
real-time expert system shell was used to code and
implement a rule-based PI-type control algorithm.

Alexander Penlidis received his Dipl. Eng. (1980)
from the University of Thessaloniki, Greece, and his
PhD (1986) from McMaster University, Canada, both
in chemical engineering. He joined the department
of chemical engineering at the University of Water-
loo in 1986. His interests lie in the area of polymer
reactor modeling, design, optimization, and com-
puter control.

William R. Cluett received his BSc (1981) from
Queen's University, Canada, and his PhD (1986) at
the University of Alberta, Canada, both in chemical
engineering. He joined the department of chemical
engineering at the University of Toronto in 1986. His
S interests lie in robust control, adaptive control, and
applications to industrial processes.
Copyright ChE Division, ASEE 1991
Chemical Engineering Education

The design and implementation effort using the ex-
pert shell was compared with the one required for
the previous control strategies.
These control projects were very beneficial from
an educational/training point, both to the students
and to the faculty members. Our observations are
briefly discussed in the following sections.

TEMPERATURE CONTROL
IN POLYMERIZATION REACTORS

Polymerization reactions are exothermic in na-
ture; the amount of heat released during monomer
conversion to polymer is considerable. Temperature
variations greatly affect the kinetics of polymeriza-
tion processes, and through the kinetics they have a
strong impact on the way the produced polymer is
structured and thus, on its physical properties and
quality characteristics. Therefore, control of reactor
temperature is critical. If the polymerization tem-
perature is allowed to increase, monomer conversion
increases and more polymer is produced. Hence, the
polymerizing mixture becomes more viscous and heat
removal becomes more difficult. This may easily lead
to disastrous reactor runaways. Reactor tempera-
ture must, therefore, be kept within the limits that
allow one to carry out a safe polymerization, i.e.,
within the system's heat-removal capabilities'2-61

THE PILOT-PLANT REACTOR

Figure la shows a schematic of the pilot-plant
reactor layout, and Figure lb is a photograph of the
whole set-up. The pilot scale stainless steel reactor
consists of a vessel with a volume of 5 1, surrounded
by a jacket. Oil is pumped through the jacket to heat
or cool the vessel as required. The reactor is mounted
on a hydraulic lift stand and has a removable top
head which is held stationary while the vessel is
lowered or raised on the stand to open or close the
reactor. A variable-speed turbine agitator is used for
stirring the reactor contents. Auxiliary lines (e.g.,
vacuum, feed, nitrogen, etc.) enter and leave the top
of the reactor.
The heat transfer system consists of a water/oil
heat exchanger with two 9 kw electric heaters con-
nected in series and a 1 hp circulation pump. One of
the two electric heaters is automatically shut off
when the temperature of the oil is above 50 C. The
oil returning from the jacket is pumped through the
heat exchanger. A thermocouple at the exit of the
second heater (TT2) measures the delivered oil tem-

OIL O11. -
SURGE o
TANK OUT

I I

PCICUIP A lol
PUMP

1 MAS [EIl

t I OCAIL COllnIOI I.rn
SrTj" TC2|._T2 7

HIEATING CIIII E)
ELEMENT WATER

FIGURE la. Schematic of the pilot-plant reactor.

FIGURE lb. Photograph of the pilot-plant set-up.

perature. This measurement is sent to a local con-
troller (TC2), which turns on or off one or both heat-
ers and opens or closes the solenoid valve that al-
lows the cold water to flow through the heat ex-
changer. The function of this internal local control-
ler is to maintain the oil delivery temperature at a
desired value. From the heat exchange unit, the oil
is sent to the reactor's jacket. A by-pass pipe con-
necting the oil output with the oil input to the unit is
provided in order to prevent an excessive pressure
buildup from the oil pump.

Three type-J insulated thermocouples constitute
the set-up's sensor system. The first (TT1), inserted
through the stationary head, measures the tempera-
ture of the reactor contents. The other two measure
the oil temperature at the jacket inlet (TT2) and
outlet (not shown). The thermocouple signals are
sent to an IBM PS/2 60 computer via an OPTOMUX
OPTO-22 data acquisition system.

The control configuration consists of the local
internal controller (TC2) in the heat transfer system
connected in cascade with the master control algo-
rithm (TC1) in the process computer. The master

Winter 1990

Many of these "unmodeled" phenomena, which are not likely to be observed in a bench-scale
laboratory, may readily manifest themselves in an environment that more closely
resembles an industrial-scale operation, such as a pilot-plant set-up.

controller receives the measured reactor's tempera-
ture and calculates the value of the oil delivery tem-
perature that will maintain the reactor's tempera-
ture as closely as possible to a desired operating
point. The calculated value is sent to the heat-
transfer system where it becomes the set-point for
the internal controller. The internal controller is
simply an on/off control system and can be cali-
brated, activated, or deactivated via the unit's con-
trol panel keypad.

ANALYSIS AND DESIGN OF THE
CONTROL SYSTEM
Several monomer systems and modes of reactor
operation were tried during the design projects. The
tests included "dry runs" with water or solvent, solu-
tion polymerization of methyl methacrylate (MMA)
with varying fractions of solvent (40%-60%), emul-
sion terpolymerizations of styrene-butyl acrylate-
acrylic acid, "soap-free" emulsion polymerizations of
styrene, and low conversion (less than 40%) bulk
copolymerizations of styrene-butyl acrylate; the re-
actor modes employed were batch and semi-batch
isothermal and temperature programming cases. For
the sake of brevity, the examples that will be used
here will be drawn from the solution MMA
homopolymerization tests. However, the control
algorithms performed in a similar way for all differ-
ent monomer systems and modes of operation.

To perform the system identification, the reactor
was filled with solvent (about 4 1), and its tempera-
ture was raised to 50 C. Toluene was used as the
solvent. Steps of 20 C were applied to the reactor
temperature set-point, and a first order plus dead-
time model was fitted to the open-loop response. The
identified transfer function follows:

Gp(s)=e tdSpS+ (1)

with t, = 70 s, and T = 500 s.

Based on open-loop dynamics, a sampling time
of four seconds was chosed to approximate continu-
ous-time control.

The three algorithms considered are very popu-
lar in industry. The Smith predictor, usually applied

in conjunction with PI or PID control, provides dead-
time compensation, while the Dahlin controller forces
the closed-loop to behave in a desired first-order plus
dead-time fashion.

Recently, increasing attention has been given to
the use of expert systems for process control applica-
tions. In one of the design projects, a commercial
expert system shell (RTES) was used to code a simple
rule-based PI-type controller for the control problem
under consideration. This algorithm included heu-
ristics to compensate for dead-time and for changes
in the operating point. The results were compared
with those from the Dahlin controller, the Smith
predictor, and the conventional PID controller. For
all controllers, the performance criteria were mini-
mum oscillation about the reactor's set-point and
minimum response time. At steady-state, an error
band of 0.5 C was allowed. These criteria were
chosen so as to minimize variations in the polymeri-
zation rate, which in turn affect conversion (produc-
tivity) and product quality.

EXPERIMENTAL PROCEDURE

A nominal operating point of 50 C was chosen.
During the development and tuning stages, the re-
actor, with about 4 1 of solvent, was heated to the
operating point. Once steady state was achieved, a
change of set-point from 50 C to 70 C was applied.
Once the reactor stabilized at the new operating
condition, a set-point change from 70 C to 50 C was
issued. To test for disturbance rejection, 300 ml of
toluene at room temperature was then injected into
the reactor. The controllers were fine-tuned by trial
and error until the closed-loop responses met the
performance specifications.

As a final test, a "real" polymerization was car-
ried out using the different control laws. 1.4 1 of
MMA, 2.5 1 of toluene, and 33 g of initiator were
mixed into the reactor. The reactor contents were
brought from room temperature to the nominal op-
erating point (50 C), and the polymerization was
allowed to proceed for two and a half hours (Stage
A). After this period, the reactor's temperature set-
point was increased to 60 C and maintained at this
value for another hour (Stage B). Finally, 5 g of

Chem mical Engineering Education

initiator (AIBN) dissolved in 300 ml of toluene at
room temperature was injected into the reactor,
representing a disturbance. The reaction was al-
lowed to continue for one more hour (Stage C), and
then the polymerization was stopped and the reactor
was cleaned.

ANALYSIS OF THE RESULTS

The results that follow are representative of an
extensive set collected in the pilot-plant during the
design projects. The aim is to show how the use of
the pilot-plant helped in the design, implementa-
tion, tuning, and evaluation of several control algo-
rithms. The results also show the progression in the
students' steps/thoughts.

In general, according to the students' observa-
tions, the use of the pilot-plant was of great value in
mimicking a "real-world" process with respect to:

the unmodeled dynamics introduced by the use of a first-
order plus dead-time model to approximate the process
(bearing in mind that the process was a polymerization
system which exhibits highly non-linear phenomena,
such as a reaction exotherm and an increase in viscosity
["gel effect"] as the reaction proceeds)

the physical limitations and operational characteristics
of the equipment constituting the pilot-plant. This was
particularly important for the heat exchange unit whose
on/off internal controller and variable time-delay
associated with driving the oil to the desired set-point
(control manipulation) made the closed-loop system
prone to an oscillatory behaviour

the training point of view. The students had the chance
to enlarge their knowledge with typical polymerization
processes, systems that are sufficiently complex to present
a challenge to the undergraduates, and furthermore,
systems largely unstudied in the undergraduate
curriculum. Thus, this combination of"undergraduate-
level control algorithms" with "complex processes" made
the undergraduate design projects much more
meaningful.

First, the development and comparative analysis
of the controllers for the reactor full of solvent is
presented. Next, two of the controllers with the more
satisfactory performance are applied to the solution
polymerization of MMA. Finally, the results of the
rule-based controller are discussed.

Figure 2 shows the closed-loop response of the
reactor using the velocity form of a PID controller.
Although the controller brings the process to steady
state, one can observe an overshoot of approximate
8 C and a settling time of more than one hour.

The control objectives are, therefore, far from being
satisfied.

To improve the PID controller performance, up-
per and lower limits were imposed on the oil tem-
perature set-point, and the internal controller set-
tings in the heat-transfer unit were readjusted fol-
lowing discussions with the manufacturer. The
closed-loop response with the new bounds and the
finely-tuned PID controller parameters is shown in
Figure 3. One can see that the controller stabilizes
the process in approximately thirty minutes, with no
overshoot. The response was considered satisfactory,
and this controller was chosen as the base case for
performance evaluation.

Figure 4 shows the step-response from 50 C to
70 C of the Smith predictor and the Dahlin control-

LLGENI[

00 0 20.0 30.

100 0 5 0 60,0
TilE [MINi

0B 0 10 1100 120.0

FIGURE 2. Closed-loop Response (Toluene, PID
controller).

70 - - -

60
S Reactor Temp
-- Oil Setpt
Reactor Setpt
50 .-r ..r-r----r i i i . i . | 1 1
0 10 20 30 40 50
Time (Min.)
FIGURE 3. Closed-loop response (toluene, bounded
finelv-tuned PID controller).

Winter 1990

ler as compared to the PID. One can see that the
Smith predictor gives a smoother step-response than
the other two controllers. Figure 5 shows the closed-
loop response to a "cold-disturbance" for the three
controllers under consideration. In this case, the
performance of the Dahlin controller was better than
those of the Smith predictor and the PID controller.

Figures 6 and 7 show the reactor temperature
record for the solution polymerization of MMA un-
der PID and Smith predictor control, respectively.
Although both strategies gave acceptable control,
the Smith predictor generated smoother control ma-
nipulations, which provided a more acceptable be-
haviour of the oil temperature variations.

Finally, based on the previous observations, a

Time (Min.)
FIGURE 4. Closed-loop response (toluene, PID, Smith
predictor and Dahlin controllers).

Time (Min.)
FIGURE 5. Disturbance rejection (toluene, PID, Smith
predictor and Dahlin controllers).

rule-based PI-type controller was designed to com-
pensate for the process dead-time and to provide
smooth step response and good disturbance rejec-
tion. From the previous process experience, three
heuristics were defined and implemented:

No control action until the reactor temperature responds
to a previous manipulation within a tolerance band of+
0.50C (to compensate for dead-time).

The initial manipulation to a set-point change in the
reactor's temperature is to set the oil set-point equal to
the reactor set-point.
No PI control action until the temperature of the inflowing
oil is within a 1C band of the oil set-point generated by
the control algorithm. This heuristic is used in order to
reduce the oscillating behaviour due to the high
capacitance of the process (volume of reaction mixture

Time (minutes)
FIGURE 6a. MMA solution polymerization (PID control,
stage A).
80

~' 70-

Reactor Temp.
------ Oil Temp.

