|
Items in cart [0] |
Questions? Call 888-624-8373 |
| Copyright © 2009. National Academy of Sciences. All rights reserved. Terms of Use and Privacy Statement |
Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter.
Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 135
OCR for page 136
136
_l MAGINE A ROOM 50 feet long and 30
feet wide, filled with cabinets contain-
_ - ing 18,000 vacuum tubes intercon-
nected by miles of copper wire. This was the
ENIAC (Figure 8.1), a 30-ton behemoth that
was the world's first electronic computer. Now
imagine that ENIAC had been given the task
of solving a set of simultaneous linear equations
that embodied 30,000 independent variables.
With a team of people working around the clock
to record intermediate solutions and feed them
back into the computer (ENIAC had a limited
amount of memory), ENIAC would still be
chugging away more than 40 years later to solve
the problem! A supercomputer using modern
algorithms could solve the problem in an hour
or so.
The speed and capability of the modern com-
puter have tremendous implications for the
practice of chemical engineering. In the future,
computer programs incorporating artificial in-
telligence or expert systems will help engineers
design improved chemical processes more ef-
ficiently. Complex computations based on fun-
damental engineering knowledge will allow en-
gineers to design reactors that can virtually
eliminate undesired by-products, making pro-
cesses less complex and less pol-
luting. New sensors, many of
which will be miniature analyti-
cal laboratories tied to miniature
electronics, will allow rapid and
accurate measurements for con-
trol that are currently impossi-
ble. New chemical products that
today are discovered predomi-
nantly through laboratory work
for instance, reinforced plastics
that are as strong as steel and
weigh less than aluminum or drugs
with miraculous properties may
be discovered in the future by
computer calculations based on
models that predict the detailed
behavior of molecules.
Chemical engineers will lead
this revolution. They will need
to be trained to use advanced
computer techniques for process
design, process control, and
FRONTIERS I.\ CHEMICAL ENGINEERING
management of process information. Advanced
engineering development will be based more
than ever on mathematical modeling and sci-
entific computation. Reliable modeling at the
microscale, the individual process unit scale,
and the plant process scale will improve our
ability to scale up processes in a few large steps,
possibly bypassing the need for a pilot plant
and saving the 2 or 3 years required to build
and operate it. Process models capable of pre-
dicting dynamic behavior, operability, flexibil-
ity, and potential safety problems will permit
these aspects of a process to be considered
more fully earlier in the design stage. Because
improved computers can perform the extensive
computations required by such models, it will
be possible to test alternative designs more
quickly.
A chemical process must be designed to
operate under a chosen set of conditions, each
of which must be controlled within specified
limits if the process is to operate reliably and
yield a product of specified quality. Accurate,
complex, computer-solvable models of chemical
processes will incorporate features of the con-
trols that are needed to maintain the desired
process conditions. Such models will be able to
FIGURE 8.1 The Electronic Numerical Integrator and Calculator (ENIAC)
and its inventor, J. Presper Eckert, circa 1946. ENIAC was the world's first
electronic computer. Courtesy, UNISYS Corporation.
OCR for page 137
COMPUTER-ASSISTED PROCESS AND C0.~L ENGINEERING
predict the effects of process excursions and
the control measures needed to correct them.
Computer management of the process operation
will rapidly initiate control correction of process
excursions. The development of new types of
process sensors will be essential to this degree
of process control and will eliminate the time-
consuming withdrawal of process stream sam-
ples for analysis.
The design of a commercial process can
generate an almost uncountable number of pos-
sible solutions for seemingly simple problems.
Even a decade's 10,000-fold increase in com-
putational capability in terms of faster computer
speeds and better algorithms does not permit a
person to search among these alternatives, nor
would such a search make strategic sense. Give
engineers a design problem for which the best
solution is obvious from today's technology,
and they will quickly write down the correct
solution without searching. If the solution is not
obvious, they will often home in on the infor-
mation needed and perform the computations
that will expose the right solution quickly and
with minimal effort. By using intuition and
experience, they eliminate the need for testing
every possible alternative. We need to under-
stand how to use computer technology in much
the same way, to solve complex problems where
many of the decisions are based on qualitative
information and insights that develop as the
problem is attacked.
Encoding this activity in the computer in-
volves a type of modeling in which the capa-
bilities of the designer and his tools, the alter-
native procedures by which complex problem
solving can be performed, and effective methods
of information management are all incorporated.
Advances in artificial intelligence, expert sys-
tems, and information management will revo-
lutionize the automation of this activity, giving
us computers that can display encyclopedic
recall of relevant information and nearly human
reasoning capabilities. A HAL 9000 of 2001: A
Space Odyssey fame may indeed exist in the
future.
Computer-generated visual information, for
example, three-dimensional portrayal of pro-
posed new designs, will be commonplace in the
future. Communication will be in natural lan
137
guage, using both pictures and voice. This
setting, which will address the need for new
chemical processes and products by harnessing
almost unimaginable computing power, will pro-
vide significant new research opportunities in
chemical engineering.
USING THE COMPUTER'S POTENTIAL
Each decade over the last 35 years has seen
the processing speed of newly designed com-
puters increase by a factor of about 100 owing
to advances in the design of electronic micro-
circuits and other computer hardware. On top
of this has been another 100-fold increase per
decade in computer speed, thanks to more
efficient methods of carrying out computations
(algorithms). It is not widely appreciated that
new algorithms have been as valuable as hard-
ware design in improving computer perform-
ance. With the combination of improved com-
puter hardware and better algorithms, effective
computer speeds have more than doubled on
average each year.
