1 November 2006
Educating the engineer
Ready for industry? Control coursework under scrutiny
EDITOR'S NOTE: In the October InTech, University of Texas Chemical Engineering professor Thomas Edgar wrote a story discussing the pros and cons of engineering curriculum in today's universities. This month, a panel of experts shares its insights.
By Joseph S. Alford
Part 2 of a two-part series
During my 35 year career in Bioprocess Control in Pharmaceutical plants and in mentoring automation engineers, I have sensed a growing gap between what is traditionally taught in undergraduate process control courses and the evolving needs of today's industrial chemical plants. In testing this perceived gap with colleagues working in other industries and with university professors teaching process control, it appears the existence of this gap is widely accepted.
Part of the gap stems from process control courses having been originally developed when processes were primarily continuous steady-state linear processes. Many of today's processes are multi-step non-steady-state, non-linear batch processes. Other new paradigms of process control have also developed (e.g. discrete manufacturing). Many new engineering graduates are weak in understanding the nuances of the components of a loop block diagram (e.g. valve sizing, stiction, hysteresis, positioners, non-linearity) and so are weak in performing control loop problem root cause analysis-or in appropriately representing such loop components (e.g. valves) in simulation programs. Too many control engineers start out thinking the solution to almost every control problem is to retune the controller, when the problem is often elsewhere. The most recent confirmation of "the gap" came as I prepared to take ISA's Certified Automation Professional (CAP) exam in 2005. A recommended reference book in preparing for the CAP exam is ISA's Automation Body of Knowledge. A comparison of this book (which contains no content on Laplace transforms, Bode Diagrams, root locus, or Nyquist/Routh stability analysis) is in stark contrast to the current syllabus of the typical undergraduate process control course.
This "gap" was mentioned in a CPC-7 (www.cpc7.org) presentation on Bioprocess Control. The conference agenda also dedicated a half-day session to this issue. A few other articles on this topic have also appeared in the literature in recent years. To summarize this ongoing discussion for InTech readers, a two-part article on this subject was developed. The first, authored by Edgar in the October 2006 issue presented information on this gap and summarizes the CPC-7 half-day session on this topic. In the November issue, a "roundtable" of highly respected process control experts from industry and academia offer their comments in response to the October story. Readers will see there is broad acceptance that a problem exists, but there is a diversity of opinion on remedies. As they say, the devil is in the detail. Regardless, the hope for this two-part article is to spark further dialog, especially in universities, as they consider possible changes to the undergraduate process control course to better prepare their students to be productive in the industrial workplace.
Douglas J. Cooper (email@example.com) is a professor of chemical engineering at the University of Connecticut; the founder of Control Station, Inc., a provider of software solutions for process controller design, tuning, and training; and editor of www.controlguru.com, a Web site focused on practical process control issues for industrial practitioners. His industrial experience includes three years at Chevron Research Co.:
I have taught more than 50 process control short courses to industrial practitioners over the past 15 years. A good fraction of the 500 or so attendees were chemical engineers who had graduated in the previous five years. The thought that so many companies would pay for instruction in a mainstream chemical engineering subject is compelling evidence that there is a disconnect between academic outcomes and industrial needs in the undergraduate process control curriculum.
At the start of my short course, I ask attendees about their background. I have heard the same story with modest variation many times. A reader of my www.controlguru.com site gives clear voice to this tale. I reproduce his opening paragraph (modestly condensed and edited) below:
"I had a course in control systems, and we used ... Matlab/Simulink. All introductory control courses are pretty much the same and involve analysis using Laplace transforms and the frequency domain ... We designed controllers, but in most every case, we started with a known process transfer function in the Laplace domain. We had a couple of exercises where we learned how to model physical system with differential equations and then obtain Laplace transfer functions ... When I started work, I became frustrated ... In the real world, transfer functions are unknown and most design procedures I learned are useless."
Indeed, academia and industry do not even speak the same language. Consider that to an academic, the "output" comes from the process. In industrial practice, it comes from the controller. All of the blame does not lie at academia's door. There are about 40 flavors of the PID algorithm in current commercial products. All must be tuned paying attention to the minor differences, all refer to the common terms with different names, and all provide (virtually) the same performance.
My belief is today's control course should provide a practical tool-box of skills for students entering the workforce while including just enough advanced theory to excite those few destined for graduate study in the subject. To that end, my control course is grounded in hands-on experimentation and the application of theory.
Students learn-by-doing as they explore the dynamics of a half-dozen "real" processes. They design experiments, collect data, and use analysis tools to understand and describe the different behaviors exhibited by these processes. They see that data is noisy and learn collecting a proper data set is among the most challenging parts of a control project. Nonlinear process behavior and its impact on control is something they experience through experimental failure on their way to discovery and success.
During the course, the students actually design, implement, and test a dozen different control algorithms and strategies on these processes. They learn through practice and observation what constitutes "best" performance. They learn for chemical process control, the objective is often disturbance rejection rather than set point tracking, and controllers must be designed accordingly. At this point, the costs and benefits of cascade and feed forward are explored through implementation and testing.
It takes half of the semester to work through the "hows" of practical hands-on process control. We then start over and move quickly through the theoretical "whys," but now the students understand the motivation for learning the theory.
