2006-1214: PROCESS CONTROL: A RELEVANT APPROACH

Barry Barkel, University of Michigan

Lecturer in Chemical Engineering, University of Michigan

Peter Woolf, University of Michigan

Assistant Professor of Chemical and Biomedical Engineering, University of Michigan

Page 11.1026.1

© American Society for Engineering Education, 2006

Introduction

Information from many sources, including both new and past graduates, employers and
even from a sitting chairman of the AIChE, suggested that the process control course at
the University of Michigan was not relevant to the workplace that our graduates were
entering. We also conducted interviews with experts actually practicing control
engineering in chemical process systems. These sources confirmed the comments we had
received. All sources, however, indicated that there are several aspects of process
control that graduates do need to know and that these areas were not being covered by
current course work.

Based on inputs that the course content provided few useful skills to our students, serious
consideration s given to dropping the process control course from the required
curriculum. Dropping this course would eliminate the wasted time and energy students
spent in a nonproductive pursuit and allow inclusion of new courses in the required
curriculum. An alternate proposal involved creating a new course that met the needs of
students by including those skills that had been identified as having value in today’s
workplace. Because the authors felt strongly that graduating engineers must be able to
control the processes they designed or managed, we undertook the task of redesigning the
process control course.

What

Several elements of process control were identified as relevant to basic understanding of

the material and needed to be included in the new process control course:

• A knowledge of sensors and control devices:

• Reading and creating P & IDs for process systems
• Developing control “philosophies” and strategies
• Control in biological systems
• Writing control logic for process systems
• Mathematical modeling of process systems
• Dynamic modeling of real systems
• Understanding of convergence, stability and response of systems
• PID controllers and their tuning
• Statistical process control

The design task for the new course was to cover the above material adequately, within a
15-week term, and to do so in a manner that that had flow and continuity. In addition,
each section should relate in a logical way.

To accommodate this material, we developed the course plan listed in Table 1. In

comparison to the more traditional course plan that we followed previously, this course
represents a significant deviation, both in terms of content and emphasis.
Page 11.1026.2
 New Course Old Course
Week 1 Introduction Introduction
Team Dynamics
Week 2 Temp & Press. Measurement & Control Laplace Transforms
Biosensors & Response Inversion by partial fractions
Week 3 Flow and Level, Measurement & Control Other properties of transformations
P&IDs Response of 1st order systems
Control Devices, Safety& Enviro. Factors
Week 4 Process Design as a control tool Examples of 1st order systems
Systems Integration 1st order systems in series
Week 5 Feedback, feedforward, cascade systems Higher order systems I
Control Logic Higher order systems II
Week 6 Blood Sugar Control Mechanisms Control Systems & Elements
Review Closed loop transfer functions
Week 7 Exam Exam
ODE Model Construction Transient response-simple systems
Week 8 Fixed Pts and Stability in ODE Eigenvalues and Stability
Linearization of ODEs Introduction to root locus
Week 9 Matlab analysis of ODEs Intro to frequency response
ODE regression Design by frequency response
Week 10 Prop. Intgrl. Deriv. & PID control Intro to State Space representation
PID control cont. Transfer function matrix
Week 11 PID tuning from experimental data Controller tuning
Bode plots, freq. response Control valves, and practical issues
Week 12 Review Review
Exam Exam
Week 13 Model Predictive Control Multivariable control
Advanced Control Design Advanced Control Strategies
Week 14 Statistical Process Control Statistical Process Control
Statistical Process Control Statistical Process Control
Week 15 Review Review
Final exam Final exam
Table 1: Schedule for the current course and the previous, more traditional process
control course. Classes meet for two hours per class, twice a week.