0 20 40 60 80 10 120
Time (minutes)

FIGURE 6b. MMA solution polymerization (PID control,
stages B and C).

Chemical Engineering Education

to be heated/cooled) in combination with the process
dead-time.

The behaviour of the rule-based controller is
shown in Figure 8 for the solvent test. Comparison
with Figure 3 reveals that it was possible to approxi-
mate the performance of the PID controller with
that of the rule-based controller.

CONCLUDING REMARKS

Working in a modern pilot-plant allowed the
students to gain experience in handling large vol-
umes of hazardous materials, process start-up and
shut-down, equipment failures, operational vari-
ations, scaling-up, equipment cleaning, and run-time
scheduling.

Time (minutes)
FIGURE 7a. MMA solution polymerization (Smith
predictor, Stage A).

These complexities not only made the design proj-
ects more meaningful from a technical point of view,
but also offered the students invaluable experience
in another important aspect which is largely un-
touched in undergraduate education-the aspect of
project/group management. By setting more difficult
objectives in a virtually open-ended project, and by
involving the undergraduates in something outside
of their "familiar turf' of departmental laboratories,
the whole project became a "group effort." Different
groups had to rely on the good or bad performance of
other groups in order to accomplish their objectives,
a fact that resulted in everyone who was involved
being interested in what others were doing. Coop-
eration with a group, coordination among different
groups, and time-planning and management became
the key features.
The operation required the participation of many
groups of undergraduate students, graduate students,
faculty members, technicians, and research engi-
neers, i.e., groups with different levels of expertise.
As one group of students put it, "...all of us benefited
from this learning experience tremendously. The
projects gave us the opportunity to supervise and to
be supervised, to plan and to organize, and to work
together in a large team...." And another group put
it, "...after these projects we appreciated much bet-
ter what one means by systems in series or in paral-
lel, interacting or non-interacting, and by the rate-
controlling step...."
Currently, we are trying to introduce new proj-
ects involving the implementation of more advanced
control techniques, such as minimum variance (con-
strained and unconstrained), predictive and/or adap-
Continued on page 60.

Time (minutes)
FIGURE 7b. MMA solution polymerization (Smith
predictor, Stages B and C).
Winter 1990

LEGEND
E SL rrl POINT
_f JV ^^

&0.0 o .
TIMC (mLn)

100.0 120.0 110.0

FIGURE 8. Closed-loop response (Toluene, bounded rule-
based controller).

i I p classroom

A ROBUST ALTERNATE

TO LEAST SUM OF SQUARES FOR

LINEAR REGRESSION

G. P. RANGAIAH
National University of Singapore
Singapore 0511, Singapore

R egression, which deals with fitting an appropri-
ate equation to a set of experimental data, is
often encountered in engineering and scientific
analyses. Of all the regression problems, fitting a
linear equation of the form y = a + px, containing
two parameters (a and p), is the most common one.
This problem is discussed in linear regression, while
multiple linear regression and nonlinear regression
deal with fitting linear equations containing three or
more parameters and nonlinear equations respec-
tively. Note that linear, in regression analysis, re-
fers to parameters to be estimated based on experi-
mental data.

LEAST SUM OF SQUARES
Given a set of data (x,, y for i = 1,2,...,n) and that
the equation for representing the data is
y = a + px, the main task in linear regression is to
find the best estimates of a and P. The popular
method used for this estimation is the familiar ordi-
nary (or unweighted, or equally-weighted) least sum
of squares (LSS). According to this method, estimates
of parameters are obtained by minimizing the sum
of squares of residuals,

n[yi-(a+bxi)]2 (1)
i=l
with respect to a and b. The quantities a and b are
the estimates of a and p, respectively, since a and P
cannot be found exactly because of inevitable experi-
mental noise.
LSS analysis is almost two centuries old. By solv-
ing the minimization problem in Eq. (1), a and b can
be obtained without making any assumptions. But

G. P. Rangaiah received his Bachelors, Masters,
and Doctorate degrees, all in chemical engineering,
from Andhra University, IIT Kanpur, and Monash
University, respectively. He worked in Engineers
India Limited, New Delhi, from 1976 to 1978. He has
been lecturing at the National University of Sin-
gapore since 1982. His research interests are in
process simulation, optimization, and control.

for subsequent inferences (e.g., on confidence inter-
vals, goodness of fit), the following assumptions are
generally implicit:'"
Equation selected for fitting the data is correct.
Independent variable, x, is free from noise.
Noise in the dependent variable, y, follows normal
distribution with mean zero and constant standard
deviation (SD).
Noise in y, and yi (i i j) is uncorrelated.
When these assumptions are satisfied, the LSS esti-
mator has nice features such as unbiasedness and
minimum variance; one can also estimate confidence
intervals and conduct goodness-of-fit tests.
However, LSS is sensitive to violation of the above
assumptions. If one or more of these assumptions
are not valid, then the estimates obtained
by LSS may be biased and/or may have larger vari-
ances. In other words, accuracy and precision of
LSS will be low. Although dictionary definitions of
accuracy and precision are similar, there is a dis-
tinction in their scientific usage.121 Accuracy and pre-
cision refer, respectively, to bias and variance
of parameter estimates obtained by a method. (For
the definition of bias and variance, see Equations 6
and 7, presented later.) Low accuracy implies larger
bias (in absolute value), and low precision means
larger variance.

Copyright ChE Division, ASEE 1991

Chemical Engineering Education

Let us discuss to what extent the above assump-
tions are likely to be satisfied in practical applica-
tions. The first assumption is likely to be valid since
knowledge of the physical phenomena relating to
the experimental data (and hence the appropriate
equation or model) is often available.
The second assumption relates to identifying the
independent and dependent variables in a given set
of data. The quantity which is free from noise or
which has negligible noise should be selected as the
independent variable, x. The other quantity, con-
taining relatively large noise, will be the dependent
variable or response, y. Frequently, this can be done
with reasonable confidence. For example, quantities
such as time, temperature, etc., can be measured
precisely; hence, one of these quantities should be x.
Often, these are the quantities which are adjusted
in engineering experiments. On the other hand,
quantities such as flow rate, concentration, reaction
rate, dimensionless groups, etc., generally contain
greater noise-so one of these quantities is a likely
candidate for y. Therefore, the second assumption
for LSS is often valid.
That leaves us with the last two assumptions.
Variance (or square of SD) of noise in y may or may
not be constant. One or more experimental points
may be outliers or wild points (see references 3 and
4 for the definition of outliers). Or there may be
systematic noise because of drift of instruments.
Hence, one often does not know whether these two
assumptions are valid or not. One might consider
fitting data by LSS, and then analyzing residuals
[i.e., the differences between y, and predicted value
(= a + bx)] for the satisfaction of assumptions."'
However, initial fitting by LSS is affected by outliers,
if they are present. This in turn will affect
residuals-thus complicating their analysis. This as-
pect, which is often not realized, will be demon-
strated later. Also, analysis of residuals is some-
times qualitative and subjective.
A few points are worth mentioning here. First, if
variance of noise in y is not constant, then the
weighted least sum of squares is a potential alter-
nate to (ordinary or unweighted) LSS. However, this
requires knowledge of variation of SD of noise for
determining the correct (relative) weights. Unfortu-
nately, often this detailed information is either not
available or is too expensive (in terms of experimen-
tal effort and cost) to obtain, thus precluding the use
of weighted least sum of squares. Therefore, LSS
hereafter refers to unweighted least-squares analy-

Since some of the assumptions ofLSS are
unlikely to be valid, there is a need for alternative
methods which are less affected by violation of
the assumptions, or that require less stringent
assumptions. These methods are generally
known as robust methods.
sis unless otherwise stated.
Second, LSS can be used to obtain parameter esti-
mates (by solving problems similar to Eq. 1) without
worrying about the satisfaction of underlying as-
sumptions. Apparently, this is what many users of
LSS do. However, in such situations accuracy and
precision of estimates obtained (as well as the use of
LSS itself) is questionable. Finally, this article is
concerned with point estimates rather than interval
estimates (such as confidence intervals).

ROBUST METHODS
Since some of the assumptions of LSS are un-
likely to be valid, there is a need for alternative
methods which are less affected by violation of the
assumptions, or that require less stringent assump-
tions. These methods are generally known as robust
methods. A robust alternate to LSS should be com-
parable to LSS (in accuracy and precision) when all
the assumptions stated above are satisfied, and it
should be better than LSS when one or more assump-
tions are not valid. Further, the method should pref-
erably be simple in principle and computations so
that it has the potential to replace LSS (which can be
readily done on many hand-held scientific calcula-
tors) in due course.
Many robust methods have been discussed in the
literature.'4' One of the promising robust alterna-
tives to LSS is the median method (MM), a nonpara-
metric method.'1561 A graphical version of MM, known
as direct linear plot, is popular with biochemists.'71
The objective of this paper is to emphasize the un-
derlying assumptions of LSS, to introduce MM, and
to show how it compares with LSS for linear regres-
sion. All estimation methods can be expected to
give nearly identical estimates if noise is negligible
or if the number of points in the data set is very
large. However, neither of these situations is true in
many applications.

MEDIAN METHOD
In order to estimate the two parameters in
y = a + bx, two points [say (xi, y,) and (xi, yj)] are suffi-

Winter 1990

cient, assuming x # x. (and absence of noise). The
estimates ai and b. are given by

Yi -Yj (2)
i xi xj

aij = Yi bij xi (3)

The subscripts indicate that ith and jth points in
the data set are used. These estimates can be termed
as preliminary estimates. Generally, there will be
more than two points. Hence, for n different x-
values, one will have n(n-1)/2 sets of preliminary es-
timates. (If there are replicates, the number of sets
will be less than this number.) The preliminary esti-
mates (of either a or b) are very likely to be different
from one another because of noise. In the absence of
noise, all preliminary estimates will be identically
equal to the true value.

In MM, the median of the above preliminary
estimates is then taken as the (final) estimate. On
the other hand, estimates of a(b) by LSS is equal to
the weighted average of all a (b..), the weights being
proportional to (x. x.)2. Hence, in MM, median re-
places the weighted average of LSS. Since median is
robust towards extreme values compared to weighted
average, MM can be expected to be less affected by
the violations) of the assumptions of LSS.

EXAMPLES

A set of simulated data with a = 2 and p = 1 is
presented as Case A in the second column of Table 1;
noise in this data set satisfies all the assumptions
underlying LSS. The data in the third column (Case
B) are similar to those in the second column except
that two points (corresponding to x; = 8 and 9) are
simulated as outliers.

Application of LSS to the two data sets in Table 1
is well known and need not be described. Use of MM
for Case A is briefly described here. In this set of
data there are 10 points, and all x. are different.
Hence one can calculate 10 x 9/2 = 45 pair-wise
intercepts and slopes using Equations 2 and 3 (with
i=1,j=2,3,...,10; i=2,j=3,4,...,10; ...; i=9, j=10). If all
x. are not different, then the number of pair-wise
estimates will be less than 45. The final estimate of
a (or p) by MM is the median of the 45 preliminary
estimates of intercept (or slope). The common defini-
tion of median is valid. For the data set under con-
sideration, the final estimates of a and b can be

found to be 1.66 and 1.03, respectively.
Estimates of a and p obtained by both LSS and
MM in Cases A and B are shown in Table 2. For Case
A, estimates by LSS are marginally closer to true
values than those by MM. However, either method
seems to be satisfactory in the ideal situation,
wherein the underlying assumptions of LSS are valid.
The results for Case B are shown graphically in
Figure 1. The fitted line by LSS is strongly influ-
enced by the two outliers at x; = 8 and 9. This
is reflected in estimates of parameters (Table 2). The
difference between a and a is about 45%, while that
between 3 and b is about 20%. These large errors
in estimates by LSS affect subsequent analysis of
residuals.

Residuals are nothing but vertical differences
between experimental points and the fitted line (by
LSS) in Figure 1, and these are plotted in Figure 2.