My first attempts at a hands-on approach started in the undergraduate lab, but getting students significant practice with real experiments proved to be impractical. So I began developing a process control training simulator called Picles (which morphed into the more-comprehensive Control Station). A training simulator is visual and hands-on. It provides students with significant real-world experience much the way airplane simulators do for pilots. And modern control installations are computer based, so a video display is the natural window through which the subject is practiced.
And most important, students begin each assignment by confronting a "real" simulated process on their own terms. Though they eventually learn the benefits of passing through the ideal world of transfer functions on the way to a solution, no project begins there. A real-world approach with significant hands-on practice from a training simulator prepares students for industrial practice while also giving them a reason to explore and appreciate the fundamental theory.
Russell Rhinehart (firstname.lastname@example.org) is Head of the School of Chemical Engineering at Oklahoma State University, holds the Bartlett Endowed Chair, and has experience in industry (13 years) and academe (21 years). He is an ISA Fellow and is Editor-in-Chief of ISA Transactions:
I think that the question should be, "How can we create control as an undergraduate discipline?"
Control is a career, and one course is inadequate for preparation. In the process industries control includes: process and instrument dynamics, MIMO interactions and modeling, noise filtering and spurious signal rejection, PID-type feedback strategies, safety/constraint strategies, model predictive control, reliability, safety instrumentation and determination of SIL, adaptive tuning, supervisory set-point assignment by optimization of plant-wide models, batch-to-batch recipe adjustment, statistical decision making, calibration of measurement devices, specification of valve size and characteristic to permit turn-up and turn-down, signal transmission mechanisms and protocol, standards for batch and security ... The subjects are diverse and complex when one considers the mechanical devices of robots and vehicles and adds vision to the sensing, autonomy to the control, and vibration to the control objectives. Added to this diversity is the control of electrical power production and distribution, of microwave transmission devices, of pacemakers, and the structure of the computers that do "control" for everyone's application. Add to this air traffic control, river and reservoir control, supply chain management, informatics, knowledge management, and so on ...
As a career, control is critical to product quality, personal and manufacturing safety, and to resource and asset management. The techniques of control extend across disciplines and are grounded in fundamental math and science. Further, the fundamentals of measurement and control support the developing opportunities in knowledge discovery and action management. Control is a career of national importance.
Control is not a subset of the primary disciplines. The dynamic nature of processes, information processing techniques, sorting fact from noise and spurious events, detecting faults, optimizing over future time, minimizing variability, risk analysis, and adapting decision-making strategies is common to control engineering regardless of the primary discipline. The math that supports all of this includes calculus, differential equations, linear algebra, propagation of probable events in complex systems, nonlinear solutions, numerical implementation, encryption, assessing sufficiency of models, optimization, statistics, finite impulse response models, neural networks, fuzzy logic, etc. Neither the topics, nor the required math skills are comprehensively covered in any of the primary undergraduate disciplines.
We can barely get chemical engineers to understand process dynamics and advanced classical PID applications in the one undergraduate course. One undergraduate course in any discipline cannot produce control engineers. There is a PE category for "Control Engineer," but no U.S. undergraduate programs to support it.
For these reasons, it appears the U.S. should have undergraduate degree programs in "automation engineering" or some such name. Many other countries have undergraduate programs that grant engineering degrees in control, automation, instrumentation, or under similar names. It would seem there is justification for the U.S. to do the same. However, higher education is not an income-generating enterprise, and our university system is dependent on state and federal funds and endowments. Unfortunately, state support for schools is diminishing. This is a national trend, which has increased the dependency on research income. Universities cannot justify the start of a new BS program because there is a need for the skill. The program needs to self-supporting from endowments or research-or have such a direct and strong relation to economic development for the state legislature to create something new to subsidize.
ISA may be the society of people to organize industry leaders to provide the government influence and economic justification necessary to create undergraduate control engineering programs and to establish the profession.
Alternately, where there is an existing interdisciplinary strength and faculty and courses in place, a university can offer, relatively easily, a MS degree in a subject to meet the education need. Oklahoma State University adopted this path and offers a MS degree in Control System Engineering, both on-campus and through distance education.
Greg McMillan, CAP, (greg.mcmillan@emersonProcess.com) is a retired Senior Fellow after 33 years at Monsanto Co. and Solutia Inc., and he is an affiliate professor at Washington University in St. Louis and a consultant with CDI-Process and Industrial in Austin, Tex. McMillan is an ISA Fellow and the author of 16 ISA books:
Terry Tolliver and I have taught a course on dynamic modeling and control at Washington University (WU) in St. Louis since 2002 that is a requirement for a degree in chemical engineering. The course uses an industrial virtual plant and the ISA book Advanced Control Unleashed. The students are very computer literate and pick up on the use of industrial software from just a few screen prints put into the laboratory exercises. The knowledge gained is generally applicable since the function blocks are based on Foundation Fieldbus used in millions of devices and by over a hundred manufacturers. The configuration environment is also consistent with the international standard IEC 61804. The students learn how to intelligently discuss and use an industrial process simulation, DCS, and data historian that form a virtual plant on their desk. There is companion course taught by Bob Heider where an actual hardware version of the same DCS is used to control the temperature, pressure, and level of vessels in a hardware lab. The three professors have a total of more than 100 years experience in industry.