Appreciation of the Realities

While the prior “classical” treatment of controls gave our graduates knowledge of the
mathematics that represented how a simple control system would respond, there was little
to tie the math to complexity of the real, physical system. We felt it was essential to
promote an understanding of how control really works. However, the first three years of
the chemical engineering curriculum involves learning the “basics” and focuses strongly
on computational solutions to problems with a single answer. Students entering the
process control course expected similar single answer computational problems. Indeed,
many process control textbooks closely follow this single answer model. However, most
real world problems in control have a range of possible solutions. We chose to
Page 11.1026.3

emphasize the reality of selecting one choice from among a number of alternate
acceptable answers as an important part of the learning process. While our students have
been exposed to open-ended problems in other courses, this exposure was limited.

To help overcome the student discomfort with alternate solutions to the same problem,
we used two assignments and an experiment early in the course to focus on the reality
and complexity of control. Below we describe these parts of the course in more detail.

The first homework assignment was to instrument an android to mimic human responses.
Students were limited to commercially available sensors and control devices to simulate
actions such as sleep, exercise, injury, etc. This helped focus them on the detect/respond
nature of control.

We emphasize in class that understanding a process and the control objectives are
essential to designing a control system. To focus on this understanding, the second
homework assignment asked student to develop a list of questions they would ask the
master distiller of distillery to enable the process to move from the “art of experience” to
effective modern control. This problem served to emphasize the depth of understanding
necessary to effect control.

To provide some hands on understanding, an experiment was developed to illustrate the

complexity of control (Figure 1). This system uses a variable flow of hot water through
a dry ice bed to generate carbon dioxide gas. The control objective is to maintain a
constant flow of the product gas. The system was specifically designed as a difficult
control system as the CO2 liberation depends on the quantity of dry ice remaining at any
given time, the geometry of the dry ice block, and the buildup of ice on the dry ice block.
None of these are measurable by the control system—instead the controller has access
only to a few temperature and flow measurements shown in the schematic. Students were
asked to run the experiment as provided and then develop improvement to the system to
make it more controllable. This experience developed a good appreciation of the fact that
control parameters are never constant and, like driving a car, keeping things within the
boundaries is the only way to manage the task.

Page 11.1026.4
Figure 1: A schematic of a CO2 production experiment used in the course to provide
hands on experience for a complex control system.

Sensors, Control Devices, and Systems

Students have only limited prior knowledge of sensors and control devices and their
application. Most of that knowledge comes from contrived situations where the
instrumentation has been selected and installed for them in a lab course. These
experiences provide little understanding of how to select a device and the device’s
limitations. Beyond this, these contrived experiences provide little to no understanding
of where to place instrumentation to properly control the system. Prior process control
courses made no attempt to teach this material despite of its critical role in the field.

In the new course we chose to include more information about the sensors themselves
and their rational placement for control. Sensors and control devices do not stand-alone,
they must be interrelated to be effective. Most traditional controls texts show this
interrelationship as “hard wiring”, a direct relationship between a sensor and a control
device. Yet modern control systems for the last 20 years or more have routed signals to
and from sensors and devices through computers. There the information is processed via
mathematical models to select the controller action. The exact routing of data and the
processing of responses can become extremely complex and is always a function of the
specific system. Yet in most process systems there is a generality of which sensors and
control devices are or ought to be interrelated. In the new course, students are taught to
recognize these generalities and explore the advantages and disadvantages of specific
relationships.

Piping and Instrumentation Diagrams (P&IDs) are a universally recognized method of

Page 11.1026.5

communicating sensor and actuator placement in a control system. This basic chemical
engineering document was not employed in the prior controls course. To remedy this,
The University of Michigan Department of Chemical Engineering has developed a
technical standard for student P&IDs that is now used in a number of core courses. This
standard is now presented first in the process control course and students are expected to
use the instrumentation principles of this standard in their homework and on exams.

From the beginning, we emphasize that controllability is not only a part of the sensors
and valves, but also embedded in the architecture of the process itself. The physical
configuration of equipment, the sequencing of operations and the conditions under which
operations are carried out have as much to do determining the control system as any of
the sensor and control devices. The philosophy of the control strategy begins in the
physical design. By control philosophy, we mean pre-determining limits on conditions so
as to assure safety and optimize the result. A decision to process a material as a high-
pressure liquid versus low-pressure gas due to easier controllability is an example of
establishing a philosophy on the basis of control. Students are taught to recognize that a
basic philosophy is important in a control design; it is the focus for the subsequent
strategies employed within that control system.