TABLE 1
Typical Data With and Without Outliers

xi yi
Case A Case B

1.0 2.68 2.68
2.0 3.74 3.74
3.0 4.79 4.79
4.0 5.76 5.76
5.0 5.60 5.60
6.0 8.54 8.54
7.0 9.08 9.08
8.0 9.80 12.8
9.0 11.2 14.2
10.0 11.0 11.0

TABLE 2
Estimates of a and 3 by LSS and MM

Estimates by
Parameter
LSS MM

Case A
a 1.72 1.66
S1.00 1.03
Case B
a 1.12 1.57
P 1.22 1.08

Chemical Engineering Education

The largest (absolute) residual corresponds to the
good point at x; = 10, and is -2.32. Residuals corre-
sponding to the two outliers at x, = 8 and 9 are 1.92
and 2.1, respectively. Hence it is not easy to distin-
guish the outliers from the rest of the data. In this
situation, one may opt to reject the point with the
largest (absolute) residual, which happens to be
a good point in the present example. If the point
(xi = 10, y. = 11) is rejected and LSS is reapplied to
the remaining 9 points, then the estimates of a and
P are 0.41 and 1.41, respectively. Now, the error in
the estimates is much more. This type of influence of
outliers on LSS, in the first place and on subsequent
analysis of residuals, is often not realized.
For the data of Case B, estimates obtained by MM
are presented in Table 2, and the fitted line is shown
in Figure 1 along with the corresponding results for
LSS. Estimates obtained by MM are closer to the true
values, compared to those by LSS. Further, estimates
by MM in Cases A and B are nearly equal. Hence MM,
unlike LSS, is almost unaffected by the outliers.
The robust performance of MM over LSS cannot
be concluded based on a single set of simulated data.
Therefore, LSS and MM are evaluated through Monte
Carlo tests. A detailed description of these tests and
results obtained on the Arrhenius equation are
presented elsewhere.181 Also, see reference 4 for some
results obtained through Monte Carlo tests. Here,
Monte Carlo tests are briefly described, and typical

11,
LSS
SMM
12 -

10

8 8

X
FIGURE 1. Case B-Data and fitted lines by
LSS and MM

results on the linear equation y = a + 3x are pre-
sented below.

MONTE CARLO TESTS

First of all, let us write the linear equation in-
cluding noise in the dependent variable, y:
yi =a+bxi+ei (4)

Noise, ei, is the origin of all difficulties in finding
precise parameter values and the need for a good
estimation method. In the tests, noise with different
assumptions regarding its distribution is simulated.
The steps in the test are as follows:
1. Assume a, P, n, xi, and distribution for e. In the present
tests, a, P, and n are 2, 1, and 10, respectively, while
values of x are those shown in Table 1.
2. Simulate noise, e (satisfying the assumed distribution),
and then calculate y, (= a + px, + e,) for i = 1,2,...,n.
Note that any two noises (e and e., i A j) will be
different.
3. For the simulated data (xi and y,, i = 1,2,...,n), evaluate a
and b by both LSS and MM. The estimates by the two
methods are unlikely to be identical.
4. In order to obtain the average performance of the
methods, repeat steps 2 and 3 many times say, 4000
times.

Hence, for each method and for each parameter
there will be 4000 estimates; these estimates are
likely to be different from one another as well as
different from the true (or exact) value. The vari-
ation of 4000 estimates is analyzed in terms of mean
square error (MSE). For example, mean square er-
ror of b is given by
4000
S(bk -)2
MSE of b= k=14000 (5)
It can be easily shown that MSE is the sum of square
of bias and variance141, defined as follows

2 *

-2

0 2 1. 6 a 10

FIGURE 2. Plot of residuals-Case B and LSS

Winter 1990

4000
1(bk )
k=l
4000

where b is the arithmetic mean of all the 4000 esti-
mates. The bias is usually small, and the main con-
tributor to MSE is the variance. Because of this, a
small MSE generally implies higher precision. There-
fore, the best (or most precise) method is one having
the smallest MSE.

RESULTS AND DISCUSSION

The results of a few Monte Carlo tests are sum-
marized in Table 3. The variation in these tests is
essentially in the distribution of noise, which is al-
ways assumed to be normally distributed but with a
different mean and/or variance. This is because nor-
mal distribution is justified in many situations. In
the first test in Table 3, SD of noise in y is constant
and equals to 0.6, while mean of noise is zero; fur-
ther, there is no noise in x. Hence, the underlying
assumptions of LSS are all satisfied. In the remain-
ing tests, generated noise is such that one assump-
tion of LSS is relaxed or violated for evaluating the
robustness of MM compared to LSS.

In tests 2 to 5, SD of noise in y depends on the
true value of dependent variable, y* (= a + Px); the
assumed SD of noise is stated in Table 3. In the sixth
test, 2 points (selected at random) out of 10 are
simulated as outliers. Mean and SD of noise for out-
liers are chosen as 3.0 and 0.6, respectively, while
the corresponding quantities for regular points are
0.0 and 0.6, respectively. In the last test, unlike tests
1 to 6, noise with mean 0.0 and SD 0.6 is included in
y as well as in x. That is, both x and y are subject to
noise with the same distribution. However, noise in
x, is neither equal nor correlated to noise in the
corresponding y.

The results in Table 3 show that in the first test,
MSE of either a or b by LSS is lower than that by MM.
This is expected since the assumptions of LSS are all
satisfied. In tests 2 to 5, in which SD is not constant,
MSE for MM is generally smaller than that for LSS.
That is, MM is more precise than LSS. MM is also
superior to LSS when there are outliers (test 6). When
the same noise is present in both x and y (test 7), LSS
is better than MM.

Hence, performance of MM is less affected by
variation in SD of noise or by outliers. However,

Bias of b= b -

Chemical Engineering Education

Variance of b

TABLE 3
Results of Monte Carlo Tests on the Performance
of LSS and MM

Test Noise Characteristics MSE of a (MSE of b) x 100
# (see text for details) LSS MM LSS MM

1 SD= 0.6 0.164 0.216 0.421 0.477
2 SD= 0.2y* 0.621 0.435 3.34 3.20
3 SD = 3.0 / y* 0.259 0.268 0.471 0.397
4 SD = 0.006 (y*)2 0.047 0.012 0.358 0.247
5 SD = 9.0 / (y*)2 0.203 0.095 0.363 0.134
6 2 Outliers 1.19 0.931 2.11 1.56
7 Noise in both x and y 0.342 0.436 0.889 0.993

when SD is constant or when noise is present in both
x and y, MM is marginally inferior to the popular
LSS. Results in Table 3 indicate that error in a is
generally much more than that in b. This means it is
more difficult to estimate the intercept precisely.

CONCLUDING REMARKS

Ordinary (or unweighted) least sum of squares
has been the workhorse for simple linear regression
during the past two hundred years. However, its
precision is affected when noise in the dependent
variable has variable standard deviation, or when
there are outliers. A promising robust method for
these situations is the median method, where the
principle and computations are simple. This article
brings this method to the attention of chemical engi-
neers and presents some results to show its robust-
ness. Interesting research is progressing on regres-
sion diagnostics for outlier detection and on robust
(or resistant to outliers) regression. The recent book
by Rousseeuw141 describes both these topics, as well
as another promising robust method known as least
median of squares.

REFERENCES
1. Gunst, R. F., and R. L. Mason, Regression Analysis and Its
Application, Marcel Dekker, New York (1980)
2. Bevington, P. R., Data Reduction and ErrorAnalysis for Physi-
cal Sciences, McGraw-Hill Book Company, New York (1969)
3. Draper, N. R., and H. Smith, Applied Regresion Analysis, John
Wiley and Sons, New York (1981)
4. Rousseeuw, P. J., and A. M. Leroy, Robust Regression and
Outlier Detection, John Wiley and Sons, New York (1987)
5. Sen, P. K., J. Amer. Statist. Ass., 63,1379-89 (1968)
6. Hollander, M., and D. A. Wolfe, Nonparametric Statistical
Methods, John Wiley and Sons, New York (1973)
7. Cornish-Bowden, A., Fundamentals of Enzyme Kinetics, But-
terworths, London (1979)
8. Rangaiah, G. P., Chem. Eng. J., 29,159-166 (1984) 1

= book review

STATISTICAL MECHANICS OF CHAIN
MOLECULES
Paul J. Flory
Oxford University Press, 200 Madison Ave., New York,
NY 10016; $49.95 (1989)

Reviewed by
P. T. Cummins, J. W. Rudisill
University of Virginia

The late Paul J. Flory's book, Statistical Me-
chanics of Chain Molecules, was first published in
1969 by John Wiley and Sons. This edition, pub-
lished by Carl Hanser Verlag and distributed in the
U.S. by Oxford University Press, is a reprint of the
1969 volume with corrections and additional remarks
by Flory.
Flory was a remarkable scientist whose career
included industrial research and development (Du-
Pont, 1934-37; Exxon, 1940-43; and Goodyear, 1943-
48) and distinguished academic teaching and re-
search (University of Cincinnati, 1937-40; Cornell
University 1948-57; Mellon Institute, 1957-61; and
Stanford University from 1961). By the time of his
death in 1985, he had received a number of prestig-
ious awards, including the 1974 Nobel Prize in
Chemistry, and many honorary degrees.
Flory devoted his scientific career to the elucida-
tion of the physical principles underlying the confor-
mational and thermodynamic properties of polymers
in solution. The theoretical framework was provided
by statistical mechanics. The systems were charac-
terized experimentally through techniques such as
light scattering, neutron scattering, and thermo-
physical property measurements. Statistical Me-
chanics of Chain Molecules brings together into one
coherent work the many contributions made by Flory,
his co-workers, and other researchers into developing
a statistical mechanical description of the conforma-
tional properties of chain molecules. The approach is
to focus on the statistical mechanics of single chains
so that the solvent is regarded as a continuum.
In the preface, Flory states that one of his goals
in writing the book was to provide full details of
mathematical derivations in order to make the book
as self-contained as possible. In consequence, the
layout of the book is quite methodical.
Chapter I introduces the concepts of spatial dis-
tributions of chain molecules, mean square end-to-

S end distance and mean square radius of gyra-
tion . Some simple models for polymer chains-
the freely jointed chain (a random flight with fixed
bond lengths, random bond angles, and free rotation
around bonds) and the freely rotating chain (with
fixed bond lengths, fixed bond angles, and free rota-
tion around bonds)-are introduced.
In Chapter II, the term random coil is introduced
to define an isolated chain molecule which, due to
the absence of constraints, is free to take up any of
the vast number of configurations allowed by rota-
tions about bonds between neighboring units in the
moelcule. Some of the experimental techniques used
to determine , , and d In IdT where T is
temperature, such as intrinsic viscosity, hydrody-
namical measurements (sedimentation velocity and
diffusion coefficient), and light scattering, are de-
scribed and representative measurements reported
for a large class of polymer repeat units.
Chapters III and IV describe the principal
mathematical techniques used to compute the parti-
tion functions of chain molecules with realistic bond
potentials and steric, disperion, and multipolar in-
teractions between atoms in the polymer. The key
simplification is the adoption of the rotational iso-
meric state approximation which assumes that, once
the conformational energy has been computed to
determine the rotational potential, the minima in
the potential are taken as the only possible confor-
mations of the bond. Thus, each bond is treated as
occurring in one or another of several discrete rota-
tional states. The rotational isomeric state approxi-
mation allows the partition function for the molecule
to be treated as a summation over a finite number of
states. The summation can be represented succinctly
as matrix products once the statistical weight ma-
trix U is known. The element U.. essentially gives the
probability that a bond in rotational state i will be
followed in the chain by a bond in rotational state j.
The elements of U can therefore be obtained from
torsional and intramolecular potentials. All the
properties of interest-such as moments of the spa-
tial distribution including the dielectric constant-
can then be obtained from the partition function by
matrix manipulation of products (including direct
products) of the statistical weight matrices.
Chapters V, VI, and VII then implement the ro-
tational isomeric state model to compute the confor-
mational properties of, respectively, symmetric chains
(n-alkanes, polyethylene, and other polymers whose
repeat unit does not contain a symmetry-breaking
Continued on page 53.

Winter 1990

curriculum

THE POWER OF SPREADSHEETS

in a Mass and Energy Balances Course

MICHAEL MISOVICH, KARYN BIASCA
University of Wisconsin
Stevens Point, WI 54481

T he use of computers for design and analysis in
chemical engineering applications has flourished
in the past decade, and most chemical engineering
curricula include a computing language requirement,
such as FORTRAN, BASIC, or PASCAL, usually
taken in the freshman or sophomore year. Upper level
courses often include instruction in the use of soft-
ware packages for process simulation and design.
We feel, however, that the use of spreadsheets for
engineering calculations has not been widely enough
integrated across the curriculum. We feel that this
kind of instruction is particularly important since
graduating engineers may not have access to sophis-
ticated software packages or programming languages,
but will almost certainly have spreadsheet software
available for their use. Although the structure and
use of spreadsheets is quite different from program-
ming languages, many of the same kind of problems
can be solved by either method.
Recent articles have appeared which described
the use of spreadsheets for chemical engineer-
ing calculations I' and their use in a senior design
class.12" We have found it fruitful to expose students
to spreadsheets at the earliest possible point in the
curriculum-in the sophomore mass and energy bal-
ances class. Because of the relative ease in learning
spreadsheet software, we find that the investment of
a few hours of class time at this point allows students
to work more meaningful problems and to examine
problems in greater detail.