Each chapter in Advanced Control Unleashed starts with a section on "Practice," continues with sections on "Opportunity Assessment" and "Application," and concludes with "Theory." The strategy is to provide the relevance and practical considerations before getting into the theory that offers a deeper understanding. For example, in Chapter 2 - "Setting the Foundation," the student gets an overview and perspective, list of opportunities, examples, application detail, and rules of thumb before getting into the theory where the focus turns to the setup of the differential equations for the material and energy balances to enable the student to learn the source of process time constants and gains in terms of process parameters. The students are not asked to solve or integrate these equations. Instead, they graphically create a dynamic simulation of processes for unit operations commonly encountered on the job. Blocks for filters, dead times, noise, periodic disturbances, and backlash and sticktion are added to make the challenge of process control more realistic. Additionally, the students configure an actual control system that can be downloaded into a real DCS. The students apply industrial embedded tools for auto tuning, statistical analysis, and model predictive control (MPC). The course centers on time response because this is what they see on the trend charts in the control room, but there is a session to show how to go from the time domain to the frequency domain.
When I recently went back to WU and gave a guest lecture on the use of PID and MPC for fed-batch control of a fermenter, a student asked "What is a batch?" I knew that students were taught to think in terms of a steady state and the material and energy balances on a Process Flow Diagram (PFD) for continuous operations, but I didn't fully realize the implications until the question.
I have had chemical engineers in industry ask, why do you need a PID or MPC when you can just set the flow shown on the PFD? In fact, the batch sequences defined by process engineers today often try to set a predetermined step sequence of flows instead of using feedback control to sort it out. I have also had experienced instrument engineers ask why do you need a Coriolis density measurement when the composition is constant as shown on the PFD? I also see ads for pressure and temperature compensated differential pressure orifice meters that claim to offer an unqualified mass flow measurement. If only the composition in all the pipelines were constant. This would sure make life easy. Product quality would be a non-issue. Obviously the importance of dynamics and disturbances for process measurement and control is often missing in action.
In a batch process, the product concentration follows a profile. In some case there is also a temperature profile, and in almost every case where a PID or MPC is used, the transfer of variability for a constant set point means there is a profile in the controller output. This understanding is lacking when chemical engineers are taught to think steady state. The lessons from batch would also be useful for the automated startups and grade transitions in continuous operations.
To improve the WU course, I would add a one-hour introduction to the ISA batch standard and basic batch sequencing so the student gets familiar with basic batch concepts and terminology. I would add another one hour session on fed-batch control to show the implication of the lack of a discharge flow and the consequential diminution of process self-regulation on temperature, composition, and pH dynamics and control. The examples of choice would be jacketed chemical and biological reactors. This means two class hours must go. I would delete two of the five hours spent on PID controller tuning. I would retain the last two hours on an introduction to MPC because it drives home the concept of process time constant, gain, and dead time and its use in the most successful advanced control tool in industry. I would keep the one-hour session on frequency response because some tuning and performance tools show Bode and Nyquist plots. However, I could be convinced to replace this with an introduction to multi-way dynamic principal component analysis of the batch profiles of manipulated variables and process variables that are not controlled. Even better would be making the required course four hours per week.
Diana Bouchard (email@example.com) retired from a 26-year scientific career in the Process Control Group at the Pulp and Paper Research Institute of Canada (Paprican). She has also been active in ISA leadership and most recently served as vice president for Publications:
Well, I am neither a control engineer nor a university professor, so to some extent I feel I walked in the door of this august assembly by accident. However, I believe I can use my experience to inform this discussion from some different angles.
Let me start with a question (as Rhinehart also did): "How can we give control engineering a future?"
I don't know the experience of others in this discussion, but in Canada, it has become more and more difficult to maintain enrollments in many chemical engineering programs because people see process plants closing, engineering staffs cut to the bare bone, and their future employment prospects in the field dropping off. Paradoxically, the advancement of measurement and control technology has in many cases made plants operable with fewer engineers and technicians. For these reasons, many professors have to expend time and energy in promotional efforts to get students into their programs. Anything we do to improve control engineering education will be for naught if there is no one to teach.
What this implies, in my view, is a need for a closer liaison between university engineering programs (control and otherwise) and the people who will employ their graduates. These discussions must include not only large continuous-process manufacturing plants, which are directly employing fewer and fewer engineers, but also consulting engineering firms, batch processors, and discrete manufacturers. A technical-vocational college I know in Montreal goes so far as to have a few industry representatives on its board and hold review meetings every few years with their industrial partners to go over their program and course offerings and fine-tune them to meet industry needs. Could this approach be adapted for a university engineering program? It would certainly reduce the element of unpleasant surprise when employers find their new engineering hires ill-prepared to work in their plants.
Another implication is we, the automation and control community, need to do a much better public relations job so more people appreciate the vast scope of human activities where control engineering plays a role. This could include, for example, books on measurement and control oriented toward the lay person, or teaching modules to present the essentials to K-12 students and get them started thinking of control engineering as a potential career.
Control has become a lot bigger than continuous process control. If control engineering does not expand its scope to include batch and discrete processes, it will remain rooted in industrial sectors, many of which are stagnant or declining in North America. By contrast, many of the sectors that use batch and discrete processing (e.g. pharmaceutical, vision systems) are growing. There is no hope of keeping up enrollments and getting control established as a discipline if we confine ourselves to stagnant or declining industries. However, I fully appreciate the challenges noted by others of cramming ever more topics into a finite number of class and laboratory hours.