The availability of computing resources is
also increasing rapidly. The actual and projected
availability of high-speed supercomputers is
shown in Table 8.1. The projection for 1990-
at least 700 computers of Cray I class repre-
sents 35 times more available computing power
than that available in 1980. While continued
substantial investment in supercomputers is
needed, support for better ways of using them
should not be neglected. It is conceivable that
a new algorithm could effectively increase the
power of supercomputing for a specific problem
by a factor of 35 overnight. During the last two
decades, many developments in numerical anal
TABLE 8.1 Actual and Estimated
Supercomputing Resources Available
to Researchers in the United States,
1980-1990
Year
Number of Cray I-CIass
Supercomputers
21
142
700-1,000
980
985
990
OCR for page 138
FRONTIERS IN' CHEMICAL ENGINEERI~G
ysis have had a profound impact
on scientific computation.
Clearly computer technology
has improved rapidly; there is
little reason to doubt that it will
continue to do so. The problem
is now and will continue to be
the lack of people trained to ap-
ply computer technology to sci-
entific and engineering tasks. The
improvements suggested in
Chapter 7 in terms of our ability
to design and control better
chemical products and processes
will be made by chemical engi-
neers who understand comput-
ers- not by computer scientists
or by software engineers. The
countries that understand this
distinction will lead the world in
chemical technology.
MATHEMATICAL MODELS
OF FUNDAMENTAL
PHENOMENA
Chemical engineers have tra-
ditionally used mathematical
models to characterize the phys-
ical and chemical interactions oc-
curring in chemical processes.
Many of these models either have
been entirely empirical or have
relied on crude approximations
of the basic physics or chemistry
of the process. This is because a typical chemical
process comprises an assemblage of interacting
Ho`vs, transports, and chemical reactions. Ac-
curate analysis and prediction of the behavior
of . such a complex system require detailed
portrayal of the physics of transport and the
chemistry of reactions, which calls for complex
equations that do not yield to traditional math-
ematics. Nonlinear partial differential and in-
tegral equations in two and sometimes three
spatial variables must be solved for regions with
complicated shapes that often have at least some
free boundary. The more accurate the model,
the more mathematically complex it becomes,
but it cannot be more complex than allowed for
by the available methods for solving its equa-
tions.
Before the advent of modern computer-aided
mathematics, most mathematical models of real
chemical processes were so idealized that they
had severely limited utility being reduced to
one dimension and a few variables, or linearized,
or limited to simplified variability of parameters.
The increased availability of supercomputers
along with progress in computational mathe-
matics and numerical functional analysis is rev-
olutionizing the way in which chemical engi-
neers approach the theory and engineering of
chemical processes. The means are at hand to
model process physics and chemistry from the
OCR for page 139
C0MPU~ER-ASSISTED PROCESS AND CONTROL E\GI.~ERI~G
molecular scale to the plant scale; to construct
models that incorporate all relevant phenomena
of a process; and to design, control, and opti-
mize more on the basis of computed theoretical
predictions and less on empiricism. Chemical
engineers, using advanced computational meth-
ods and supercomputers, can now readily iden-
tify the important phenomena in a complex
chemical process over the entire range of ap-
plicable conditions by exhaustive solution of
detailed models. The benefits of investing in
less empirical, more fundamental mathematical
models are becoming clear:
.
The capability to construct mathematical
models that more fully incorpo-
rate the basic chemistry and
physics of a system provides a
mechanism for assessing under-
standing of fundamental phe-
nomena in a system by compar-
ing predictions made by the model
with experimental data.
~ Better models can replace
laboratory or field tests that are
difficult or costly to perform or
identify crucial experiments that
should be carried out. In either
case, they will significantly en-
hance the scope and productivity
of chemical engineering re-
searchers in academia and in-
dustry.
~ In process design, it is fre-
quently discovered that many of
the basic data needed to under-
stand a process are lacking. Be-
cause most current mathematical
models are not sufficiently ac-
curate to permit direct scale-up
of the process from laboratory
data to full plant size, a pilot
plant must be constructed. As
models are improved, it may be-
come possible to evaluate design
decisions with more confidence,
and bypass the pilot plant stage.
Process technologies for which
the use of more comprehensive
mathematical models can result in major im-
provements include those for biochemical re-
action processing; high-performance polymers,
plastics, composites, and ceramics; chemical
reaction processing (e.g., reaction injection
molding, reaction coating, chemical vapor dep-
osition); microelectronic circuits; optical fibers
and disks; magnetic memory systems; high-
speed coating; photovoltaic and semiconductor
materials; coal gasification; enhanced petroleum
recovery; solution mining; and hazardous waste
disposal. To date the most extensive use of
supercomputer modeling has been in space age
weapons technologies, where objectives, eco-
nomics, and time frames differ from those in
OCR for page 140
/
- - -
the chemical process indus-
tries. It is clearly in the national
interest to stimulate the more
extensive use of advanced com-
putational methods and super-
computers in other industries
critical to our worldwide com-
petitive position. A program to
encourage the greater dissemi-
nation of advanced computa-
tional techniques and hardware
will offer challenges and oppor-
tunities to computational math-
ematicians and numerical ana-
lysts, to engineering scientists,
to applications and software
experts in firms that develop
and manufacture supercompu-
ters, and, above all, to perceptive
leaders in high-technology pro-
cess industries.
The following sections de-
scribe in more detail a number
of areas in chemical engineering
in which the ability to develop
and apply detailed mathematical
models should yield substantial
rewards.
Hydrodynamic Systems
Much of the current compu-
tational modeling research in chemical engi-
neering is concerned.with the behavior of flow-
ing fluids. The general system of equations that
describe fluid mechanics, called the Navier-
Stokes equations, has been known for more
than 100 years, but for complex phenomena the
equations are exceedingly difficult to solve.
Only recently have methods been devised to
treat such phenomena as shock waves and
turbulence. Further difficulties arise when dis-
parate temporal and spatial scales are present
and when chemical reactions occur in the fluid.