ADVANTAGES OF SPREADSHEETS
Many engineering students find that using FOR-
TRAN or a similar language makes a problem more
difficult to solve because the bulk of their effort must
be centered on the details of programming syntax.
As instructors, we sometimes compensate for this by
not making the engineering part of the problem too
difficult or too time-consuming. Students then get

the message, "I know this is an easy problem, but
look how hard it becomes when using the computer,"
and as a result, they may subsequently avoid using
the computer and avoid solving complicated prob-
lems for which hand-solution is not feasible.
Using spreadsheets alleviates some of the prob-
lems that students encounter with traditional pro-
gramming languages. The syntax of spreadsheets
essentially consists of algebraic formulas and func-
tions, concepts which are familiar from elementary
algebra. On the other hand, programming languages
require the mastery or memorization of numerous
keywords, punctuation marks, and special symbols,
which in most cases must be combined by using rigid
rules. So it is not surprising that students can learn
to do meaningful calculations much more quickly by
using a spreadsheet than they can with a program-
ming language.
Spreadsheets also help reduce the proliferation
of syntax errors and logical errors. If a syntactically
incorrect formula is entered, the programmer is im-
mediately notified of the error and can correct it be-
fore continuing. Since the numerical result of every
formula is displayed after it is entered, grossly in-
correct results can be seen immediately, and logical
errors in formulas can be corrected before being
propagated through the remaining calculation steps.

MichaelJ. Misovich is an assistantprofessor
in the paperscience departmentat the Univer-
sity of Wisconsin-Stevens Point. He holds
a PhD in chemical engineering and an MS
in computer science from Michigan State
University.

Karyn L. Biasca isanassistantprofessorin the
paper science department at the University of
Wisconsin-Stevens Point. She holds a PhD
From the Institute of Paper Chemistry anda BS
Sin chemical engineering from the University of
California at Los Angeles.
0 Copyright ChE Division ASEE 199I1
Chemical Engineering Education

In typical programming languages, such errors may
be difficult to debug without adding numerous PRINT
or WRITE statements which will list results of all
intermediate calculations leading up to the wrong
answer.

It is easy for students to document their spread-
sheet work because of the convenience of entering
text labels next to numeric cells. When program-
ming languages are used, students often avoid such
documenting features as comments and PRINT state-
ments to label output because of the extra effort
needed to include them.

Using a spreadsheet also helps students to effi-
ciently organize their calculations. The grid struc-
ture facilitates creation and organization of one- and
two-dimensional tables. Contrast this to FORTRAN,
where creating a two-dimensional table requires the
correct nesting of an implied DO loop inside an ordi-
nary DO loop, manipulating array subscripts in the
proper order, and correct interaction of the implied
DO loop with its FORMAT statement.

Flexibility of data input is more apparent in
a spreadsheet, since results can be recalculated
when any cell is changed, at any time. In program-
ming languages, only variables listed in READ or
INPUT statements can be changed while the pro-
gram is being executed. Students may circumvent
this by using an assignment statement: an example
of this in BASIC would be "LET TEMP = 298" rather
than "INPUT TEMP". This makes testing several
cases or ranges of values for independent variables
much less convenient.
Since spreadsheet software includes graphics ca-
pability, students can generate graphical results with
ease. Displaying the results of calculations in a chart
or graph often provides additional insight into the
physical significance of their calculations. When us-
ing programming languages, such graphics may be
generated only by special subroutines or procedure
calls which are not familiar to many students.
The human element of programming has been
studied in the fields of software engineering and
structured programming,'3' and it is believed that
excessive flexibility in program design leads to pro-
grams which are complicated to understand and dif-
ficult to debug. Spreadsheet programming offers
fewer of these pitfalls because it works with an in-
herently structured format. We believe that the use
of spreadsheet software aids students in organizing
their work on problems done by hand. The rows and
columns of a spreadsheet encourage the habit of
Winter 1990

setting up tables to solve mass and energy balance
problems. This approach is often helpful (particu-
larly to novices) but is not apparent when problems
are solved by writing conventional computer pro-
grams.

One criticism of spreadsheet software is that it
lacks the power of IF-THEN-ELSE and DO state-
ments found in programming languages. Yet most
spreadsheets do contain functions which implement

Wefeel... the use of spreadsheetsfor
engineering calculations has not been widely
enough integrated across the curriculum....
students can learn to do meaningful calculations
much more quickly by using a spreadsheet...

IF-THEN-ELSE logic, and most of them support it-
eration (which is probably the most common use of
repetition logic in engineering calculations) by circu-
lar recalculation. Most also support some type of
"macro," "program," or "project" processing, which
constitutes a programming language for manipulat-
ing the spreadsheet. Examples of this are given in
Reference 4.
When the above factors are considered, the func-
tionality of spreadsheet software is comparable to
programming languages for most types of engineer-
ing calculations. Actually, however, we do not need
to make use of these advanced features in the mass
and energy balances class since this type of logic is
not usually necessary in balance calculations. A
straightforward accounting of inputs and outputs
around a system boundary usually suffices, and
spreadsheets are the ideal software for such calcula-
tions.

Spreadsheets are not a substitute for modular
simulation packages such as Aspenplus.15' For com-
plicated, multiple-unit balances, these packages are
preferred because the balance calculations and physi-
cal data are built in. To students who are solving
simpler problems and are learning to solve such
problems for the first time, the overhead associated
with simulation packages makes the problem seem
more difficult to solve. Also, students do not practice
setting up and solving balances when the package
does it for them.

SPREADSHEET INSTRUCTION
We present several spreadsheet computing prob-
lems during the mass and energy balances class.
Typically, students entering this class have taken
FORTRAN as freshmen or sophomores, but few have

had experience with spreadsheets. Since spread-
sheets are a powerful and convenient tool which will
be of value to the student throughout the entire
chemical engineering curriculum, we felt that we
should offer instruction at the earliest possible time.
That is why we chose this particular class as the ve-
hicle for spreadsheet instruction. As a result, in-
structors in subsequent classes who want to set up
computing problems will find that they have stu-
dents who are familiar with spreadsheets.
One of the principal advantages of spreadsheet
computing is that it is simple to learn. In a two-hour,
computing lab period, students can learn enough to
complete a mass balance for a combustion process in
which an independent variable such as the percent-
of-excess air or percent-of-incomplete combustion is
varied. They can construct tables and graphs illus-
trating these results.

Many mass and energy balance problems can be
completed with only knowledge of a small set of op-
erating concepts and commands. The concept of cell
addresses is vital, as is the distinction between
numeric data, text data, and formulas. Commands
for formatting cells, editing existing cells, and copy-
ing cells are needed. For the copying operation, the
distinction between absolute and relative cell ad-
dresses is important. Although some spreadsheets
refer to this as "copy without adjustment" and "copy
with adjustment," all perform the same function-
they hold either the row or column number of a cell
fixed during copying. This set of commands is suffi-
cient to create and edit a spreadsheet and to display
results on the monitor. In addition, commands to
produce printed results and to save and reload pre-
vious work are needed. Finally, commands for pro-
ducing graphics are used to identify the coordinates
to be graphed, to produce labels and titles, and to
display or print the graph.

Students are encouraged to obey a few custom-
ary rules of spreadsheet usage in our class. We ask
that all numeric input data be placed at the top and
all formulas be placed at the bottom of the spread-
sheet, with the sections separated by a visible line.
All values, whether input data or calculated out-
put, should be labeled. Circular (or forward) refer-
ences, in which a formula contains a cell which is in
a later row or column, should be avoided unless
iteration is specifically intended. These may lead to
errors when input data are changed because the
formula will calculate results using cells which them-
selves have not been recalculated based on the new
data.

We expect students to work substantially through
a combustion mass balance problem, applying these
spreadsheet concepts, in a two-hour lab period. Each
step is explained in a written handout, and the in-
structors are available to give immediate assistance
to any student who becomes confused.
The first time this introductory exercise was given,
in 1985, students completed a follow-up survey which
included the question, "Would you find solving prob-
lems like the one we did easier using a spreadsheet,
or by writing a computer program in FORTRAN,
BASIC, PASCAL, etc.?" Of forty-one students who
responded, 80% preferred the spreadsheet, 17% did
not know, and only 3% (one student) preferred using
a programming language. Some of the student com-
ments we received were:
Using a spreadsheet keeps the problem more or-
ganized and actually saves time.
It's easier in spreadsheet because it simplifies the
process of solving. The other languages tend to
complicate the material balance.
In the spreadsheet, all data was visible at all times
and the data was set up in a nice tabulated form
that was easily changed.

Mistakes were easier to correct

Since this first experience, we have refined the
exercise by improving handouts and becoming
acclimated to the potential problems caused by a
relatively large student group using a public com-
puting lab.

ASSIGNMENT EXAMPLES
We typically assign from five to eight spread-
sheet problems during a quarter or semester class.
These problems are given in sequence with the ma-
terial as it is discussed in lecture. They also coordi-
nate well with the presentation of subject matter in
textbooks we have used.'6'71 Copies of these assign-
ments are available from the authors upon request.

The combustion mass balance problem described
above is given first. The goal is to calculate wet and
dry analysis of a flue gas. Typical variables include
the empirical formula of the compound burned, de-
gree of incomplete combustion, and percentage of ex-
cess air supplied. Later, in another problem, this
same spreadsheet is expanded to include gas con-
cepts such as volumetric flow rate, partial pressure,
vapor pressure, humidity, and dew point. After en-
ergy balances have been studied, the same problem
is expanded a third time to include temperature and
enthalpy data, with the goal of calculating adiabatic

Chemical Engineering Education

combustion temperature. This three-problem series
is always used.
A fourth problem is used, when real gases are
studied, to demonstrate the link between equations
of state and generalized compressibility charts. To
students, these methods may seem entirely dissimi-
lar. Using a spreadsheet, several constant tempera-
ture or constant volume curves of a compressibility
chart are plotted with PVT data generated from an
equation of state. When students compare their own
curves to those shown in the textbook charts, the
reaction is often one of amazement over the striking
similarity of the results. A sample of typical output
from this assignment is given in Figure 1.
The graphics capability of spreadsheets is also
applied in two other optional assignments. The first
of these problems asks the students to generate an
enthalphy-concentration diagram for a binary sys-
tem (such as water-sulfuric acid) using tabulated
heat capacity and enthalpy of solution data. The
second problem is the construction of a psychromet-
ric diagram from equations describing vapor pres-
sure and enthalpy. As with the compressibility dia-
gram assignment, these assignments tie together
the algebraic, tabular, and graphical techniques of
problem solving. Students see that similar results
can be obtained by using different methods. A re-
cent article'8l recommends using "computer friendly"
techniques in place of, or in addition to, graphical
techniques whenever possible; the above assignments
provide examples of this.

Other problems that are sometimes used are 1)
the generation of enthalpy of vaporization data as a
function of temperature from various forms of the
Clausius-Clapeyron equation or the Watson correla-
Compressibility Chart for
Ammonia
using SRK eqn of state

Reduced pressure pr

FIGURE 1. Sample compressibility chart.
Winter 1990

tion, and 2) generation of an enthalpy table (like the
steam table) for a substance other than water. These
problems demonstrate the correspondence between
equation-oriented and tabular approaches for esti-
mating enthalpies.