I agree that control engineering is way too big to be covered adequately in a one-semester course. In my own experience, introductory courses don't make you an expert; they enable you to have an intelligent conversation with someone who really knows the field. Maybe it would be fruitful to review the control engineering curriculum from this vantage point. What would a chemical engineer need to know to have an intelligent conversation with a real control expert?
I can imagine control engineering being structured as an introductory course (first semester) followed by one or more courses in the second semester. Someone who just wanted a nodding acquaintance with control wouldn't have to take a second-semester course; someone who wanted to learn more about batch processing, discrete automation, or control system design would take a second course. Of course anyone who wanted to become a real control specialist would go on to get a Master's or Ph.D. degree in an appropriate discipline. Almost all the people I have known whom I would describe as "control specialists" have such a background.
Interesting that in many universities outside North America, control engineering is a full-year course. In the pulp and paper industry, it is widely believed control is more effectively practiced, and advanced control concepts are more often applied, in Europe, particularly in Scandinavia, than in North America. Some of this may be due to mills' hiring control people with advanced degrees. M.Eng. holders are widespread in Scandinavian pulp and paper mills, almost unheard-of over here. The difference does seem to show.
Data variability and uncertainty are crucial topics, not only for control engineering but for other areas of engineering. Many students believe you can set a variable to a value and it stays there automatically. Use of computers makes this misconception worse because the numbers you input to a computer don't wander around seemingly of their own volition the way measurements in a process plant do. This makes it all the more important to introduce data variability and uncertainty concepts early in the engineering program. Students will need a little time to get themselves into the head space where numbers aren't necessarily stable or trustworthy and don't necessarily agree with other plant measurements.
Establishing control as an engineering discipline is a longer-term goal. For now, it is essential to improve control engineering education along the lines suggested in this discussion.
Cecil Smith, Ph.D., PE, (firstname.lastname@example.org) has over 40 years experience in process control, the first 13 of which were on the LSU faculty in various positions, including Professor of Chemical Engineering and Chairman of the Department of Computer Science. Since leaving academia, he has provided extensive "continuing education" training and consulting services to industry in continuous and batch processes:
The "process control" course as generally taught in academia is the result of applying linear systems theory to processes. We need a fresh start, one based on "you have to understand the process." Processes involve basic mechanisms such as heat transfer, mass transfer, vapor-liquid equilibria, reactions, etc. The process control course should start with these basic principles and develop control strategies (such as P&I diagrams) for the process. In such an approach, LaPlace transforms will never arise, and indeed LaPlace transforms and all derivative topics are irrelevant to the practice of process control. One will also find steady-state issues are far more important than dynamics. With such an approach, maybe we can finally make progress on integrating process design and process control. Process control is definitely a subject within the discipline of chemical engineering, and the undergraduate course must reflect this.
The focus should be on control applications whose solution requires more than a superficial understanding of the process. Based on my experiences, the following aspects deserve attention:
- Limitations on operations, especially those that arise within the process itself.
- Nonlinearities from equipment such as heat transfer.
- Interaction between process variables. From my experiences, it is all of the steady-state variety.
- Propagation of variance from one variable to another.
None of these have anything to do with dynamics. Contrary to popular belief, dynamics are the root cause of very few process control problems. Tuning difficulties are symptoms that appear when issues such as the above are not properly reflected in the P&I diagram.
Dynamic simulations based on basic process mechanisms (not s-domain transfer functions) are a valuable teaching tool. I use these routinely in the continuing education courses I teach. With encouragement from Monsanto many years ago, I developed a PC-based simulator with the "look and feel" of commercial control systems, a variety of PID algorithm options, a versatile set of control blocks for advanced control, process graphics, and versatile trending. It makes for an excellent learning experience, but we need to be realistic about the use of dynamic simulation in practice. We have the technology to dynamically simulate our processes, but when will we do it? The day we have to start up with every loop tuned and in automatic may come someday, but probably not in my lifetime. In the meantime, we are missing opportunities to extract information from our steady-state simulations to develop better process control strategies. Dynamic simulation can only assess a proposed strategy; our priority must be on control strategy design. In this regard, steady-state is far more important than dynamics. And this even applies to batch processes, where the root causes of problems usually prove to be factors in the steady-state relationships (such as a change in heat transfer rate of 50:1 during the batch).
Here are my specific suggestions:
- Teach the basics. I have no problem with a theoretical course, provided the theory is relevant to engineering practice.
- Focus on developing P&I diagrams. Get this right, and everything else will fall into place. Stress the steady-state relationships and the steady-state simulation model as the primary source of the process understanding required to develop a good P&I.
- Do everything in the time domain. Do not even mention LaPlace. I developed a CBT on process control with an excellent search facility. I take great pride in demonstrating that the word "LaPlace" is nowhere to be found.
- Explain PID in the time domain, including issues such as position/velocity, parallel/series, reset windup (students should recognize what is required for the usual windup protection to be effective), initialization/tracking (including reset feedback), etc. Explain how PID works in the time domain and even technicians with only a superficial understanding of integrals and derivatives will get it.
- Introduce time constants and transportation lags. Explain how the PID tuning coefficients are related to these parameters and the impact of transportation lags on loop performance. If there is no time for inverse response, so be it.
- Deemphasize tuning. If P&I is right, tuning will be straightforward. However, this does not mean that loops are tuned properly. Explaining how to recognize when a loop is poorly tuned would be more useful than discussing tuning techniques and automated tuning.