Solutions of the Navier-Stokes equations can
be smooth and steady, or they can exhibit
regular oscillations or even chaos. In some cases
the fluid flow is enclosed by a rigid boundary
with a complex shape, as in the extrusion of
polymers; in others the flow is effectively un
bounded and the solution must extend to infin-
ity, as in atmospheric systems; and in still
others, such as the flow of blood in vessels, the
boundary is deformable. Solution of the Navier-
Stokes equations for systems of technological
interest remains an exceedingly challenging task;
supercomputers are needed to treat those sys-
tems that can be solved.
Polymer Processing
The development of polymers and polymer
composites will benefit greatly from the avail-
ability of better computers and better algo-
rithms. The inherent properties of a polymer
are governed by the chemical structure of its
molecules, but the properties of a finished poly-
mer product are affected by the interactions
OCR for page 141
among these molecules, which are strongly influ-
enced by the way in which the material has
been processed. While it is now possible to
predict certain properties of polymers from their
molecular structure, the ability to predict the
effect of polymer processing steps on polymer
properties is just being developed. Ideally it
would be desirable to model all steps from the
formation of the polymer through its processing
and then predict the final properties of the
material from structure-property relationships.
Although such modeling is a formidable prob-
lem, it is becoming feasible with the advent of
supercomputers and improved algorithms.
Petroleum Production
Computation is widely used in petroleum
exploration and production by exploration geo
~Air ~
,:.
physicists, petroleum engineers,
and chemical engineers. As more
sophisticated techniques are de-
veloped for locating and recover-
ing petroleum, mathematical
modeling is playing an ever-ex-
panding role.
Once regions that may contain
petroleum are located, local geo-
logical features must be sought
that might have trapped the hy-
drocarbons. The discovery pro-
cess is based on a kind of seismic
prospecting in which geologic
maps are constructed from re-
flected seismic signals generated
by explosions or vibrations at
the surface of the earth. These
signals are reflected or refracted
in varying degrees by different
rock strata and are recorded by
a set of receivers. Thus, the prob-
lem of interpreting signals can be
likened to that of analyzing light
beams reflected by an array of
variously curved plates of glass
of different reflectivities sepa-
rated by liquids of different re-
fractive indexes. The inverse
mathematical problem of deter-
mining the earth structure and
the properties of the strata from
the recorded signals is extremely difficult.
After a hydrocarbon reservoir has been lo-
cated, the flow of oil, water, gas, and possibly
injected chemicals in the reservoir must be
modeled. This challenge is particularly appro-
priate for chemical engineers working with pe-
troleum engineers because of the important role
played by molecular level interactions between
oil, subsurface water, and rock. Models for fluid
flow in porous media comprise coupled systems
of nonlinear partial differential equations for
conservation of mass and energy, equations of
state, and other constraining relationships. These
models are usually defined on irregular domains
with complex boundary conditions. Their nu-
merical solution, with attendant difficulties such
as choice of discretization methods and grid or-
ientation, is a challenging intellectual problem.
OCR for page 142
Once wells have been drilled into the for-
mation, the local properties of the reservoir
rocks and fluids can be determined. To construct
a realistic model of the reservoir, its properties
over its total extent- not just at the well sites-
must be known. One way of estimating these
properties is to match production histories at
the wells with those predicted by the reservoir
model. This is a classic ill-posed inverse prob-
lem that is very difficult to solve.
When or if the reservoir is successfully sim-
ulated, the engineer can turn to optimizing
petroleum recovery, and theoretical ideas can
be applied to models for various enhanced
recovery methods to select optimal procedures
and schedules (see Chapter 71.
Combustion Systems
Combustion is one of the oldest and most
basic chemical processes (Plate 61. Its accurate
mathematical modeling can help avoid explo-
sions and catastrophic fires, promote more ef-
ficient fuel use, minimize pollutant formation,
and design systems for the incineration of toxic
materials (see Chapter 81. For example, mod-
eling the initiation and propagation of fires,
explosions, and detonations requires the ability
to model combustion phenomena. Models of
the internal combustion engine can shed light
on the influence of combustion chamber shape
or valve and spark plug placement on engine
performance. Models at the molecular level can
provide a fundamental understanding of how
fuels are burned and how gaseous and particu-
late pollutants are formed. This can lead to
ways to improve the design of combustion
systems.
Mathematical models of combustion must
incorporate intricate fluid mechanics coupled
with the kinetics of many chemical reactions
among a multitude of compounds and free
radicals. They must also consider that those
reactions are taking place in turbulent flows
inside chambers with complex shapes. Because
complete models of real combustors, incorpo-
rating accurate treatment of both fluid mechan-
ics and chemistry, are still too large for present
computers, the challenge is to construct simpler,
yet still valid, models by using critical insight
FRO.\TEERS I^Y CHEMICAL ENGI.~G
into the important chemical and physical phe-
nomena found in combustors. Chemical engi-
neers have the mix of expertise necessary to
accomplish this.
Environmental Systems
The environment can be likened to a giant
chemical reactor. Gases and particles are emit-
ted into the atmosphere by industrial and other
man-made processes, as well as by a variety of
natural processes such as photosynthesis, vul-
canism, wildfires, and decay processes. These
gases and particles can undergo chemical re-
actions, and they or their reaction products can
be transported by the wind, mixed by atmos-
pheric turbulence, and absorbed by water drop-
lets. Ultimately, they either remain in the at-
mosphere indefinitely or reach the earth's surface.
For example, the hazes of polluted atmospheres
consist of submicron aerosols of inorganic and
organic compounds, which are formed by chem-
ical reaction, homogeneous nucleation, or con-
densation of gases (Plate 71.
Models of atmospheric phenomena are similar
to those of combustion and involve the coupling
of exceedingly complex chemistry and physics
with three-dimensional hydrodynamics. The
distribution and transport of chemicals intro-
duced into groundwater also involve a coupling
of chemical reactions and transports through
porous solid media. The development of ground-
water models is critical to understanding the
effects of land disposal of toxic waste (see
Chapter 71.