Case studies are used to illustrate the above concepts and the students do homework
assignments involving similar process systems to reinforce the principles. A team based
project is also assigned where students are given basic chemistry and processing
requirements and are required to assemble appropriate equipment and develop the P&ID
of their system showing all instrumentation and the interrelationships among sensor and
control devices.

Control in Biological Systems.

Control in and of biological systems had not previously been included in the Michigan
control course. The strong emergence of biological options in chemical engineering has
made it imperative to expose students to this area and to demonstrate that, in general, the
same control issues apply to living organisms and large chemical complexes alike. In
addition, research into biological control mechanisms has suggested a number of such
mechanisms that could potentially be applied to non-biological control problems [1].

For the course we introduced two case studies related to biological systems: blood
glucose control and bacterial chemotaxis. Throughout the semester we introduced the
problem of glucose control via insulin in the human body. Initially, we introduced the
physiology of the problem and used a simple pharmacokinetic model of insulin as an
alternative example for a logical controller. This model was extended to the Bergman
Minimal Model [2, 3], to provide a more accurate differential equation model for glucose
regulation. Students were assigned a team project to design an insulin injection schedule
for a diabetic patient. For this project, students were provided with historical patient data
on blood glucose and insulin injections. One of the student programs was tested on the
patient—one of the faculty teaching the class.

A second biological case study in the course was the problem of bacterial chemotaxis. In
Page 11.1026.6

response to a food gradient, a bacterium will perform a biased random walk toward the
food source. This is a well-studied problem, both biologically [4], and from a controls
perspective [5], making it ideally suited for a senior level engineering course. In this case
study, students were asked to read the scientific literature and explored the similarities
and differences between bacterial chemotaxis and control in a conventional chemical
process.

By introducing bacterial chemotaxis, we achieve two ends. First, we expose the students
to a relatively unexplored set of questions where the physics, biology, and mathematics
are all incomplete. Second, we also introduced the students to the idea of decentralized
control, a concept critical to regulating nano-scale machines.

Operating Logic

Even the most sophisticated hardware system will not function unless it is programmed.
The program is what makes the installed control system operational. Previous control
courses had not dealt with this issue, yet it is often an area that chemical engineers
become involved with early in their careers. Programming can take many forms but it is
the general concepts that are important not the specifics. A simple IF… THEN program
format was used in the course. The simplicity of this format proved useful as it forced
the students to consider actions/reactions in small steps and minimized overlooking
factors that might be obscured in a more sophisticated format. Asking the students to
write a “story” spelling out what they wanted to the program to accomplish proved to be
a very useful tool in developing programming skills. Students usually captured in words,
operations that they tended to overlook when working directly a programming format.

Batch systems rather than continuous operations were the primary focus of programming
exercises, as is shown in Figure 2 and 3. Continuous systems tend to be less interesting
when operating at steady state and the intricacies involved in start-up and shutdown are
too complicated to present in an undergraduate controls class. Batch operations have
well defined steps in ramping up and ramping down and generally contain a steady state
segment in their sequence. From a controls perspective, batch processes are therefore
more interesting and illustrative for use in the case studies used in both class and
homework.

Because understanding of the effects of program instructions was felt to be as important

as knowing how to develop the program, several methods of program analysis were
introduced. Truth tables and graphical methods were used to examine the system
response to specific inputs, an example of which is shown in Figure 4. Students were
also given an assignment where Matlab was used to simulate a process controller.
Page 11.1026.7
 BLOWER

N2
C

CW

1) Please position the sensors and control devices in this system

2) Please write the control logic for this batch reaction

A + B --> C + ABC-CBA Reaction is exothermic

It goes to 100% completion

A and B are liquids B is reacted slowly into a full charge of A

C is a gas and the product
ABC-CBA is a waste liquid
Tcw< TRx<TA<TB<<Tabc-cba
Where: TRx is the reaction temperatue,Tcw is the water temp
and other Ts are the BPs of of the designated species
The exotherm is sufficient to get the reaction mass to TRx
The reaction is run at 0.5atm. C is stored at 1.5 atm

Figure 2: Example control programming problem of a batch process.