- Definitely include final control elements, mainly pumps with variable speed drives and control valves. (About a third of the problems brought to my attention are related to the control valve.) Converting inherent characteristics to installed characteristics is a good use of a flow simulation. Linear and ideal behavior should be the objective, but students need to understand if the fluid is caustic slurry, this is unlikely to be realized. Stay out of control valve sizing. Oversized control valves are a tradition in this industry, so devote the time to explaining their consequences.
- Use dynamic simulation as a teaching tool, but make the students aware in practice most problems are solved without a dynamic simulation.
- Batch needs to be integrated throughout the chemical engineering curriculum. It is not the responsibility of the process control course to introduce students to batch processes.
- Focus on basic regulatory control, and do it well. Topics such as optimization and model predictive control are important, but there is only so much that can reasonably be covered in a three-hour course. An M.S. or Ph.D. in chemical engineering with an emphasis on process control should be available to those wishing to make process control their career.
- Some propose to essentially integrate process control and process safety. Especially as pertaining to personnel safety, I prefer the "old school" approach of keeping them separate-separate equipment and separate people. The process controls should not take an action that would elicit a reaction from the safety system, and this aspect is appropriate for the process control course. But with the objective of running the process in the best possible manner, the process controls are relatively complex and frequently modified. Are there defects? Very likely, despite our best efforts. So let's assume there is some sequence of events that will cause the process controls to do something utterly stupid. Safety department, your job is to stop it.
Chemical engineers are the crème de la crème of engineers, so take the high road. They deserve education, not training. Stay out of the systems (bits and bytes) issues. For chemical engineers, this stuff is pretty simple (after all, every vendor claims to be "user friendly"), and the trend is for this work to be done in cheap labor markets. And stay out of ladder logic, discrete manufacturing, and all other topics that require no more than a superficial understanding of the process.
In the October story (www.isa.org/intech/20061003), Edgar mentions Dr. Brian Ramaker, now retired from Shell Oil. I will never understand why companies let people with such expertise walk out the door. Over the years, Ramaker made millions for Shell, applying process control in refineries. Unfortunately, this is only one example of what seems to be a common practice throughout industry. But there is a larger question: Could this reflect the attitude of today's management toward automation? Do they view automation as enhancing profits and deserving further investment, or is automation an unavoidable cost that must be reduced at every opportunity? Upon retirement, does the person who chose a career in automation get the accolades that make new hires want to pursue the same career path, or is the retirement "party" a perfunctory event held in the company cafeteria? These are not official corporate functions under the control of management, but people are watching and management can exert influence if it so chooses. Take a key from Louisiana politics-look at what they do, not what they say. But then such practices by industry provide opportunities for others. In Ramaker's case, his expertise would be invaluable in developing a real process control course, in fact, a new course with a different number so those who have taken the previous course could take the new one and really learn something about process control.
James B. Riggs (email@example.com) has been a professor of chemical engineering at Texas Tech University since 1983. He co-founded the Texas Tech Process Control and Optimization Consortium (www.che.ttu.edu/pcoc/) in 1992, has over five years of industrial experience, and has over 80 technical publications and books on process modeling, control, and optimization:
The issues Edgar raised with regard to modifying the undergraduate process control class are associated with the classical question of theory versus practice in engineering education. Clearly, a course that deals only with industrial practice of process control will not provide our students with the fundamental understanding of the subject necessary to understand and adapt to the changes that occur over the lifetime of their career. On the other hand, a course based on applied mathematics that deals only with the theoretical aspects of process control provides most students few, if any, industrially relevant skills. Moreover, such courses alienate many students toward process control due to the pure mathematical presentation of the material. Therefore, the key, in my opinion, is to provide control courses that provide basic industrially relevant skills while also providing a fundamental understanding of process control and process dynamics. Furthermore, the challenge to this approach is to fit it into a single undergraduate class.
Using control theory to provide fundamental insight into the behavior of process dynamics and control systems is essential for providing our students with the conceptual knowledge for their future careers. Some of the key important to a sound conceptual understanding of process control are understanding the unique characteristics of proportional, integral and derivative control, the concept of stability, the different behavior for linear and nonlinear systems, the characteristics of positive feedback, stabilization of open-loop systems by feedback action, and the effect of sensor filtering on closed-loop dynamics. While it is possible to address these issues in the time domain, the Laplace domain can be used to derive results for general cases while time-domain approaches can usually only be used to demonstrate behavior for specific cases. Moreover, important terminology for the process control field comes directly from the theoretical analysis of linear control systems. On the other hand, it is a harder to justify an in-depth study of frequency response analysis, although the concept of the frequency dependency of disturbance on closed-loop performance is important.