PROCESS DESIGN
The primary goal of process design is to
identify the optimal equipment units, the optimal
connections between them, and the optimal
conditions for operating them to deliver desired
product yields at the lowest cost, using safe
process paths, and with minimal adverse impact
to the environment. Design is a complex prob-
lem that involves not only the quantitative
computing depicted in the previous section, but
also the effective handling of massive amounts
of information and qualitative reasoning.
OCR for page 143
CO~PUTER-ASSISTED PROCESS AVID CO1YTROL ENGINEERING
Computer-Assisted Design of New Processes
Designs for new processes proceed through
at least three stages:
· Conceptual design the generation of ideas
for new processes (process synthesis) and their
translation into an initial design. This stage
includes preliminary cost estimates to assess
the potential profitability of the process, as well
as analyses of process safety and environmental
considerations.
· Final design a rigorous set of design cal-
culations to specify all the significant details of
a process.
· Detailed design preparation of engineer-
ing drawings and equipment lists needed for
construction.
The key step in the conceptual design of a
new chemical manufacturing process is gener-
ating the process flowsheet (Figure 8.24. All
other elements of computer-aided design (e.g.,
process simulation, design of control systems,
and plantwide integration of processes) come
into play after the flowsheet has been estab-
lished. In current practice, the pressure to enter
the market quickly often allows for the explo-
ration of only a few of the process alternatives
that should be considered. To be fair to today's
designers, it is possible to generate a very large
number of alternative process paths at the
conceptual stage of design, and yet experience
indicates that less than 1 percent of the ideas
for new designs become commercial. Thus, the
challenge in computer-aided process synthesis
is to develop systematic procedures for the
generation and quick screening of many process
alternatives. The goal is to simplify the synthe-
sis/analysis activity in conceptual design and
give the designer confidence that the initial
universe of potential process paths contained
all the pathways with reasonable chances for
commercial success. The advances in computer-
aided process synthesis that are possible over
the next decade are dramatic. They include both
an increasing level of sophistication (e.g., the
synthesis of heat exchanger networks, se-
quences of separation processes, networks of
reactors, and process control systems) and com-
putational procedures that should make possible
~3
the identification of the most viable process
option in a relatively short amount of time.
As the designer moves from conceptual design
toward final design, he or she must analyze a
number of alternatives for the final design. The
development of large, computer-aided design
programs (so-called process simulators) such as
FLOWTRAN, PROCESS, DESIGN 2000, and
ASPEN (or other equivalent programs used in
various companies J has significantly automated
the detailed computations needed to analyze
these various process designs. The availability
of process simulators has probably been the
most important development in the design of
petrochemical plants in the past 20 years, cutting
design times drastically and resulting in better
designed plants.
Although the available simulators have done
much to achieve superior design of petrochem-
ical processes, there is considerable room for
improvement. For example, better models are
needed for complex reactors and for solids
processing operations such as crystallization,
filtration, and drying. Thermodynamic models
are needed for polar compounds. Moreover, the
current process simulators are limited to steady-
state operations and are capable of analyzing
only isolated parts of a chemical plant at any
given time. This compartmentalization is due
to the limitations on computer memory that
prevailed when these programs were first de-
veloped. This memory limitation resulted in a
computational strategy that divided the plant
into "boxes" and simulated static conditions
within each box, iteratively merging the results
to simulate the entire plant. With today's su-
percomputers, it is possible to simulate the
dynamics of the entire chemical plant. This
opens the way for dramatic advances in mod-
eling and analysis of alternative process designs,
because the chemical reactions that occur in
manufacturing processes are usually nonlinear
and interdependent, and random disturbances
in the process can propagate quickly and threaten
the operation of the entire plant. To nullify the
effects of such disturbances, the designer must
know the dynamics of the entire plant, so that
control failure in any one unit does not radiate
quickly to other units. It is now within our
reach to integrate this sophisticated level of
OCR for page 144
~4
rip
in
A)
.
_ LL
a'
I
in
-
._
~ o
co ~ cn ~
~ I, ~ c
~ - ' At o
Cat ~ -
·- CO
C,,, ._ ._
C C
o o
al ~ ~
Cat , ~
o o
cn On
CO
in
U.
V'
-
CO
-
CO
~ 1
4
~ o
~_
_ ~ cn
-
~n
An,
\/
to
On
~ Lo_
_
cn
C~
1 ~ ~
u)
~< ~
(D
_
~n
1
4
C~
~n
cn
1 ~
c~ 1
1~
tD 1
~\
UD
CO
~r ~ .
cn~
f ~
U1, ~
~ ~:CO
m' 1 _
C~
rn
CO
CO
cn
CO
03 ~
~7
_ ~=
Tm
4
1
!
mT
1
4
l
-
C~
C~
y.:..:i
2..::3
y..:1
~ 3
o ....1
it .3
~ .::.3
e~, . ~
o 3
..... 3
y..:3
y::1
~. ~
: :.y, .: ~ y : : ::y :.:: y y .. :y: y y :y.:y _ ::::: y
ey~ y ~y y ~ ~ ~ ~y y~ --y--y · ·~- ~y ~ y-y ~ y ~ y y- ~ y-
. y.:::: :.:. y : y y ~ : : y ~ ,: , .y:'.: .. .. .. : y :::
::.y: :.: - . :y: .* - , y :.:: :.y :.: : :: :.,. - ~- . ~ : :y ~ y ~ . yy 0 ~ .,, ,: .,,,y _
[7 . ~ .:.::y.:.::: ~ .:.y :.::.: ~_ 7.:.**y7*~ : ?. ~ y* ~ ~ y ~ *. . yZ xy - x. . .y.y~ x.y :. o .y
~ _ ~ ~ ~ ~ . ~ . ~
·$* ~ ye.~ y..y .; - *$ ° · - .y. I * $ · * * ~ ·: $ y: *$ y.- y.:.7 .i . :. N 7 - *$ ·-::: $
~ .y.* y y , y :.: ~ * : : · :y :. . * - : y -,y, ..,y..~ .y y y y ~:
[.: :.:.:.: : .::.: :y.:::. ::. :: ~ * :.: :::: ::.::y$ y.:y: :: ::
. , ~=~=
.=
~ .