Page 11.1026.8
Figure IV

High Pressure Air Line

AIR
(Mandrill Inflation)
Vent

Fixed side Movable side

of mold of mold
Vent

M
A
N
D
A R
I
L
L
AIR
B (pneumatic
ram)

Electric
HEATER

POWERSTAT
A part is manufactured by reactive molding in a 2-piece mold. The 2 halves of the mold are clamped
together using pneumatic pressure. Predetermined amounts of A and B are then injected into the mold
which has been preheated with an electric heater. And expandable mandrill is then “inflated” with air until
the desired mold pressure is reached. A and B are then allowed to react for time,
R at
t temperature TR
and pressure PR The mandrill is then “deflated” and the mold allowed to open so that the part can be
removed. The reaction is:2A + B --> C (to completion)
The reaction is endothermic
1) Please place the sensors and control devices on this equipment, Show relationships.
2) Please write the logic program for this process. Note correct order is essential
Page 11.1026.9

Figure 3: An example control programming problem for a molding operation

Figure 4: An example of a truth table and state transition graph for a control program
developed in the course. Note the parallels between the stability analyses done in
continuous systems.

Classical Content

The concept of system performance is important to understanding control systems and

tends to be the central focus of classical control courses. Yet few chemical engineers end
up using theses skills in actual practice. Our focus was to provide an appropriate level of
understanding of these issues without delving too deeply into the details.

Feed forward, feedback and cascade control were incorporated implicitly from the very
beginning by way the P & IDs and case studies. Later in the course we analyzed these
architectures using differential equation based models.

In general, our students were not good at looking at a real process system and writing a
simplified yet representative mathematical model of that system. Understanding the
natural response in a real system is basic to understanding control of that system. The
Page 11.1026.10

key to understanding the natural response of a real system is the ability to analyze a real
system, determine the most influential variables and mathematically model that system so
that the effect of imposed control can be evaluated. Material was presented to emphasize
the relationship between experimental data and a physical system with models based on
ordinary differential equations.

Throughout the course we opted not to teach traditional frequency domain, Laplace
transform techniques. While useful in some applications, most frequency domain
methods tend to obscure the student’s physical and intuitive understanding of the process.
In contrast, much of the chemical engineering curriculum is based in the language of
ordinary differential equations. As such, students come into the course with some ability
to solve, analyze, write, and interpret ordinary differential equations. In the spirit of
extending the students ability to solve problems, not perform rote calculations, we cut
most of the frequency domain topics from the course.

Convergence, stability and response are concepts necessary to understanding how a

control system will behave. To illustrate these concepts, we introduced some concepts
from advanced control theory and from nonlinear dynamics, such as phase portraits and
incidence graphs [6, 7]. For both of these presentations, we emphasized less the
theoretical properties of the method, and more the applied uses and intuitive behaviors.
As an example, for phase portraits students constructed a possibly nonlinear model of a
process, identified stable points, and used Matlab to calculate the trajectory that the
system would take to approach the stable point from another point in the space.
Traditional approaches to linearization and stability analysis were also presented and
modeled numerically to demonstrate their range of applicability.

Incidence graphs were used as a simplified model of the interactions between process and
control variables. Incidence graphs are a way to describe immediate causal relationships
between pairs of variables [1]. An example of an incidence graph problem is shown in
Figure 5. Both incidence graphs and phase portraits were used in class and as homework
assignments.

PID controllers and tuning were covered as particular extensions of the more general
nonlinear analysis described above. Here too, incidence graphs and phase portraits
proved useful in describing the systems.

Page 11.1026.11
Figure 5: This example problem shows how a series of causal activating or repressing
relationships can be logically propagated to yield useful insights for control.