My objective when teaching an undergraduate process control class (or writing a control textbook) has been to attempt to also teach the students the skills necessary to function as entry level control engineers (e.g., maintain regulatory control systems for a portion of a plant). To be able to perform these duties, I concluded they must be able to troubleshoot control loops, tune control loops, and make some basic control design decisions. To troubleshoot a control loop, a student must understand P&ID diagrams and the control relevant aspects of the hardware (control computers, actuator systems, and sensor systems) that comprises a control loop and know how it fits together as well as have a systematic approach to control loop troubleshooting. The study of controller tuning should be directed toward tuning methods that are industrially relevant (e.g., a tuning method that requires a process model to use is not usually efficient in an industrial setting). To this end, I present field tuning for fast control loops and autotune-based tuning for the slow loops. Finally, students should be able to select the proper mode of a PID controller and know when and how to use advance PID approaches (e.g., cascade, ratio, feedforward, and inferential control). It is my experience that good control practice for SISO systems is a prerequisite for the control of MIMO systems, and that the extension from SISO systems to MIMO systems is relatively direct once the control of SISO systems is mastered. Therefore, I emphasize SISO system control and overview MIMO issues (e.g., coupling, disturbance sensitivity, and dynamic differences) near the end of the class. In addition, our control laboratory emphasizes the control relevant skills (e.g., instrument calibration, controller tuning, cascade, ratio and feedforward implementation, and control loop troubleshooting), allowing the students to practice these tasks on a laboratory DCS.
Unless faculty are willing and able to effectively integrate control theory and industrial control practice into undergraduate control classes, these courses will not fully serve the needs of today's students and industry. It is very difficult for a professor to know what is industrially relevant and what theoretical concepts are relevant to industry unless they have direct exposure to the industrial practice of process control.
Ron Artigue (firstname.lastname@example.org) is professor of chemical engineering at Rose-Hulman Institute of Technology. He has taught process control for over 25 years and has practiced as a consultant with Pitman Moore and Eli Lilly and Co.
Atanas Serbezov (email@example.com) is an associate professor in the Department of Chemical Engineering at Rose-Hulman Institute of Technology. He has experience in process control as a practitioner (Honeywell, Inc., Praxair, Inc., and Eli Lilly and Co.) and as an educator (University of Rochester and Rose-Hulman Institute of Technology):
At Rose-Hulman Institute of Technology, our required process control class covers both process dynamics and control. Our course goals are that our students will have a fundamental understanding of process dynamics, be able to design and tune simple control systems, have had practical experience for immediate application as process engineers, and have established a basis for advanced study.
A number of topics we still teach are being suggested as irrelevant, LaPlace Transforms (LT) and Frequency Response (FR) to name a couple. Students are introduced to a number of topics that many have bemoaned as being useless because they "never use them (directly) in their job." For example, Bode diagrams and frequency response (FR) fit this description for many. However, Bode diagrams are an excellent way to visualize frequency response. And if one truly wants to fundamentally understand how to filter out process and environmental noise and yet retain real process dynamical information, it is important to understand FR. Moreover, many fear using derivative control, since they do not understand how noise and its negative effect on derivative control can be eliminated with the proper filter. Here too, the open loop FR can help one to visualize the effect of filters on control system performance. With expanding use of computers, it is also important to complement computations with estimates that show that results are reasonable. The Bode plot is a very simple way of exploring many properties of a system. When computers are used to generate Bode plots, FR concepts can be taught effectively in just a couple of lectures.
Control (Engineering) education is a path function and not just a change in state. The topics one studies, the thinking of the discipline, the problems, and solutions worked on, all contribute to the excellence of the educational experience. Business executives are problem solvers, and the calculus or process control classes they took at school helped them become problem solvers, never having to solve a differential equation again notwithstanding.
At Rose-Hulman, we also emphasize the use of simulation tools like Control Station and MatLAB/SIMULINK. Students are taught how to use Control Station to analyze, model, select, and test control system parameters for both general transfer function model systems and simulated nonlinear systems. This tool and others allow for the direct import of real process data so students can safely test their control system designs usually in faster than real-time simulations. These tools are essential to help students learn and apply the fundamentals of process control.
Process control today is implemented on complex networked systems where computers (PLC or DCS) are connected to external devices, which perform sensing, actuation, and control. Due to the intricacy of the software, even simple control algorithms (such as PID) are not straightforward to implement. This is why it is important to introduce the students to the architecture of the modern control systems and point out the IT aspects of current process control practice.
At Rose-Hulman Institute of Technology, students have the opportunity to use a DCS (DeltaV) as part of the Unit Operations laboratory. The architecture of the DCS system is discussed in the introductory control course. Also discussed in detail are the components of a typical feedback loop, such as sensors, transmitters, control valves, and actuators. Show-and-tell demonstrations (in class or in the UO laboratory) help uncover the mystery of the control hardware. Sensors, transmitters, and control valves are no longer abstract symbols on a P&ID, but electro-mechanical systems of complex design.
Below are the changes in the control course at Rose-Hulman Institute of Technology that have been introduced in recent years:
- Computer programs (Maple or Mathcad) are used to perform LaPlace transforms and frequency response analysis. This has allowed the instructor to spend less time on the mechanics of the transforms and focus on the fundamental process control applications. This has also allowed the instructor to include extra material in the course.
- Computer simulation software is used that allows students to collect realistic process data, analyze the data, design, and tune appropriate controllers.
- Students are introduced to the computer and networking aspects of modern process control implementation by examining the architecture of a DCS system.
Tom Marlin (www.chemeng.mcmaster.ca/faculty/marlin/default.htm) is a professor of chemical engineering and director of the Advanced Control Consortium at McMaster University in Hamilton, Ontario, Canada. He worked in industry for 16 years:
Let's start with the goal statement, "to operate equipment they design, how to control processes and understand the dynamic nature of processes." I would add, "safety, reliably, and profitably over a range of process conditions," which introduces designing process structures and selecting equipment with the capacity and flexibility to achieve good dynamic performance.