We have taken a multiple-effect evaporator prob-
lem from a textbook."' This problem is readily adapt-
able to a spreadsheet solution as the problem state-
ment requests that a table of properties be filled in
for each effect. Each effect obeys similar equations,
so a copying operation can be used to fill in the entire
table once the equations for one effect have been
entered. The manipulation of one cell containing an
independent variable (the vapor flow from the last
effect) allows recalculation of the entire table. Stu-
dents adjust this value by trial-and-error until the
desired liquid composition leaving the last effect is
generated. This problem demonstrates that trial-and-
error can be a practical solution technique for com-
plicated problems when the computer is performing
the calculations. Students may at first resist using
trial-and-error because they have been taught it is
not mathematically "elegant" and can be labor-in-
tensive when done by hand.
Students have successfully solved the problems
described here using several spreadsheet pack-
ages, including Lotus 1-2-3,19' Supercalc3 and Super-
calc4,'10' and the Smart Spreadsheet,'111 on several
types of IBM and IBM-compatible microcomputers.
The only major difference between these packages is
the adjustment of absolute and relative cell refer-
ences in copying operations. Other differences be-
tween these spreadsheets are usually immaterial to
students in our class.
Once a spreadsheet has been designed, students
can use it to answer questions or to gain additional
understanding or insight into the problem. For ex-
ample, we typically require students to solve the adi-
abatic combustion temperature problem for two dif-
ferent compounds, and then they compare the re-
sults and report on their findings in a short memo-
randum similar to one described in an article by
McKean and Hanzevack.'12' The class can then pool
its results and discuss questions such as, "What
chemical characteristics appear to lead to higher
combustion temperatures for fuels?" or "Why do we
use carbon (coal) and methane (natural gas) as fuels
when their combustion temperatures are among the
lowest of the compounds under the conditions we
studied?" Similar discussion can be facilitated for
many of the other problems by assigning different
combinations of independent variables to different
Continued on page 52.

o lass and home problems

The object of this column is to enhance our readers' collection of interesting and novel problem
in chemical engineering. Problems of the type that can be used to motivate the student bypresentinj
a particularprinciple in class, or in a new light, or that can be assigned as a novel home problem, ar
requested as well as those that are more traditional in nature, which elucidate difficult concepts
Please submit them toProfessor James 0. Wilkes and Professor T. C. Papanastasiou, ChE Departmenl
University of Michigan, Ann Arbor, MI 48109.

AMUNDSON'S MATRIX METHOD

FOR BINARY DISTILLATION REVISITED

J. J. J. CHEN
University ofAuckland Further, r, and r2, the roots of Eq. (3), will
Auckland, New Zealand .. .,,
faui~ E24 5)JJ ThSL

A mundsonll' expressed the binary distillation prob-
lem as a matrix difference equation. In this
paper, matrix power equations will be used to solve
and simplify the same problem, making it suitable
for illustrating the application of matrices in courses
of engineering mathematics or separations processes.

SOME RELATIONSHIPS IN MATRICES
Consider the matrix A of order 2:

A= a2[
Arbi b2
whose characteristic equation, I A rII = 0, is

a1 -r
b,

a2 =0
b2-

r2 -(al+b2)r+(ab2 -a2b)= 0 (3)
where I is the unit matrix of the order of A. From the
Cayley-Hamilton theorem, the matrix A also satis-
fies its own characteristic equation. Thus
A2 -(a + b2)A+(alb2 a2b)I=0 (4)
where O is the zero matrix of the order of A.
By using equations like Eq. (4) for higher powers
and substituting from the lower power equations, it
can be shown that

AP = aA + p (5)
where a and 1 are numerical constants which de-
pend on the matrix A and exponent p.
Copyright ChE Division, ASEE 1991

Also

satslly q. k,. Lus
rp = ar + (6)

r = (r2 + P (7)
Eqs. (5), (6), and (7) will be applied to binary distilla-
tion. However, first we need to formalize some of
Amundson's treatment.

BINARY DISTILLATION

Following Amundson, by assuming constant vola-
(1) utility, the equilibrium line is
y x (8)
A+Bx
(2) and the operating line is
y =mx + b (9)

Taking the top product composition as d, the use of
a total condenser gives y, = d. The plate numbers are
counted from top to bottom. The liquid leaving plate
1 is obtained from Eq. (8). Thus
X= Ayl (10)
-By +1
and y2, obtained from the substitution of x, from Eq.

John J. Chen is an Associate Professor in chemical
and matenals engineering at the University of Auck-
land, New Zealand.

Chemical Engineering Education

s
g
e
$.
t,

(10) into the operating line equation [Eq. (9)], is
Y2 = mX1 + b
Simplifying

(mA Bb)yi + b
Y2 ByI +1
-By, +1

and assuming constant volatility and molal over-
) flow, the only unknowns are a and p. Furthermore, P
may be solved explicitly in terms of a by using Eq.
(18). Thus

(11b)

[-Bynp+n +Yp+n -(mA- Bb)y -b]c
Yn -Yp+n

(19)

Now we can define y n, the composition of the
vapour leaving the (p+n)t' plate, as

Yp+n =p+n/p+n) (12a)
where

Yp+n 1[ a a21plyn
yp+n bLi b2

(12b)

The plate number from which we begin the step-
ping-off process is n. The value of y when p = 1
(i.e., Yl+n), is thus given by

Yl+n alYn + a2
Yl+n a.n. (13)
Yl+n blyn + b2
By comparing Eqs. (11b) and (13), where p = 1 and
n=l,

a = mA-Bb

a2 =b

b, =-B

b2 = 1

(14a)

(14b)

(14c)

(14d)

Thus
y2. (mA-Bb) by (15)
Y2 -B 1.1 1
With reference to Eq. (12b), it may be shown by
induction that
Yp+n*r 1 (mA-Bb) bpyn
JL B b]P[Y] (16)
yp+n** [ -B
Applying Eq. (5), Eq. (16) may be re-written as

SYp+n1 '[a(mA-Bb) ab][Yn][p O][y,]
Lyp+n*J L -ab ac Ll] Lo pi 17)
The composition of vapour leaving the (p+n)th plate
may then be written in terms of the vapour leaving
the nth plate as
Yp+n= Yp+n* [(mA -Bb)+ P]Yn +b (18)
~p+n ** -aByn +a+P
In Eq. (18), given a binary distillation problem

It is now possible to eliminate a and P in Eqs. (6) and
(7), and to evaluate p, the number of plates. The
factors r, and r2 are the roots of the characteristic
equation of the square matrix given in Eq. (16), and
they may be readily shown to be

r1,r2= [(mA-Bb+1)+ (mA-Bb+1)2 -4mA (20)

Dividing Eq. (6) and Eq. (7), substituting the value
for P from Eq. (19), and eliminating a

rl ri(yn -Yp)+[yp+n(1 Byn)-(mA-Bb)y -b]
r2 r2 (-Yp+)+[yp+n(1-Byn)-(mA-Bb)yn-b]

(21)
Thus, the number of plates p between tray number
(p+n) and n is given by
^1(y ypy )+[yp^(l- Byj -(mA Bb)y, b]
r (yn-yp+n )+[yp+n(1-Byn)-(mA-Bb)yn-b]
r2 (yn p+n) + p+n (1 -Byn) (mA -Bb)yn -b]

APPLICATION
We shall now apply Eq. (22) to the same problem
considered by Amundson in solving the distillation
of a 0.40 mole fraction benzene mixed with toluene
introduced at its bubble point. The equilibrium curve
is given by
x
y --- (23)
0.41+0.59x
The top produce is 0.995 benzene, and the bottom is
0.005 benzene. The operating lines above and below
the feed are, respectively
y =0.75x + 0.249 (24)

y= 1.3773x + 0.001886 (25)
The roots for the characteristic equations for above
and below the feed are, respectively

r1 = 0.75130

r2 = 0.40920

(26)

Winter 1990

r = 0.99744

r20.56615

Applying Eq. (22) to above the feed position,
n = 1, y, = 0.995, inserting the appropriate values for
A, B, m, and b from Eqs. (23) and (24) with reference
to Eqs. (10) and (11), and using r, = 0.75130 and
r, = 0.40920, we obtain y1,+ by substituting the feed
composition into the operating line as the feed is in-
troduced at its bubble point. Thus yn l = 0.549 (or
0.553 using the lower operating line).

Applying the values

r 0.75130

yl = 0.995
B= 0.59
A= 0.41

r2 = 0.40920

Ypl = 0.549
m= 0.75
b= 0.249

results in 9.54 plates above feed position.
Below the feed position, Eq. (22) may be applied
by taking n = 9.54, i.e., y954 = 0.549. We obtain y,,,
by substituting values for A, B, m, and b from Eqs.
(23) and (25), and using r, = 0.99744 and r, = 0.56615.
We obtain y 4 by substituting x = 0.005 into the
equilibrium line to give yp+9.4 = 0.0121. Thus

r,= 0.99744

Y9.54 = 0.549
B= 0.59
A= 0.41

r2 = 0.56615

Yp+9.54 = 0.0121
m= 1.3773
b= 0.001886

These values, when substituted into Eq. (22) yield
9.91 plates below feed position.

CONCLUSIONS

The binary distillation problem considered by
Amundson was re-examined, and a simpler method
involving powers of matrices has been given and an
explicit solution obtained. This approach is suitable
for use in engineering mathematics or separation
processes courses to illustrate the application of
matrices to engineering problems.

ACKNOWLEDGEMENTS

The author is grateful to his colleague, Kevin
Free, for improvements in the clarity of this paper.

REFERENCES
1. Amundson, N., "Application of Matrices and Finite Differ-
ence Equations to Binary Distillation," Trans. AIChE, 42,
939(1946) 7

POWER OF SPREADSHEETS
Continued from page 49.

students. Insight into the relative importance of
variables and sensitivity of results to changes in the
input can be gathered from such an exercise. The
ease of changing input data also allows instructors
to efficiently check calculations made with different
combinations of independent variables.

CONCLUSIONS
Our experience with spreadsheet computing has
proved to us that it is feasible to provide instruction
on spreadsheet use as part of the mass and energy
balances class. Within a time-frame of approximately
two hours, students can learn sufficient fundamen-
tals to use spreadsheets as a tool for solving a vari-
ety of problems in the class. After solving five to
eight problems, most of them have enough confi-
dence and experience to apply the techniques in
future engineering classes.

The use of spreadsheets also encourages organi-
zation in problem solving which hopefully will carry
through to the student's non-computer work. The
flexibility and convenience of spreadsheets allows
students to solve more meaningful problems and to
examine the solutions in detail by manipulating in-
dependent variables to determine their effect. The
built-in graphics capability also helps to tie together
graphical and algebraic solution techniques when
such alternate methods exist for a given problem.

REFERENCES
1. Rosen, E.M., and R.N. Adams, "A Review of Spreadsheet
Usage in Chemical Engineering Calculations," Computers
and Chem. Engg., 11(6), 723 (1987)
2. Grulke, E.A., "Using Spreadsheets for Teaching Design,"
Chem. Eng. Ed., 20,128 (1986)
3. Dijkstra, E.W., A Discipline of Programming, Prentice-
Hall Inc., Englewood Cliffs, NJ (1976)
4. Rosen, E.M., "The Use of Lotus 1-2-3 Macros in Engineer-
ing Calculations," Chem. Eng. Ed., 24,100 (1990)
5. Aspenplus, Aspen Technology Corporation, Cambridge,
MA
6. Himmelblau, D.M., Basic Principles and Calculations in
Chemical Engineering, Prentice-Hall, Inc., Englewood
Cliffs, NJ (1982)
7. Fielder, R.M., and R.W. Rousseau, Elementary Principles
of Chemical Processes, John Wiley & Sons, New York
(1986)
8. Wankat, P.C., "What Will We Remove From the Curricu-
lum to Make Room for X?" Chem. Eng. Ed., 21, 72 (1987)
9. Lotus 1-2-3, Lotus Development Corporation, Cambridge,
MA
10. Supercalc, Computer Associates International, San Jose,
CA
11. Smartware, Informix Software Inc., Lenexa, KS
12. McKean, R.A., and E.L. Hanzevack, "The Heart of the
Matter: The Engineer's Essential One-Page Memo," Chem.
Eng. Ed., 23,102 (1989) f
Chemical Engineering Education

REVIEW: Chain Molecules
Continued from page 45.