Ct .
C~
_ ;>
C~
Q
O
~ _
._
C) ~
.O
04
_ ~
Ct ~
Ct Ct
,.~ V
~ ^
O
, · ~
O ~
'' (V
- (L)
C~
(V ~
~ I .
oc ~
~L) o
~.
O
_ ~
._
~ 50
_
~ s
O
C~ ~
U) {-
~o - o
C)
,=
~, .
. _ .
: - ~
C) . _
s~ (V
O
C)
;^ ~
O
Ce Ct
~ ~,
O
o~ c
C)
V,
3
O ~
O
Ct
O ~
,'
~ 5
Ct -
~ 3
C~
X n
._
rn
,_
C:
_
OCR for page 145
COMPUTER-ASSISTED PROCESS AND CONTROL ENGINEERING
design and analysis on a plantwide scale (in-
cluding design and performance modeling of
plantwide control systems) into the computa-
tional tools used to analyze and optimize the
performance of individual processes in the plant.
In the detailed design stage for a chemical
manufacturing process, a chosen process design
must be converted into a list of equipment items
to be purchased and a set of blueprints to guide
their assembly. The design is presented as a
detailed process flow diagram (PFD), from which
is constructed a list of all needed items of
equipment, and piping and instrumentation dia-
grams (PIDs) that show the equipment and its
interconnections. The next task is to establish
the physical layout for the entire plant. Ad-
vanced computer-based drafting tools aid in all
these activities.
Computer-Assisted Process Retrofitting
The preceding section focused largely on the
design of new plants. However, these proce-
dures can also be adapted to the retrofitting of
existing plants. Retrofitting is generally under-
taken to increase the capacity of a plant; to
make use of a new technology such as an
improved catalyst, a new material of construc-
tion, or a new unit operation; or to respond to
a significant change in the cost of energy or raw
materials. A fair amount of retrofitting in the
chemical process industries in recent years has
been undertaken to improve energy efficiency
through plantwide energy integration retrofit-
ting of about 50 processes to incorporate modern
heat-exchanger network synthesis concepts has
reduced energy requirements in the chemical
industry by 30 to 50 percent. Retrofitting will
continue to play a major role in the design of
chemical plants as new procedures for com-
puter-aided synthesis of separation systems are
applied and research on process synthesis be-
gins to yield large savings by helping existing
petrochemical plants produce the same mix of
products through more economical chemical
reaction pathways.
We need to develop a systematic approach
to analyzing the impact of making changes in
the connections between process units or in the
~5
size of units that are undertaken to improve
operating costs, plant flexibility, or safety.
Research Opportunities in Process Design
The overall goal of process design research
is to develop a systematic procedure, probably
in the form of an interactive computer program,
that contains design heuristics and interchange-
able approximate and rigorous models that can
lead an engineer from an initial concept to a
final design as quickly as possible. The final
design must include considerations of econom-
ics, controllability, safety, and environmental
protection. We need to extend the conceptual
and final design procedures that have been
developed for petrochemical processes to pro-
cesses for producing polymers, biochemicals,
and electronic devices. We also must develop
systematic synthesis/analysis procedures for
studying batch processes analogous to the pro-
cedures that have been developed for studying
continuous petrochemical processes.
There are some aspects of process design in
which decisions are based primarily on past
experience rather than on quantitative perform-
ance models. Problems of this type include the
selection of construction materials, the selection
of appropriate models for evaluating the phys-
ical properties of homogeneous and heteroge-
neous mixtures of components, and the selec-
tion of safety systems. Advances in expert
systems technology and information manage-
ment will have a profound impact on expressing
the solutions to these problems.
In summary, systematic procedures must be
developed for the following:
· generation of process alternatives;
· quick screening of process alternatives us-
ing both rule-of-thumb and short-cut calcula-
tions;
· inclusion of controllability, safety, and en-
vironmental factors in the initial design;
· more detailed screening of a small number
of promising design alternatives;
· design of the process and management of
its construction;
· use and extension of expert systems con-
cepts to handle aspects of design that deal with
OCR for page 146
a mixture of qualitative and quantitative infor-
mation; and
· retrofitting of existing plants.
PROCESS OPERATIONS AND CONTROL
Process operations and control have a tre-
mendous impact on the profitability of a man-
ufacturing operation. In some cases, they can
determine the economic viability of a manufac-
turing facility. For example, Du Pont's Process
Control Technology Panel has estimated that if
Du Pont were to extend the degree of computer
process control that has been achieved at a few
of its plants across the entire corporation, it
would save as much as half a billion dollars a
year in manufacturing costs. If Du Pont's num-
bers are representative, the entire chemical
industry could save billions of dollars each year
through more widespread application of the best
available process control. This could be the
single most effective step that the U.S. chemical
process industry could take to improve its global
competitive position in manufacturing.
Why are such savings suddenly possible?
Because of the explosive developments in com-
puter technology, research on operations and
control is no longer constrained by lack of
computing power. In particular, the traditional
boundaries between design, control, optimiza-
tion, simulation, and operation are disappearing.
Control is becoming a part of process design;
simulation and optimization are becoming com-
ponents of control design.
Research opportunities in process operations
and control lie in three areas:
· collection of information through process
measurements;
· interpretation and communication of infor-
mation by use of process models; and
· utilization of information through control
algorithms and control strategies for both nor-
mal and abnormal operation.