The most striking item in Part I is the great disparity between the topics chosen by academics and industrial practitioners to achieve the educational goal. I believe the reason for the difference is a more holistic view of the problem taken by industrial practitioners. An appropriate academic view also exists; in Chemical engineering, it is termed Process Systems Engineering (PSE). PSE includes process control, but it integrates control with modeling, statistics, and optimization. Clearly, this broad range of topics cannot be addressed in one course. The appropriate educational program offers a suite of integrated courses that enable an undergraduate to build expertise in PSE. We at McMaster have five required and six elective undergraduate courses addressing PSE. For example, process control topics covered in complementary courses include: alarms, safety valves, and safety interlock systems in the design course; model-predictive control (MPC) the advanced control course; economic optimization of plant operations in the optimization course; and experimental design and statistical process control, including multivariate methods, in the advanced statistics course. Extensions to discrete manufacturing and PLCs are available in a Mechatonics course from our Mechanical Engineering Department. Details of this popular program will be presented at the American Association of Engineering Education Annual Meeting in 2007.
A second issue is the appropriate balance between quantitative calculation skills and qualitative process understanding. We should strive to teach qualitative methods that can be applied and are testable. For example, we can teach students to evaluate a closed-loop dynamic response and recommend tuning changes that will improve performance. Also, students can be able challenged to estimate the fastest possible set-point change that can be achieved for a process from knowledge of the process dynamics. Finally, laboratories provide experiences that are usefully recalled throughout a professional career.
Regarding the undergraduate course, too many topics in a course results in little learning, other than short-term memorization. I propose we integrate process systems topics throughout the required courses and offer a set of focused electives for those students who wish to build expertise and are considering a career in PSE.
As Edgar rightly points out, continuing education is essential for many topics beyond the typical undergraduate's background. Consultants and vendors offer many valuable courses, and some universities are belatedly responding to this professional need. McMaster is offering new graduate education programs tailored to the industrial practitioner that complements (but does not replace) our research-oriented graduate programs. Practicing engineers are eager for opportunities to improve their skills and knowledge, but many North American companies are regrettably unwilling to invest in their employees, which is "penny wise and pound foolish."
Vernon Trevathan (firstname.lastname@example.org), consultant, is an ISA Fellow, an ISA vice president with responsibility for training and certification, and the editor of ISA's Automation Book of Knowledge. He worked for Monsanto/Solutia for 35 years in process control and for an engineering company in non-process automation:
Loops continue to be an important area where chemical engineers can make contributions to process control, and learning how to do so requires strong emphasis on these basic topics:
- Time-domain responses, what type of process situations produce them and how these responses impact controllability. (Computer simulators can help, and Greg McMillan's approach seems very good.)
- How these responses relate to the process design such as how effective dead time in a mixed vessel relates to the intensity of the mixing. (Most chemical engineers seem to have a very poor understanding of this.)
- How the various response situations are controlled with a PID controller and very basic controller tuning.
And some familiarity with the following topics:
- Process and instrumentation diagrams
- Loop selection and interaction
- Basic transfer functions
- Advanced regulatory control (cascade, feedforward) and advanced control (model predictive control, fuzzy logic)-when they are needed and how they work
- How various unit operations are typically controlled (Experiments with actual process equipment would be useful.)
- Control hardware: DCS, transmitters, analyzers, and valves
But, control topics have expanded dramatically over the past 10 years. To make the most useful contribution to process control installations today, chemical engineers need to know more than just loops. (Less than 1/4 of the book, A Guide to the Automation Body of Knowledge, is devoted to all of the topics listed above.) Thus the following topics should also be covered at least enough to give some familiarization:
- Elementary logic implementation
- ISA-88 batch control
- Safety Instrumented Systems
- HMI display design concepts
- Alarm management concepts
Chemical engineers should have an understanding of dynamic modeling and simulation, though this would ideally be covered in several courses and not relegated to the process control course. All engineers should have an understanding of Laplace transforms and frequency domain representations, but this should also not fall to a controls course. However, frequency domain analysis techniques (bode, Nyquist, etc.) as well as discrete techniques (z-transforms) have proved to be rarely useful in industry except in isolated areas such as motion control; and I believe a better understanding of process control can be achieved by focusing on the time domain-so I don't think these techniques deserve any treatment in an undergraduate process control course. I am not impressed with undergraduate work on large commercial computer software-simulation, Model Predictive Control, or DCS configuration-except maybe as a brief demonstration to know it exists; but I would prefer the available time be spent on basic principles no matter how limited this must be given the time constraints.
Most of the undergraduate chemical engineers will not be responsible for complete control system installations, and most will not work with controls outside of the chemical process area, so they will not need to know the bulk of the other 3/4 of the Automation Body of Knowledge. Those that do will need to learn about communications, system integration, Manufacturing Execution Systems, programming, motors and drives, electrical installations, their country's codes and standards, and many other topics. Some combination of more college courses, short courses after college, a passion for all things computer, and years of self-study will be required to achieve that. Electrical Engineering courses cover some of these topics, and most automation professionals have that background. In addition, since most full time automation professionals today work for suppliers, engineering contractors or system integrators rather than end-user companies (and that trend is advancing), it is probable their work will include non-process application areas, such as packaging, machine automation, parts manufacturing, quality inspection, discrete material handling, Web processes, and motion control. One or a few undergraduate courses in control cannot cover even a small fraction of this material, and the process control course in chemical engineering should not even try. Since a degree in engineering does not usually indicate much automation knowledge, the Certified Automation Professional (CAP) credential is proving to be very important to demonstrate that knowledge. (However, I agree with Rhinehart that we need undergraduate majors in Automation Engineering.)