sidechain), asymmetric vinyl chains (chains whose
repeat unit does contain a sidechain), and polypep-
tides and proteins. For each repeat unit studied, the
statistical weight matrix is derived from physical
considerations and direct computation of the confor-
mational energies. Relevant moment properties are
then computed and compared with experiment.
Chapter VIII contains a more detailed statistical
mechanical analysis of the freely jointed and other
model chains, and Chapter IX describes the theoreti-
cal background required to relate optical properties
and radiation scattering measurements to moments
of the spatial distribution.
In summary, this text contains a very complete
description of the application of the rotational iso-
meric state model. The mathematical manipulations
in Chapter I-III were found to be quite straight-
forward and followed easily from material contained
within the text. Derivations in later chapters are
not as transparent; however, references to the origi-
nal papers are very complete. The physical and chemi-
cal arguments used to derive statistical weight ma-
trices are very informative in understanding confor-
mational properties of polymers. Thus, in general
Flory does achieve his goal of a self-contained trea-
tise. He has written a clear, complete overview of the
statistical mechanics and physical basis of confor-
mations in isolated chain molecules in solution. For
researchers interested in this subject area, this book
is excellent.
However, the book may prove to be too special-
ized to attract much attention from the general chemi-
cal engineering audience. For example, the interest
of chemical engineers is often in the bulk thermody-
namic properties of polymer solutions and/or their
theological properties; this book does not touch on
either of the subjects (except obliquely by, for ex-
ample, describing methods for calculating the mean
square radius of gyration which can be related to
hydrodynamic radius).
It is therefore unlikely that the book could be
used as the text for an undergraduate or graduate
course in chemical engineering. Since the book was
written as a research monograph, it does not lend
itself to use as a textbook-for example, there are no
exercises or assignable problem sets. Faculty who
are teaching courses in applied statistical mechanics
courses may find it useful in preparing several lec-

tures on the rotational isomeric state model and its
application to real polymer chains. This would cer-
tainly serve as an extension of the material on the
statistical thermodynamics of polymers found in
typical statistical mechanical textbooks, such as D.A.
McQuarries' Statistical Thermodynamics.
In summary, the text is recommended to research-
ers interested in the physical basis and mathemati-
cal description of polymer conformations, and some
of the material in Chapters I, II, III, and V might be
suitable as part of an upper-level graduate course in
statistical mechanics. 0

books received )

Cooling Technology for Electronic Equipment, by Win Aung;
Hemisphere Publishing Co., 79 Madison Ave., New York, NY
10016-7892; 838 pages, $125 (1988)

Transport Properties of Fluids: Thermal Conductivity, Viscosity,
and Diffusion Coefficient, by Kestin and Wakeham; Hemisphere
Publishing Corp., 79 Madison Ave., New York, NY 10016-7892;
344 pages, $98 (1988)

Properties of Inorganic and Organic Fluids, by Liley, Makita,
and Tanaka; Hemisphere Publishing Corp., 79 Madison Ave.,
New York, NY 10016-7892; 309 pages, $80, (1988)

Specific Heat of Solids, by Cezairliyan; Hemisphere Publishing
Corp., 79 Madison Ave., New York, NY 10016-7892; 484 pages,
$98(1988)

Flexible Manufacturing Systems in Practice, by Roger Bonetto;
Hemisphere Publishing Corp., 79 Madison Ave., New York, NY
10016-7892; 208 pages, $37 (1988)

Standard Methods of Hydraulic Design for Power Boilers, by
Lokshin, Peterson, and Schwarz: Hemisphere Publishing Corp.,
79 Madison Ave., New York, NY 10016; 345 pages, $52.50 (1988)

Encyclopedia of Engineering Materials: Part A, Polymer Science
and Technology, edited by N. P. Cheremisinoff (Vol. 1 of 3);
Marcel Dekker, Inc., 270 Madison Ave., New York, NY 10016;
783 pages, $185 (or $157.25 each for all 3), (1988)

Natural Rubbers Science and Technology, edited by A. D. Roberts;
Oxford Science Publications, 200 Madison Ave., New York, NY
10016; 1136 pages, $150 (1988)

Adsorption and Ion Exchange: Fundamental and Applications,
edited by LeVan, Knaebel, Sircar, and Wankat; AIChE, 345 East
47th St., New York, NY 10017; $18 members, $35 non-members
(1988)

Resource Recovery of Municipal Solid Wastes, Peter J. Knox,
Editor; AIChE, 345 East 47th St., New York, NY 10017; $23
members, $45 others (1988)

Winter 1990

i1c1urriculum

USE OF PC BASED MATHEMATICS

SOFTWARE IN THE UNDERGRADUATE

CURRICULUM

JOSEPH M. SLAUGHTER, JAMES N. PETERSEN,
RICHARD L. ZOLLARS
Washington State University
Pullman, WA 99164-2710

W ith the advent of the personal computer, many
software packages have been developed which
minimize the required programming, thus allowing
the user to concentrate on understanding the prob-
lem rather than on debugging. This is very desirable
from an educational point of view. However, if stu-
dents are required to learn the use of many different
packages, any advantages gained by reducing pro-
gramming requirements are offset by increasing the
time spent learning how to use the various pack-
ages. In this article we will describe the experiences
of one of our undergraduate students in learning
and using three such programs (MathCAD ", Point
Five 121, TK Solver Plus 131) which we have examined
for incorporation into our curriculum.

Obviously, there are more packages available than
the three cited above, but we sought to find a single
package which met all of the following criteria. First,

Joseph Slaughter is a graduate student in chemical
engineering at Washington State University. He re-
ceived a BS degree in chemical engineering and a BA
degree in foreign languages and literatures, French,
from Washington State University in 1989. His current
research is in the area of bioseparations using large-
scale electrophoresis.

James N. Petersen is currently an associate professor
of chemical engineering at Washington State Univer-
sity. He received his BS degree from Montana State
University in 1976 and his PhD from Iowa State Univer-
sity in 1979. His current research interests are in the
adsorption of heavy metals from aqueous streams by
biological materials, and modeling and on-line optimi-
zation of biological processes.

it had to be a general purpose package, capable of all
the calculations (solving sets of linear and/or nonlin-
ear equations, iterative calculations, vector/matrix
manipulations, curve fitting, simple statistics, re-
gression) which are typically encountered by a stu-
dent progressing through our curriculum. For this
reason, spreadsheet programs were not included
since, although many typical problems can be solved,
the structure of the program is not optimal for all
of the calculations (solution of ODE's, sets of nonlin-
ear equations, etc.). Likewise, such packages as
Matlab i' and GAUSS 151, which concentrate on
matrix manipulations, were not included.

Second, the program had to be capable of opera-
tion on the PC's available in our department
(IBM PC/XT or AT's, or compatibles), which are
also typical of those owned by the students. Thus,
many of the symbolic manipulators (Macsyma 161)
were not considered.

Finally, we wanted the packages to be priced
within the grasp of a typical student so that they
might be purchased for use away from the academic
setting. Of the three packages evaluated, both
MathCAD and TK Solver Plus are available as stu-
dent versions for approximately $50, while Point
Five can be site licensed for a reasonable fee. There
are also a number of packages available with inter-
faces and structures very similar to TK Solver Plus,
such as Eureka 7"I and FORMULA/ONE 18'. Of this
group, however, only TK Solver Plus was evaluated,

SRichard L. Zollars is a professor of chemical engi-
neering at Washington State University. He received
his BChE (1968) from the University of Minnesota
and his MS (1972) and PhD (1974) from the Univer-
sity of Colorado. His current research interests in-
clude adsorption, colloidal and interfacial phenom-
ena, and bioseparations.

Copyright ChE Division ASEE 1991

Chemical Engineering Education

In this article we will describe the experiences of one of our undergraduate students
in learning and using three programs (MathCAD, Point Five, and TK Solver Plus) which we have
examined for incorporation into our curriculum.

due to its availability in a student version and its
greater power.

All of the packages evaluated can be run on IBM
PC/XT's or compatibles with no more than 512K,
two 5 1/4" floppy disk drives, a graphics card, and a
dot matrix printer. For this evaluation we used an
IBM PS/2 Model 50Z personal computer equipped
with a 80287 math co-processor. Feature-by-feature
comparisons of many of these software packages have
appeared [9-11, but this type of comparison does not
indicate how easily the package can be learned and
used, nor its ability to easily solve typical chemical
engineering problems and create a readable report
(such as would be required if a student used the
package to solve homework problems). Therefore,
one of our senior chemical engineering students (J.
Slaughter) was asked to solve typical homework
problems from a number of areas (thermodynamics
1121, unit operations 131, reactor design I4', kinetics 1151,
and numerical analysis 1161) using MathCAD, TK
Solver Plus, and Point Five. Each program was then
evaluated for its utility.

MATHCAD

MathCAD is a free-format scientific scratchpad
supporting 69 built-in functions and 29 symbolic op-
erators. It comes on two disks with a well-written
user's guide and a MathCAD reference booklet. Af-
ter reading the first three chapters of the user's
guide we felt fairly confident about setting up solu-
tions to typical problems. When difficulties were en-

6 Calculates the Reynolds number using
i interatic values for the fanning friction.

4.07
10
Re -

J 1 Log-log Plot
.1

.001
001 Re le*007-
Figure 1

FIGURE 1. MathCAD screen showing features such as
symbol operators and imbedded graphics.
Winter 1990

countered, help could be found either in the user's
guide or with the on-line help facility. One of the
strongest characteristics for use in education is its
format. Graphically created symbols are used in-
stead of function names, i.e., /(arg) instead of
sqrt(arg), as shown in Figure 1. Also, eighteen com-
monly used Greek letters are available for use in the
equations and text. These symbols make it easier for
the user to find errors and to create a readable
report.

The free-format style of the problem files makes
editing quite easy, much like erasing an error on
a paper scratchpad. However, moving about in
MathCAD can become tedious in large files since the
maximum cursor movement is either 80% of a page
or to the beginning or end of the current region.
Moving through the file is slowed even further if
MathCAD is in its automatic calculation mode. This
latter problem can be overcome by switching to the
manual calculation mode. MathCAD can also write
and read ASCII files so that data created from other
software can be analyzed.

MathCAD solves simultaneous equations (linear
and non-linear) using a solve block technique (see
Figure 2). This technique is initiated by guessing a
value for the unknown, entering the equation to be
solved by using a "given" command, and requesting
the solution by using a "find" command. If the con-
vergence criteria is not met, MathCAD will supply a
message to that effect, and the "find" command can
be replaced by a "minerr" command to find the result

Guess...
2
S10 -3
k = 5 10 sec
a LkoJ

giuen

D 3

n 1 find S]
k := find [h a

4
2
k Ca k' C'
a as a as

n = 3

2
3
13 I -1
k = 6.683 10 "<-- ---.------ ..
aF
Figure 2

FIGURE 2. Example of MathCAD use of a solve
block and units.

with the smallest error. The results obtained using
"minerr" are good estimates which can then be used
as new guess values.

MathCAD performs iterative calculations using
vector notation and an iteration counter (range vari-
able), as shown in Figure 3. Only one equation may
appear inside this iteration loop, but multiple func-
tions may be used. In Figure 3, for example, a user-
defined function called "gradf(s,y)" is defined by us-
ing the built-in derivative function. Other user-de-
fined functions [H(x,y), s(x,y), and k(x,y)] are subse-
quently defined using "gradf', and all of the func-
tions are combined into a single equation which is
used in the calculation.

Graphs can be created very easily in MathCAD
and can be imbedded into the report. Graphs can be
formatted for size, type (linear, semi-log, or log-log),
and number of subdivisions for each axis using
six different symbols with or without a connecting
line. One graphing feature that we found desirable
was the ability to define the x-axis with more than
one variable (neither Point Five nor TK Solver Plus
were able to do this), as shown in Figure 4. MathCAD
also creates tables, but they were limited to fifty
elements per column and it takes practice to create
presentable tables.

MathCAD supports a wide array of vector and
matrix operations and functions, although they are
limited in certain aspects. The largest array that we
were able to create using the data editor was limited
to fifty elements (newer versions can create 100-ele-
ment arrays on a 10-by-10 matrix). The absolute size
limit for arrays created by iterations and/or equa-
tions, is 8000 elements. We did not feel that this was
a significant limitation since problems larger than
this would typically not appear in classroom assign-
ments. A more significant problem occurred when
intermediate results in a calculation produced a
1-by-1 array-a value which should be used as
a scaler. MathCAD only recognizes this value as
an array so that further computations, such as the
multiplication of another array by this "scaler," were
prohibited.

Another feature that we found to be very useful
was the ability of MathCAD to use units. MathCAD
uses four base units: L (length), M (mass), T (time),
and Q (charge). Any of these can be redefined if nec-
essary (we found it convenient to redefine the base
unit for charge as a base unit for temperature). These
four base units are then used to derive other units
such as force, energy, velocity, acceleration, etc..
During calculations, MathCad converts all units into

the base units, checks for compatibility and gives the
results in the base units. The user is able to change
to any defined unit by simply entering the desired
unit. Not only does this feature provide a means of
detecting errors, but it also makes the report much
more readable (see Figure 2). MathCAD provides
three files (for mks, cgs, and US customary units)
that contain most of the desired unit conversions.