Measurements
The essence of process control is to take
appropriate and quick corrective action based
on measured information about the behavior of
the process. The concept of the process is
FRO~Tl,~RS IN CHE1441~AL E:iGINEERING
contained in a process model, and measure-
ments are used to evaluate the degree to which
the process conditions deviate from those of
the model. When a mismatch occurs between
actual process conditions and those postulated
by the model, a control and operating strategy
is invoked to correct the process conditions. A
critical interrelationship exists between mea-
surements and operating/control strategy, one
that is too often neglected. The control strategy
depends on what information is or can be
available, even while it dictates which mea-
surements are needed. This is perhaps best
illustrated in a number of manufacturing pro-
cesses in cutting-edge technologies, where con-
trol and operating strategy is circumscribed by
the lack of appropriate sensors for many critical
process variables. Conventional estimation
techniques are used that infer the values of
unmeasured variables from measured variables,
but these provide imperfect guidance for process
control.
Even something as empirical as process mea-
surement cannot be divorced from the need for
good process modeling. In the absence of a
good model, it may not be known what variables
affect process operation or product quality and
should therefore be measured or estimated.
Process measurements are subject to errors.
Random (stochastic) disturbances are ubiqui-
tous, and gross or systematic errors can be
caused by malfunctioning sensors or instru-
ments. The detection and elimination of these
errors are essential if the data are to be used
for process operations and control. The success
of this screening depends on the measurements
themselves, the failure data available, and the
process control strategy. At the present time,
diagnostic programs are not applied to most
sensor failure data. The detection and remedia-
tion of significant errors in measurements for
process control pose interesting research op-
portunities.
Interpretation of Process Information
The quality of an operation and control strat-
egy depends on the quality of the model on
which it is based (Figure 8.31. We are only
beginning to understand this relationship quan
OCR for page 147
Ct3.~PtJTE1R-4SSIST]ED P3RGitCiESS A Dry ~ C)Ai7~OL JE.~ ~.~1lNG
-
r
~ i'
...and in 1 /10,000 of a second it can compound
the process model's error 87,500 times!"
FIGURE 8.3 The importance of accurate process models.
Copyright 1988 by Sidney L. Harris.
titatively, even in the relatively simple context
of the feedback control of linear systems. Even
if it is assumed that the structure of the process
model is correct, we do not yet know how to
translate uncertainties in model parameters into
uncertainty in the performance of the control
system. A more difficult problem is to assess
the effect of an incorrect model structure, such
as a wrong set of basic equations, on the
performance of the control strategy on which it
is based. Understanding the effect of model-
process mismatch on control system perform-
ance provides a critical research opportunity.
In the context of operations and control,
simulations can be used to test new process
strategies as well as to train operating personnel
to control the process and to respond to emer-
gencies. The increasing use of simulation in
process control requires that the cost of dynamic
simulation be brought down. This could be done
by taking advantage of new developments in
computers, such as new user interfaces, com-
puter architectures, and languages, and by de-
veloping faster numerical integration algorithms
for ordinary differential equations.
Alarm management also requires research.
Modern chemical plants usually have audible
and visual annunciators to warn operators when
key variables deviate from acceptable or safe
values. A process upset in a plant that has
several interconnected units with many feed-
back controls can set off multiple alarms, and
the consequences of misinterpreting the alarms
can range from inefficient process operation to
outright disaster. When the alarm sounds, the
operator must decide quickly what action to
take. A hybridization of expert systems and
process control systems can assist the operator
in interpreting process status after an abnormal
event. The need for better handling of abnormal
events makes research in artificial intelligence
of great importance to the chemical industry.
Integration of Process Design with Control
Most continuous plants are now designed for
steady-state operation with little regard for the
ease (or difficulty) with which the steady state
can be maintained through control. Such a plant
can be difficult to control once it deviates from
the steady state.
Design and control have traditionally been
treated separately for the following reasons:
· The problems in each area alone are ex-
ceedingly complex.
· The interactions between design, control,
and optimization are poorly understood.
· The computational requirements of an in-
tegrated approach to design and control have
been beyond the capability of available hard-
ware and software.
For example, a chemical plant might be de-
signed to achieve high efficiency by integrating
the operation of many individual process units
across the plant (e.g., by using waste heat from
one unit as an energy source in another unit).
However, the tight coupling of process units
generally makes the entire plant more difficult
to control. Therefore, this is a factor that must
be considered at the design stage. No method-
ology currently exists for including this consid-
eration in plant design; its development consti-
tutes a significant research opportunity.
The supercomputer power that is becoming
available will provide the opportunity to com
OCR for page 148
- - c'
bine process and control sys-
tem design-including optimiza-
tion into one large problem that
can be solved in a way that ac-
counts for their interactions. The
success of such a consolidation
will depend on the develop-
ment of approximate compatible
models and of techniques to re-
late model quality to perform-
ance.
Robust and Adaptive Control
Control systems are designed
from mathematical models that
are generally imperfect descrip-
tions of the real process. It is
essential that control systems op-
erate satisfactorily over a wide
range of process conditions. Thus,
the control algorithm must pro-
vide for control of the process
even when the dynamic behavior
of the process differs signifi-
cantly from that predicted by the
model. A control system with
this characteristic is sometimes
called robust. In fact, a tradi-
tional disregard for the model
error problem is one of the main
reasons for the frequently cited
gap between theory and practice
in process control. Industry needs
algorithms that are robust rather
than ones that "get that last half
percent performance." Control
strategies that work all the time
within reasonable limits are bet-
ter than those that work optimally some of the
time but that frequently require reversion to
manual control.
Because over time a process often changes,
its model parameters must be continually up-
dated; in extreme cases, the basic model must
be reformulated. An adaptive system is a control
system that automatically adjusts its controller
settings or even its structure to accommodate
changes in the process or its environment. The
problem of model-plant mismatch is of crucial
importance in the design of adaptive controllers
for processes since it is that very mismatch that
drives changes in the controller parameters.
The engineering theory and methodology for
designing reliable adaptive controllers for chem-
ical processes are in the earliest stages of
development.
Finally, there are always process operations
in which neither classical nor modern control
is effective. Such operations may require qual-
itative decision making or the use of past knowl
OCR for page 149
CO.~JPU1~-~SSIS~D PROCESS A.~D CON~L E\~.~G
edge. Artificial intelligence techniques offer
promise for control system design in these cases.