Venkat Venkatasubramanian (email@example.com) is University Faculty Scholar and professor of chemical engineering at Purdue University. His research contributions are in process fault diagnosis and supervisory control, hazard analysis, informatics, and complex adaptive systems, using knowledge-based systems, neural networks, genetic algorithms, and mathematical programming approaches:
Edgar and Dr. Joe Alford, and other colleagues, have raised important concerns about the typical undergraduate process dynamics and control course currently taught in our universities. They also offer valuable suggestions for revamping the course. In the following, I would like to offer my general perspective on this topic, and make some suggestions.
Let me start with a little of the background of control education at Purdue to provide a context of my perspective. For decades, Purdue has had a strong tradition in process systems engineering research and education in general, process control in particular, as exemplified by the classic textbook in control authored by Purdue faculty members Coughanowr and Koppel. I have had the privilege and pleasure of teaching a course on process dynamics and control to senior chemical engineering undergrads for many years at Purdue University. The course has lectures three times a week and a computer simulation lab component, based on MATLAB/SIMULINK modules, that meets once a week for two hours. Our senior class has averaged about 100 students in the last two decades. The lectures are delivered to the entire class, but the class is divided into smaller groups of about 15 students each for the lab sections.
As we all recognize, we need to balance the educational priorities as seen by us academics with the needs of the industrial practitioners in structuring our courses. In this regard, I think it is important we remain cognizant of the difference between education and training. Simply stated, education is a broader learning activity fostering critical thinking. It is about asking fundamental questions and identifying and formulating problems. On the other hand, training is a narrower activity focused on teaching specific solutions and tools to well-formulated problems. Such a simplification has its limitations, to be sure, but it is still useful as it articulates the difference in the emphasis between asking questions and knowing answers.
I strongly believe anyone who graduates with a chemical engineering degree must be required to have learned process dynamics and control as well as process/product design. Without systems courses such as design, control, and optimization, I believe their education is quite incomplete. After all, it is the systems thinking that sets an engineer apart from chemists, biologists, physicists, and mathematicians. However, my informal survey of the senior class over the last two decades at Purdue suggests only 10% or less of the graduating class ended up with control-related jobs in the industry. The others, who did not go to graduate schools, went on to pursue all kinds of other career opportunities, in the process and other industries. This naturally raises the question of how much should we emphasize specific control tools, hardware, and software issues, which are more practical training oriented over the broader fundamental control concepts, methodologies, and techniques.
I am not defending the status quo in undergraduate control education, which obviously needs revamping. However, I think we should be careful about drastically altering the balance between education and training as we go about revamping the syllabus. In my opinion, the best solution is to offer a two course sequence in dynamics and control. The first one will be a required core course that covers the broad fundamental concepts, methodologies, and techniques. The second one, offered as an elective, is more specialized, aimed at students who are going to pursue a career in control. In this course, one can get into the specific tools, hardware, software, and other such details of industrial practice. However, since this solution is unlikely to be adopted by most schools, we are stuck with a single course where we have to comprise on competing objectives.
I agree with many of the changes proposed. However, one problem I have with the typical control syllabus is it does not stress the connections between control and process safety and process hazards beyond some perfunctory remarks. I try to address this deficiency in my course by motivating the course through case studies of major chemical plant accidents such as the Bhopal Gas Tragedy, Piper Alpha Disaster, Flixboro accident, and so on. I try to teach process control in this broader context, which then naturally leads to exposure to supervisory control concepts, alarm management, relationship between design and control, and so on. Of course, I don't have the time to treat any of these topics in depth, but it at least makes the course less of an exercise in applied mathematics and makes it more relevant to industrial practice.
Another technique I use to keep the control industrially relevant is to have industrial practitioners to deliver guest lectures to the students. For instance, Alford has given lectures on fermentation control and batch control to our students. I happen to think batch control should be a part of the revamped curriculum. I also recommend the idea of using MATLAB/Simulink like tools for interactive open loop and closed-loop simulations of different kinds of processes. We have had success with this approach for a number of years. In addition to the simulations, I would like to incorporate actual experiments to the course. We are not doing this yet.
In summary, the time has come for the undergraduate control course to be revamped. However, in doing so, we need to balance the pedagogical priorities and practical needs in order to best serve a wide cross-section of our chemical engineering students.
About the Author
Joseph S. Alford, Ph.D., P.E., CAP (firstname.lastname@example.org) is an engineering advisor at Eli Lilly and Co., having spent most of his career in bioprocess automation. His primary contributions have been in process mass spectrometry, advanced control and data analysis, alarm management, and use of on-line artificial intelligence technologies. He has received national awards from ISA and AIChE for his work, is a Distinguished Alumnus from both of his alma maters (Purdue and the University of Cincinnati) and is an ISA Fellow.
- There is a growing gap between what professors traditionally teach and the needs of today's plants.
- Experts agree: Problems exists, but opinions vary on a solution.
- One suggestion calls for more hands on training: "Students learn-by-doing."