Statistical Process Control

Statistical process control (SPC) is an important form of process control, yet is seldom
mentioned in any chemical engineering curriculum. SPC is a control method heavily
used in almost every field where chemical engineers work. Industry sources repeatedly
mentioned the importance of this type of control and the relative weakness of our
graduates in the basic concept of it and the underlying statistical methods.

Somewhat unexpectedly we discovered that to present this topic effectively, a review of

statistical principles was required. Remedying this deficiency occupied more time than
expected, and will likely occupy a larger part of the course in future versions.

Briefly, SPC is class of methods for improving the reliability of a process and for
identifying when a process is out of a user defined control range. The underlying
modeling skills for the topic can be simple or involved, depending on the application.
From a process control perspective, SPC is valuable in that it tells the user when a control
strategy is not working or when the system has changed.

To teach SPC, we chose to emphasize the concepts of six sigma and control charts.
Explaining six sigma required the introduction of probability, probability distributions,
normal distributions, and z-values. From this foundation, we introduced control charts
and how these charts are used to detect a process upsets.
Page 11.1026.12
Difficulties in Course Design

As with any new course, we found a few surprises and unexpected difficulties along the
way.

As mentioned elsewhere, student background for the new material was not uniform.
Areas of basic knowledge sometimes had to be reviewed before control material could be
presented as planned. This should become less of a problem in the future as we adapt the
presentation to the material and minor modifications can be made elsewhere in the
curriculum once we can more fully define the background needs for this new course.

Of particular concern was the degree of anxiety among student for non-computational,
open-ended problems. Remedies for reducing the anxiety associated with open-ended
problems are being investigated, as this was not the only class where the problem was
observed. Efforts will be made to increase the increase the open-ended content earlier in
the curriculum to remove this distraction from learning at the senior level.

Finding an appropriate textbook is also an outstanding challenge for this course. Some
texts such as those by Bequette[3], Stephanopoulos[8], Marlin[9], and Riggs[10] cover
portions of the material, but tend to emphasize mathematical techniques more heavily
than control strategies for process systems.

Currently the background material for lectures and the course readings are drawn from a
large number of sources listed in detail in Table 2. In general, background material is
distributed to the class or posted on the class website. Suggested reference readings are
also listed for students looking for additional information. However, for many students
the lack of a single, comprehensive text is frustrating.

Grading of open-ended solutions such as P&IDs, control logic programs and team
projects is time consuming. Some strides have been made in standardizing solutions
through “sectioning” problems into component parts not affected by other parts of the
solution. At the same time we recognize that too much standardization will smother the
individual creativity, which the course is designed to promote. Complicating the grading
issue is that there are relatively few graduate students with a background in many of the
areas covered in the new course. Primary instructors must do a large percentage of the
grading themselves. It is felt that this new approach to process control will always
require a higher than normal work requirement from the instructors.
Page 11.1026.13
Topic Source/Reference
Team dynamics Original material
Temperature and pressure, measurement [11, 12], + original material
and control
Biosensors and signaling [2, 4, 5, 13], + original material
Flow and level, measurement and [11, 12], + original material
control
P&IDs Original material
Control devices, safety, and Original material
environmental factors
Process design as a control tool Original material
Systems integration Original material
Feedback, feedforward, and cascade [3, 10]
architectures
Control logic [1, 14],+ original material
Blood sugar control [2, 3]
ODE modeling of chemical systems [3, 12]
Fixed points and stability, phase [6, 7]
portraits, and linearization
Matlab analysis of ODEs [3],
PPlane(http://math.rice.edu/~dfield/dfpp.html),
+ original material
P, PI, and PID control [10, 12], + original material
PID tuning from data Original material
Bode plots, frequency response [9, 12]
Model predictive control [3], + original material
Statistical process control [12, 15], + original material
Table 2: Sources and references for course handouts and readings.