Batch Process Engineering
The production of fine and specialty chemi-
cals, which are usually made by batch pro-
cesses, is becoming increasingly important and
competitive. The efficient operation of multi-
product and multipurpose batch plants offers a
variety of challenging research problems for
chemical engineers. Most industrial batch chem-
ical operations are now scheduled by intuitive,
ad hoc methods that consist of modest variations
around historical operating patterns and that
make little or no use of computer technology.
It is now widely recognized that the scheduling
problems associated with batch processes are
immensely complex and, in fact, are among the
most difficult combinatorial problems known.
Limited progress has been made in using math-
ematical models in the simplest types of batch
process scheduling. Current algorithms are too
computationally demanding and complex for
industrial use. An important intellectual chal-
lenge is to generate a unified field of batch
process engineering theory and to put it into a
practical context by using case studies.
Linear control theory will be of limited use
for operational transitions from one batch re-
gime to the next and for the control of batch
plants. Too many of the processes are unstable
and exhibit nonlinear behavior, such as multiple
steady states or limit cycles. Such problems
often arise in the batch production of polymers.
The feasibility of precisely controlling many
batch processes will depend on the development
of an appropriate nonlinear control theory with
a high level of robustness.
While startup and shutdown occur relatively
infrequently in large continuous plants, they are
inherent in batch plant operation. Most startup
and shutdown procedures, whether devised em-
pirically or theoretically, are designed to follow
a recipe of actions with no feedback. Thus, if up-
sets occur, there is often no way to change the
startup or shutdown in time to avoid unwanted
process excursions. Procedures are needed that
incorporate feedback and adaptive techniques
to the problem of plant startup and shutdown.
PROCESS SENSORS
If we had a completely accurate model of a
process and accurate measurements of process
disturbances at their inception, then corrective
action could be taken directly without the need
to measure the output streams from the process
after the disturbance has propagated through it.
But because we generally do not have adequate
models, the output streams of processes must
be measured for the purpose of feedback con-
trol.
The sensor is the "fingertip" of the process
control system. The principal challenge in pro-
cess sensing is the development of analytical
sensors, particularly for determining process
stream composition. Such sensors eliminate the
need to withdraw samples to determine process
and product parameters, a practice that should
be minimized because of inherent problems
(e.g., samples of reactive intermediates may be
toxic or otherwise dangerous, or the interven-
tion represented by withdrawing a sample may
affect process operation). Since it is important
for process control not to disturb the normal
operation of the process, sensors are needed
that can operate in the environment of the
process stream. The key to meeting this chal-
lenge is a fundamental understanding of the
physical and chemical interactions at the sen-
sor-environment interface and, in particular,
the transport and kinetic processes that occur
there.
Future Sensor Developments
The techniques used in the chemical pro-
cessing of electronic microcircuits (see Chapter
4) are being adapted to the microfabrication of
two- and three-dimensional structures for solid-
state sensors. These techniques will permit the
integration of transducers, optoelectronics, sig-
nal-conditioning and data-processing devices,
and micromechanical devices into extremely
small packages. Reduced size offers advantages
in thermal uniformity and response speed; shock
and acceleration resistance; and reduced weight,
volume, power, and cost.
Solid-state sensors may be developed that
will be responsive to a broad range of acoustic
OCR for page 150
150
inputs, electromagnetic radiation, ionizing ra-
diation, and electrochemical stimuli. Response
elements may be tailored for high selectivity
among ions, free radicals, or specific com-
pounds. Alternatively, elements with low selec-
tivity are also useful because information from
an array of such sensors, each with a different
but known broad response, can be processed
to provide quantitative analysis of a complex
mixture. Complex mixtures also lend them-
selves to chromatographic analysis. It has been
shown that gas chromatographic data can be
analyzed on a silicon chip, although with some
loss in recognition reliability. Combinations of
gas or liquid chromatography or capillary elec-
trophoresis of microsamples with mass spec-
trometry may be developed to provide superior
performance.
The development of biological sensors is
taking place at a rapid pace. Biological sensors
analyze chemical mixtures using biological re-
agents of exquisite specificity for example,
enzymes, immunoproteins, monoclonal anti-
bodies, and recombinant nucleic acids. Such
sensors may permit the analysis of fast reactions
of species in very dilute media. Multicomponent
biological sensors may be able to perform com-
plex analyses that involve multiple reactions,
with automatic regeneration of the biochemical
reagents or removal of interfering species.
Unfortunately, current biological sensors are
extremely delicate. Even when the biological
reagents are immobilized on a solid carrier,
such sensors require careful construction and
frequent recalibration, are not always amenable
to automation or unattended operation, and
sometimes have inconsistent dynamic response
and limited life. Although biological reagents
are ideally suited for some applications, partic-
ularly those in relatively mild environments,
they may not survive the harsher conditions
often found in process industries. Here again,
miniaturization of the biological sensor and its
direct integration into an optoelectronic trans-
ducer, potentiometric electrode, or membrane
are promising approaches. With the current
worldwide interest in biotechnology, major in-
novations in biological sensors can be antici-
pated.
Further advances in optoelectronics will allow
FRONTIERS IN CHEMICAL ENGINEERING
d_
L OUTPUT
TRANSMITTER
0~
/COUPLER
RECEIVER
OPTO ELECTRONIC
HUB
r FIBER
/LIGHT
-. ~ ~ ~ OR ·-~ ~^ .-
~ \
/ EXTRINSIC:
OPTICAL
SENSOR
(OPTRODE)
SENSOR
NETWORK
OR
·-~-
INTRINSIC FIBER SENSOR NETWORK
FIGURE 8.4 Configurations for several different kinds of
optical fiber sensing systems are shown. The common factor
in ail these systems is the use of an optical fiber as an
integral element in the system, either to carry light to and
from discrete sensors (often referred to as optrodes), or as
sensitive elements themselves (intrinsic fiber sensors).