Perhaps MathCAD's strongest point is the reada-
bility of its output, as shown in Figures 1 through 4.
The graphical representation of operators, ability to
mix text, plots, and calculations, and the inclusion of
units on the calculations makes the output appear
much as it would if the user were using a paper and
pencil. Indeed, one does not need to know much
about MathCAD in order to be able to read the out-
put. The only noticeable idiosyncracy in the output is
the use of the symbols ":=" and "=". MathCAD re-

Il M TN. ]mim mZ "9 Tt

geadf(a.u) a in0
dx
geadf(xa,)
d.
MI-Y.y) :_ d
gradf(x.y)
dx I

d
- gradf(at)
as ii
d
- graddf xyy
dy I1

s.a,5) := I[H(x'y]_1 gr.df(ly)

...gradient of f(x,y)

...Hessian matrix

...direction vector

M(x.!) := -((gradf(x.y)l' S(X.5)) (s(a.(y)al H(x.y) s(xa.)) ... step sine
i 0 .10

i: :1 l -- 1 s - xiayi 1 T

...itrative calculations

Figure 3
FIGURE 3. Iterative calculation in MathCAD.

Boiling Point Diagram
120

temp .temp

70
0 Xben Ybn
i i

Figure 4

FIGURE 4. Example of boiling point diagram created by
MathCAD with multiple definition of the ordinate.
Chemical Engineering Education

quires that all variables be defined first (using the
":=" symbol) before a value can be calculated (using
the "=" symbol). This arrangement requires extra
keystrokes since any computed variable must be en-
tered twice: once to define the variable and once to
compute the value.

MathCAD has also recently become available in
a student's version. This version contains all of the
power of the full version, but is limited to only 120
lines of output (two pages). We found this limitation
on the output file size to be very restrictive for all
but the most elementary of problems.

TK SOLVER PLUS

TK Solver Plus is a structured equation solver
supporting 68 built-in functions and 13 relational
and arithmetic operators. It comes on six disks with
three manuals, a well-written (and much-needed)
reference manual, an introduction to TK Solver Plus,
and application notes. Three of the disks contain 103
models and functions which are separated into 13
categories ranging from finance to matrix manipula-
tions. As with MathCAD, a student version of TK
Solver Plus is available which contains everything
included in the full version with the exception of the
application notes and three library disks. TK Solver
Plus is also able to read and write ASCII, DIF, and
WKS files.

Of the three programs that we evaluated, we
found TK Solver Plus to be the most difficult to
learn. It was necessary to complete the introduction
manual before feeling comfortable enough to start
working. In part, this is due to the manner in which
TK Solver Plus is organized. It is divided into nine
main worksheets and several subsheets, with each
sheet containing only a specific type of information.
The variable sheet is used to enter values for the
variables and to view the results. The rule sheet

Cir) Rule: "This program finds the time needed for the batch reaction des 25+/P9
VARIABLE SEETI
St Inpt n- te 'pat -Init -- Cole. t
0 i ao Initial conversion of A
.65 X Conversion of A
Cao .00298507 in ol/ft Inital concentration of A
Cho .00268657 lbaol/ft^ Initial concentration of B
ra fan Reaction rate fraction
6 a :abher of steps used In SIMPSoN
rate 8097. 7946 Vale of ilteyal
time 24.172521 ais Tine needed for batch reaction
ROLE SHEET
probe 4-A7 in ELESMITS OF CHEMICAL REACTION EIGIIEERIIS FolgerCi986)
"It ases the library l model SIMPSON to solve the integral
.Find nit ee for Cao and Ch
Cao-=ao/V
Cto- lbo/V
"Solve integral of fan from lao to X sing Simpson's role
call Silapson(f .,Xo X, ;rate)
tine-Xao/Y*rate
Ni Help Fa Cancel F5 Edit F9 Solve / Comnands Sheets ; Window switch
Figure 5
FIGURE 5. Variable sheet and rule sheet from TK Solver
Plus using the direct solver mode.
Winter 1990

In an attempt to give some impression of the
individual strengths and weaknesses of these
various packages, we rated them on .. .ease of
learning, ease of use, matrix operations, equation
solving capability, versatility, use of units,
generation of graphs and tables, etc.

defines the rules to be used, i.e., specifies the rela-
tionships between variables, and initiates user-de-
fined functions. Other sheets are used to create user-
defined functions and algorithms, lists, tables, and
graphs, to define unit conversions, and to format the
sheets and variables.

Editing and moving about in TK Solver Plus was
not difficult. Convenient window commands are used
to move from one sheet to another, and a split screen
can be used to display two sheets at a time (see
Figure 5). Errors were marked in the status column
on each sheet, along with a brief description and an
indication of the line on which the error appeared.

The strongest attribute of TK Solver Plus is equa-
tion solving. If it cannot solve an equation directly, it
automatically goes into its integration mode. We were
able to solve twenty linear equations with twenty
unknowns, using two guesses, in a matter of sec-
onds. If TK Solver Plus could not converge using the
guesses provided, it would leave the last calculated
values as the new guess values, and we only had to
re-initiate the solve command until convergence
was reached. On those occasions where convergence
could not be obtained, TK Solver Plus would indicate
this by displaying a convergence value in a high-
lighted window and leaving the output column blank.

Another strong attribute was the capability of
TK Solver Plus to let the user create functions and
algorithms. The functions and algorithms may be
defined as rule, list, or procedural functions. Rule
functions are those which call other functions or de-
fine equations. Procedural functions create a proce-
dure which the variables follow in order to obtain a
result (this is where understanding how to program
in TK Solver Plus is important, and we found that
reviewing the models in the library disks was help-
ful). List functions are used in creating tables, inter-
polating, and displaying intermediate values during
iterative calculations.

Creating tables and graphs is done on the table
and plot sheets, respectively, and were fairly easy to
generate. The graphs and tables could each be stored
and transferred from one file to another. The only
way to print a graph, however, was to first view it

and then print it using the print screen key or com-
mand. Three types of graphs may be created: line,
pie, and bar. A line graph could produce as many
plots on one graph as lists entered, using fifteen dif-
ferent colors (for line only plots) and as many sym-
bols as can be created using a standard IBM key-
board and the approximately 249 ASCII characters,
as extended by IBM (including fifteen Greek letters).
An example of a line graph illustrating the path
taken in the optimization problem can be found in
Figure 6. Plotting two lines, each having points at
different values of the independent variable (as in
Figure 4) was quite complex using TK Solver Plus
and involved concatenating two separate lists into a
single list so that the resulting list could be plotted.

One feature that we felt could use some impro-
ement was the unit sheet and the unit column on the
variable sheet. TK Solver Plus does not recognize
units either in an equation or while performing cal-
culations and so will not give a warning of incompat-
ible units, as MathCAD does. On a few occasions we
entered incorrect units but did not find our mistakes
until after printing the report.

TK Solver Plus does not contain matrix opera-
tions within its basic package, although they are
provided on the library disks or by creating your
own algorithms. This is due to the algorithms TK
Solver Plus uses when solving sets of equations which
do not rely on matrix operations. In those situations
where matrix operations were desired, we found that
the calculations could be performed, but were gener-
ally more trouble than they were worth.

Printed output from TK Solver Plus is obtained
by separately printing each sheet, graph, and/or table.
We found these reports difficult to read and follow,
especially for the beginner or non-user. Comment
statements, as shown on Figure 5, did help the read-
ability, but the reader must be familiar with TK
Solver Plus to understand and follow the report as a
whole.

POINT FIVE

Point Five is a mathematical scratchpad contain-
ing over 150 built-in functions divided into six cate-
gories: arithmetic, statistics, finance, matrix opera-
tions, data transformation, and procedure functions.
This wide range of functions indicates that it is
meant to be used by both engineering and non-engi-
neering students alike. Point Five comes with one
systems disk (or can be installed with two master
disks), a well-written user's guide, an introductory
tutorial manual, and a quick reference card. The tu-

trial was easy to follow and took only about one
hour to complete. After completing the tutorial, we
felt comfortable with the scratchpad and had very
little trouble creating the desired models. Point Five
has very nicely- organized help files for quick, on-
screen help. Like the other programs, Point Five is
able to read and write ASCII and DIF files, thus
permitting it to interact with other software.

Point Five consists of two windows on one split
screen (see Figure 7). The lower window is the
scratchpad where formulas are entered and edited.
The upper window is where the results are scrolled.
Variables can be entered and edited either directly
on the scratchpad or by using the data editor. Using
the data editor makes analyzing and manipulating
data quite easy; one simply enters a set of data,
computes the results, and then enters a new set of
data without having to change the scratchpad.

The two best features of Point Five are the built-
in functions and the matrix operations. The variety
of functions helps limit the amount of programming

Newton' s Method
.4

.2

-.2

-.6
-.8

-1.2
1 1.2 1.4 1.6 1.8 2 2.2
XI

Figure 6

FIGURE 6. Line graph created by TK Solver Plus for
optimization problem.

04 -n=rovs (nil). maietable(plot.n), addtotable(plot,xi,x2)
041--plot
Xi X2
oo. 0. 00
1. 33 -0. 50
1. 67 -0, 86
i. 98 -i 06
2. 00 -1, 00
2. 00 -1. 00
-or Educ.tional Vse Only at g-shington State Unlv.-
001 \Perforing a manual iteration technique for finding a miAMniUl sing
002o \Mleton' Ltiod.
0031 \Exaa ple 6,5 niglln [(i. Z2)-(l-2)^l4+(ol-2)^2*x2^2+(x2+i)^2
0041 1-0
005 xl[li-l. x2[i]-O initiali val. es for xl and x2
006
007 ;Calculate the Partial derivative to find te gradient.
0081
0091 i=i+1
0 01 .ntrad[l]-4l (x[l ]-2)32*x2[i]2(.l [i] -2)
xe65 Kem,451/151 Replace InaPad Caps-OF Disp-0I Calc-I Prnt-OFF
Figure 7
FIGURE 7. Screen from Point Five illustrating an
interactive calculation.

Chemical Engineering Education

required to create various models, which makes
working with Point Five easier for the beginner. We
found the matrix functions to be the best of the three
software packages that we evaluated. We were able
to enter an 89-by-89 matrix with the data editor,
invert it, and then check the results by multiplying
the inverted matrix by the original matrix, all on one
line. The results were quite accurate.
We were also able to solve simultaneous linear
equations, such as the optimization problem, by us-
ing an augmented matrix and one function,
SIMEQ(var.). Graphs and tables are quickly and
easily created in Point Five. The graphing abilities
are limited, however, to only scatter plots, line graphs,
and bar graphs. Line graphs are limited to six lines,
and the graphics are only CGA resolution (see Fig-
ure 8). Also, plots cannot be stored in separate files
and can only be printed by doing a screen print.

Unlike TK Solver Plus and MathCAD, Point Five
is not an equation solver; it is more like a powerful
programmable calculator. This difference is particu-
larly noticeable in that Point Five requires that all
equations be in their explicit form. For example, we
were not able to solve for the Fanning friction, f, in
the equation
1 / = 4.07 log(NRe Vf) 0.6

used in Figure 1 since it could not be reduced to an
explicit form.

Iterative calculations are performed in two ways:
by line, using a "FOR...DO..." statement, or in a
marked block by using the EXECUTE N TIMES
command. Point Five limits the use of marked blocks
to one per file, thus limiting the use of the EXE-
CUTE command.
I Newtons Method of Optimization

Figure 8
FIGURE 8. Line graph created by Point Five for optimiza-
tion problem.
Winter 1990

Readable reports could be made in Point Five
with the use of comment statements. Turning the
DISPLAY function off permits only comment state-
ments, blank lines in a model, and results to be
printed. Turning DISPLAY on prints each line exe-
cuted. When using a marked block to perform itera-
tive calculations, with DISPLAY on, long reports
may be generated as each result computed during
every iteration is displayed. We were able to avoid
this problem by entering each line once to obtain a
printout of the procedure and then executing the
block, but this was only working around a weakness
in Point Five. Unlike either MathCAD or TK Solver
Plus, Point Five has no capability for working with
units.

DISCUSSION AND CONCLUSION

Of the five example problems we selected, both
MathCAD and TK Solver Plus were able to solve all
five. Point Five was able to solve four of the five, but
could not solve the friction factor problem due to the
implicit equation used. TK Solver Plus appeared to
be the fastest of the three packages, although all
were rapid enough for the example problems so as
not to be an inconvenience.