Course Acceptance & Results

Student acceptance of the course was good as determined by standard post course
assessments. Scores on questions such as “This was an excellent course” and “ I learned
a lot in this course” were in the same range as those for other core courses. Many
students reported on the open assessment forms that they had learned “ a lot”. Several
students who had served internships in industry reported that they immediately
recognized the value of the new material. Many others liked the inclusion of material
that directly related to professional practice. There was considerable anxiety early in the
course on the “open ended” nature of the material. Some students came to enjoy the
creativity that it allowed. Other students, however, were still uncomfortable with the
non computational nature of problems even after they had completed the course.
Page 11.1026.14

Grades in the class had a normal distribution for a senior level class. Most students had a
good mastery of the material by the end of the course, as demonstrated by exam and
project results throughout the semester.
Including more relevant material in the process controls course has had a significant
effect on senior process design course, which most students take the following term.
Students in process design are required to submit a P&ID for their project design. In the
past these students had only a vague idea of how to control the critical elements in their
design. Now students are capable of a fairly sophisticated process analysis and submit
P&IDs very close to the “first pass” control designs found in industry.

Conclusions

Overall, we feel we have created a new undergraduate process control course that is
focused on material relevant to the skill requirement in today’s workplace. The efforts
described here represent a work in progress. We still have much to learn both in content
and in presentation methods. But these first steps have given us a glimpse into how to
make what we teach relevant to the needs of industry and our graduates.

References

1. Sontag, E., Some new directions in control theory inspired by systems biology.
Systems Biology, IEE, 2004. 1(1): p. 9-18.
2. Bergman, R.N., L.S. Phillips, and C. Cobelli, Physiologic evaluation of factors
controlling glucose tolerance in man: measurement of insulin sensitivity and beta-
cell glucose sensitivity from the response to intravenous glucose. J Clin Invest,
1981. 68(6): p. 1456-67.
3. Bequette, B.W., Process control: modeling, design, and simulation. Prentice-Hall
international series in the physical and chemical engineering sciences. 2003,
Upper Saddle River, N.J.: Prentice Hall PTR. xxix, 769.
4. Alon, U., et al., Robustness in bacterial chemotaxis. Nature, 1999. 397(6715): p.
168-71.
5. Yi, T.M., et al., Robust perfect adaptation in bacterial chemotaxis through
integral feedback control. Proc Natl Acad Sci U S A, 2000. 97(9): p. 4649-53.
6. Khalil, H.K., Nonlinear systems. 3rd ed. 2002, Upper Saddle River, NJ: Prentice
Hall. xv, 750.
7. Strogatz, S.H., Nonlinear dynamics and chaos: with applications to physics,
biology, chemistry, and engineering. 1st pbk. print. ed. 2000, Cambridge, MA:
Westview Press. xi, 498, 4 of col. plates.
8. Stephanopoulos, G., Chemical process control: an introduction to theory and
practice. 1984, Englewood Cliffs, N.J.: Prentice-Hall. xxi, 696.
9. Marlin, T.E., Process control: designing processes and control systems for
dynamic performance. 2nd ed. 2000, Boston: McGraw-Hill. xxxiii, 1017.
10. Riggs, J., Chemical Process Control. Second Edition ed. 2001, Lubbock, TX:
Ferret Publishing.
11. Richardson, J. and D. Peacock, Chemical Engineering Volume 3: Chemical and
Biochemical Reactors & Process Control, ed. J. Richardson and D. Peacock.
Page 11.1026.15

Vol. 3. 1994: Elsevier.

12. Lipták, B.G., et al., Instrument engineers' handbook. 3rd ed. 2002, Boca Raton,
Fla.: CRC Press. xxxii, 874.
13. Hanahan, D. and R.A. Weinberg, The hallmarks of cancer. Cell, 2000. 100(1): p.
57-70.
14. Raeymaekers, L., Dynamics of Boolean networks controlled by biologically
meaningful functions. J Theor Biol, 2002. 218(3): p. 331-41.
15. Breyfogle, F.W., Implementing Six Sigma: smarter solutions using statistical
methods. 1999, New York: John Wiley. xxxvii, 791.

Page 11.1026.16