Courtesy, AT&T Bell Laboratories.
the development of instrumentation with no
electrical components in the sensor (Figure 8.41.
These devices will operate by transmitting probe
light from a remote source to the process sensor
with an optical fiber light guide. In the sensor,
the light signal will be altered by the sensed
environment (e.g., by absorption of certain
wavelengths, fluorescence, or scattering) and
will thus be "encoded" with information. The
encoded signal is transmitted through the optical
fiber to a transducer that produces an electronic
signal. The advantages of such systems include
inherent safety, low signal attenuation, and the
ability to multiplex signals in the optical fiber.
Such instrumentation can incorporate additional
chemical, biological, and electronic components
and is likely to play a major role in many future
sensor systems.
Future sensors and their associated data pro-
cessing elements will need capabilities beyond
those required for the simple measurement of
process variables, such as periodic self-calibra-
tion against known standards, automatic com-
pensation for environmental or other interfer-
ences, signal conditioning including linearization
or other variable calculation, and fault recog-
nition and diagnosis. Some of these capabilities
are now available to a limited extent. Others
will become available with the continued de-
velopment of integrated sensors and data pro
OCR for page 151
CO'WPUTER-ASSISTED PROCESS A1YD CO1~L E^~1VEERING
cessing instrumentation. Arrays of sensors have
already been mentioned in connection with
complex mixture analysis, but they may also
be used to provide redundancy, fault detection,
and data reconciliation.
Research Opportunities
The availability of high-quality, real-time in-
formation on the conditions and composition of
the process stream will permit engineers to
develop a completely new generation of process
control strategies. The physical, chemical, and
biological phenomena at the sensor/process in-
terface must be understood and translated into
sensor technology. Chemical engineers are well
positioned to contribute to the development of
improved process sensors in a variety of ways,
including
· work in interdisciplinary collaborations with
electronic engineers, biologists, analytical
chemists, and others to elucidate the biological,
chemical, and physical interactions to be meas-
ured;
· application of fundamental principles of
reaction engineering and transport phenomena
to the design of sensor surfaces;
· development of new process control sys-
tems and operation strategies in response to
improved capabilities for measurement; and
· determination of the implications for pro-
cess design of wholly new types of process
sensors.
PROCESS ENGINEERING INFORMATION
MANAGEMENT
In the next decade, competition among in-
dustrialized countries will be influenced by the
way in which information and knowledge are
managed in industry. The challenge is to be the
first to find relevant information, to recognize
the key elements of that information, and to
apply those elements in the manufacture of
desired products.
Computer technology will continue to provide
new generations of hardware and software for
fast information processing and low-cost storage
and retrieval. The use of computers for infor
151
mation management and decision making will
be essential, and advanced capabilities in user
interfaces and networking will bring new di-
mensions to this application. For example, cur-
rent on-line literature search systems allow data
sharing among many users, significantly increas-
ing individual productivity. However, this tech-
nology is generally only used to manage well-
organized data; basic research is needed to apply
it to engineering data, which are not as well
organized.
A process engineer will need to be able to
store and access relevant data rapidly in order
to carry out process development and design in
less time and to solve problems arising from
new and complex design requirements (e.g.,
designing for multiple objectives of profitability,
safety, reliability, and controllability). To pro-
vide for rapid data storage, access, and transfer,
new generations of computer hardware (bulk
storage, network, and work stations) and soft-
ware (data bases) will be needed. Some specific
research challenges for chemical engineers,
working together with computer scientists and
information specialists, follow.
· Most chemical manufacturing processes in
the future will be monitored and controlled by
computer. Process data will be collected con-
tinuously and stored either locally or in a central
data base. Research is needed to develop mu-
tually compatible, efficient algorithms for stor-
ing and retrieving process data. In addition,
computerized procedures will be needed for
sifting through voluminous process data for
information to use in process improvement and
in the generation of new processes.
· Methods must be developed for storing
judgments, assumptions, and logical informa-
tion used in the design and development of
processes and models.
· Procedures and methods will be needed for
retrieving and operating on other types of im-
precise data.
· Efficient transfer of information among en-
gineers and designers will depend on how easily
these data can be accessed. The special needs
of chemical engineers in this area the partic-
ular ways in which they generate, manage, and
use information merit study.
OCR for page 152
152
IMPLICATIONS OF RESEARCH
FRONTIERS
The speed and capability of the modern com-
puter, as well as the developing sophistication
of chemical engineering design and process
control tools, have tremendous implications for
the practice of chemical engineering. Chemical
engineers of the future will conceptualize and
solve problems in entirely new ways. There are
two bottlenecks to the application of these
powerful resources, though. First, there are not
enough active research groups at the frontiers
of computer-assisted process design and con-
trol. A larger effort is needed in order for the
field to keep pace with the expanding power of
available computers. Second, many chemical
engineering departments lack the computational
resources needed to fully integrate advances in
design and control into their curricula. For the
full potential of the computer to be realized in
improved design of chemical products and the
improved design and operation of processes to
FRAN TIERS IN CHE`~ ICAL E.~.N'EE,~G
produce them, chemical engineers must be
broadly versed in advanced computer technol-
ogy. This can only happen if they have access
to state-of-the-art computational tools through-
out their educational careers, not in an isolated
course or two.
Making this broad access and utilization of
the computer in education possible will require
substantial government, academic, and indus-
trial funding to provide both hardware and
software. In some cases, groups may need
remote access to networks of supercomputers.
In other cases, dedicated array processors and
other computational hardware may be required
in the chemical engineering department. If this
country is to maintain a leadership role in
chemical technology, critical needs for both
research support and facilities acquisition must
be addressed. The status of funding in this field
and a specific initiative to achieve the goals
outlined above are discussed in Chapter 10 and
Appendix A.
Representative terms from entire chapter:
chemical engineers