Comparing the three packages is very difficult
since they are completely different, not only in struc-
ture but in emphasis of capabilities. Keeping in mind
that these programs were evaluated for use in an
engineering curriculum, we found Point Five to be a
very useful tool for performing quick, on-the-spot
calculations and data manipulations. Its library of
functions was impressive and very useful for creat-
ing educational models and statistical and financial
analysis. It was very easy to use and a good tool for
those who have little experience with personal com-
puters, but the limitations encountered for non-lin-
ear equation solving reduced its utility.
TK Solver Plus is a very powerful equation solver
with the capability of a programming language. It is
probably the most powerful of the three programs,
but is also the hardest to learn. For quick, on-the-
spot calculations (especially matrix operations) it is
almost not worth the trouble. However, once the
user becomes familiar with TK Solver Plus, its abili-
ties are practically unlimited. Its rigid structure
makes learning the program somewhat easier, but it
also makes printed reports hard to read.
We found that we preferred going to MathCAD
when attempting to solve a problem because of
MathCAD's versatility, its ease of use, and the read-
59

0.3

0.0

c4 -0.3

-0.6

-0.9

1.2 1.5 1.8 2.1 2.4

Figure 9
FIGURE 9. Comparison of Point Five, TK Solver Plus,
and MathCAD. Scale: 0=unsatisfactory, 1 =poor, 2=fair,
3=good. 4=excellent.
ability of its reports. We were able to efficiently per-
form matrix operations, create user functions, and
solve series of simultaneous equations, without hav-
ing to know many programming techniques. We en-
joyed the feature of creating and editing graphs with-
out having to leave the scratchpad every time we
wanted to examine a plot a feature especially help-
ful when trying to find roots. But one of the most
impressive features is that the printed results from
MathCAD are very readable. The free-format use of
units and graphic abilities make it easier for the
user and non-user to read and understand a
MathCAD file.
In an attempt to give some impression of the in-
dividual strengths and weaknesses of these various
packages, we rated them on the accompanying docu-
mentation, ease of learning, ease of use, matrix op-
erations, equation solving capability, versatility, use
of units, generation of graphs, generation of tables,
readability of output, and overall impression on a
scale from 0 (unsatisfactory) to 4 (excellent). While
each of the programs that we evaluated has particu-
lar strengths eee and weaknesses, as indicated in Fig-
ure 9, MathCAD was our selection as the program
best suited for general use in the chemical engineer-
ing curriculum.
ACKNOWLEDGEMENT
This work was supported by a Teaching Develop-
ment Grant from the College of Engineering and Architec-
ture, Washington State University.
REFERENCES
1. MathCAD, MathSoft, Inc., One Kendell Square, Building
200, Cambridge, MA 02139
2. Point Five, Pacific Coast Software Inc., 887 NW Grant Ave.,
Corvallis, OR 97330
3. TK Solver Plus, Universal Technical Systems, Inc., 1220
60

Rock Street, Rockford, IL 61101
4. Matlab, The Math Works, Inc., 20 N. Main Street, Suite 250,
Sherborn, MA 01770
5. GAUSS, Aptech Systems, Inc., 26250 196th Place SE, Kent,
WA 98042
6. Macsyma, Symbolics Inc., 8 New England Executive Park
East, Burlington, MA 01803
7. Eureka, Borland International, 4585 Scotts Valley Drive,
Scotts Valley, CA 95066
8. FORMULA/ONE, Soft-Sense, 12 Rockaway Lane, Arlington,
MA 02174
9. Shacham, M., and M.B. Cutlip, Chem. Eng. Ed., 22(1), 18
(1988)
10. Heller, M., Personal Engineering & Instrumentation News,
5(9), 39 (Sept. 1988)
11. Smith, A.L.,Academic Computing, 2(3), 36 (1987)
12. Himmelblau, D.M., Basic Principles and Calculations in
Chemical Engineering, 4th ed., Prentice-Hall, Inc., Engle-
wood Cliffs, NJ, p. 591 (1982)
13. McCabe, W.L., J.C. Smith, and P. Harriott, Unit Operations
of Chemical Engineering, McGraw-Hill Book Company, New
York, p. 85 (1985)
14. Fogler, H.S., Elements of Chemical Reaction Engineering,
Prentice-Hall, Inc., Englewood Cliffs, NJ, p. 164 (1986)
15. Fogler, H.S., Elements of Chemical Reaction Engineering,
Prentice-Hall, Inc., Englewood Cliffs, NJ, p. 572 (1986)
16. Edgar, T.F., and D.M. Himmelblau, Optimization of Chemi-
cal Processes, McGraw-Hill Book Company, New York, p.
208(1988) 1I

POLYMERIZATION PILOT-PLANT
Continued from page 39.
tive. Our efforts reflect our belief that it is time to
introduce more sophisticated control algorithms
(which, at present, are usually encountered at the
graduate-student level) at the undergraduate level,
combined with more complicated real processes.
ACKNOWLEDGEMENTS
Financial support from the Natural Sciences and En-
gineering Research Council (NSERC) of Canada, the Insti-
tute of Polymer Research (IPR), University of Waterloo,
the Manufacturing Research Corporation of Ontario
(MRCO), and the Computer-Aided Process Engineering
(CAPE) Laboratory at the University of Waterloo, is grate-
fully acknowledged.

Many thanks also go to the "brave students": Mike
Nakagawa, Gary Jubien, Mel deSouza, Howard Leung,
Jerry Lin, Cam Phan, and Mike Kuindersma.
REFERENCES
1. Stephanopoulos, G., Chemical Process Control: An Introduction to
Theory and Practice, Prentice Hall, New Jersey (1984)
2. Hoogendoorn, K., and R. Shaw, "Control of Polymerization Processes,"
Proc. IFAC PRP-4 Automation, Chent Belgium, Pergamon Press, p
623 (1980)
3. MacGregor, J.F., A. Penlidis, and A.E. Hamielec, "Control of Polymeri-
zation Reactors: A Review," Poly. Proc. Eng., 2, 179 (1984)
4. Richards, J.R., and P.D. Schnelle, Jr., "Perspectives on Industrial Re-
actor Control," Chem. Eng. Prog., 84, 32 (1988)
5. Davidson, R.S., "An Intelligent Temperature Controller for Jacketed
Reactors," Am. Cont. Conf., 2, 1380; CEP, 82, 18 (1986)
6. Kern, A.G., "Simplify Batch Temperature Control," Chem. Eng., pg 61,
March 28 (1988) 1
Chemical Engineering Education

AUTHOR GUIDELINES

This guide is offered to aid authors in preparing manuscripts for Chemical Engineering Education
(CEE), a quarterly journal published by the Chemical Engineering Division of the American Society for
Engineering Education (ASEE).

CEE publishes papers in the broad field of chemical engineering education. Papers generally describe
a course, a laboratory, a ChE department, a ChE educator, a ChE curriculum, research program,
machine computation, special instructional programs, or give views and opinions on various topics of
interest to the profession.

Specific suggestions on preparing papers.

TITLE Use specific and informative titles. They should be as brief as possible, consistent
with the need for defining the subject area covered by the paper.
AUTHORSHIP e Be consistent in authorship designation. Use first name, second initial, and
surname. Give complete mailing address of place where work was conducted. If current
address is different, include it in a footnote on title page.
TEXT e Manuscripts of less than twelve double-spaced typewritten pages in length will be
given priority over longer ones. Consult recent issues for general style. Assume your
reader is not a novice in the field. Include only as much history as is needed to provide
background for the particular material covered in your paper. Sectionalize the article
and insert brief appropriate headings.
TABLES Avoid tables and graphs which involve duplication or superfluous data. If you
can use a graph, do not include a table. If the reader needs the table, omit the graph.
Substitute a few typical results for lengthy tables when practical. Avoid computer
printouts.
NOMENCLATURE Follow nomenclature style of Chemical Abstracts; avoid trivial names.
If trade names are used, define at point of first use. Trade names should carry an initial
capital only, with no accompanying footnote. Use consistent units of measurement and
give dimensions for all terms. Write all equations and formulas clearly, and number
important equations consecutively.
ACKNOWLEDGMENT Include in acknowledgment only such credits as are essential
LITERATURE CITED References should be numbered and listed on a separate sheet in the
order occurring in the text.
COPY REQUIREMENTS Send two legible copies of the typed (double-spaced) manuscript
on standard letter-size paper. Clear duplicated copies are acceptable. Submit original
drawings (or clear prints) of graphs and diagrams, and clear glossy prints of
photographs. Prepare original drawings on tracing paper or high quality paper; use
black india ink and a lettering set. Choose graph papers with blue cross-sectional lines;
other colors interfere with good reproduction. Label ordinates and abscissas of graphs
along the axes and outside the graph proper. Figure captions and legends may be set in
type and need not be lettered on the drawings. Number all illustrations consecutively.
Supply all captions and legends typed on a separate page. If drawings are mailed under
separate cover, identify by name of author and title of manuscript. State in cover letter
if drawings or photographs are to be returned. Authors should include brief
biographical sketches and recent photographs with the manuscript.

Do You Qualif for Product Development
in the U.S.A. orInternational?

u-rn-
r~iw47

CHEMICAL ENGINEERS

...The World is Yours!

...iEl Mundo es Tuyo!

...Le Monde est a Vous!

...Die Welt ist Dein!

*091 %mWmrttft0)*/

Join Us
Career!

and Enjoy an Exciting

Procter & Gamble has several entry-level product
and process development openings for BS and MS
Chemical Engineers in Asia, Europe, Mexico, South
America and the U.S.A.
To readily qualify for an international
assignment you must be bilingual (including
English) possess appropriate Citizenship,
Immigration Visa, or Work Permit from one
or more of the following countries:
Austria, Belgium, Brazil, Chile, Colombia,
Denmark, Egypt, France, Germany, Hong
Kong, India, Ireland, Italy, Japan, Lebanon,
Malaysia, Mexico, Netherlands, Peru,
Philippines, Portugal, PuertoRico, Saudi
Arabia, Singapore, Spain, Taiwan, Turkey,
UnitedKingdom and Venezuela.

Procter & Gamble total sales are over 23 billion dollars
world-wide. Major product categories include beauty
care, beverage, detergent, fabric care, food, health care,
household care, paper, and pharmaceutical consumer
products. Our technically-based corporation spent nearly
700 million dollars in research and product development
last year.
We offer a stimulating environment for personal and
professional growth, highly competitive salaries, and
excellent benefits package including pension, health
care and paid relocation.
If interested, send your resume, including country
qualifications and language fluencies, to:
F.O. Schulz, Jr.
U.S.A. & International Ch E Openings
The Procter & Gamble Company
Ivorydale Technical Center (#6CEE)
5299 Spring Grove Ln.
Cincinnati, OH 45217

l PROCTER &GAMBLE
SAn Equal Opportunity Employer

	Front Cover
	Table of Contents
	Louisiana State University: Development...
	Jim Stice of the University of...
	A membrane gas separation experiment...
	An engineering applications laboratory...
	Positions available
	Engineering education verses
	A "user-friendly" program for vapor-liquid...
	Teaching effective oral presentations...
	Use of a modern polymerization...
	A robust alternate to least sum...
	Book review
	The power of spreadsheets in a...
	Amundson's matrix mthod for binary...
	Books received
	Use of PC based mathematics software...
	Back Cover

MILLISECOND	CLASS.METHOD	MESSAGE
0	sobekcm_page_globals.constructor
0	sobekcm_page_globals.constructor	Application State validated or built
0	sobekcm_database.verify_item_lookup_object
0	sobekcm_page_globals.constructor	Navigation Object created from URI query string
0	sobekcm_database.verify_item_lookup_object
0	sobekcm_page_globals.display_item	Retrieving item or group information
0	sobekcm_page_globals.get_entire_collection_hierarchy	Retrieving hierarchy information
0	sobekcm_assistant.get_entire_collection_hierarchy
0	cached_data_manager.retrieve_item_aggregation
0	cached_data_manager.retrieve_item_aggregation	Found item aggregation on local cache
0	item_aggregation_builder.get_item_aggregation	Found 'all' item aggregation in cache
0	system.web.ui.page.page_load (ufdc.page_load)
0	sobekcm_page_globals.constructor.on_page_load
0	html_echo_mainwriter.add_style_references	Adding style references to HTML
0	html_echo_mainwriter.add_text_to_page	Reading the text from the file and echoing back to the output stream
60	html_echo_mainwriter.add_text_to_page	Finished reading and writing the file


UFDC Home	<%MYSOBEK%> \| Help \| RSS