CT2011-Control-Talk-cartoon-web

The real deal with process control education

Nov. 11, 2020
We desperately need to close the undergrad preparedness gap. A virtual master's degree led by practice leaders would help, but who would pay for it?

This Control Talk column appeared in the November 2020 print edition of Control. To read more Control Talk columns click here or read the Control Talk blog here

Greg: Many of us with practice experience recognize that there is a gap between graduation and an engineer’s being able to implement and manage control systems. I’ve asked Dr. Russell Rhinehart, emeritus professor at Oklahoma State University School of Chemical Engineering, to exchange ideas about this—why it is and what can be done. Both Russ and I were keynote speakers at the 2020 Texas A&M Instrumentation and Automation Symposium. These conferences are a great way of getting a connection and some synergy between universities and industry.

Russ, there are great examples of how university professors such as yourself have advanced our understanding and practices for basic and regulatory control using the workhorse of the process industry, the proportional-integral-derivative (PID) controller. Some examples I recall include reaffirming the ability of PID to deal with process input disturbances (e.g., McAvoy), filtering noise (yourself), inferential measurements (e.g., McGregor and yourself), tuning (e.g., Astrom, Harriott, and Skogestad), alerting us to the effects of valve and sensor design (e.g., Riggs), and quantifying the effects of lags on the window of allowable gains for runaway processes (Luyben). You and other professors (e.g., Edgar, Luyben, Mellichamp, and Seborg) have helped us develop dynamic simulations and you, in particular, have helped us develop optimization techniques to get help get the most out of our control strategies. 

I think process control labs, particularly wet labs that use an actual industrial control system and simulation labs that use a digital twin, provide a lot of practical hands-on experience. It was key part of the courses that industry graduates like Bob Heider and I did at Washington University when we retired from Monsanto and Solutia. These labs exist in many universities as noted in the Control Talk columns “Hands-on labs build real skills” and “How universities help revitalize process automation professions.” Internships bring the real world to life as exemplified in the Control Talk Column “The education and motivation of engineers.” Pilot plants, such as at the University of Texas, with industrial control systems doing industry research (e.g., wireless control of distillation columns), are a great way of educating students and connecting universities with industry.

What we need is more of the above with a greater emphasis on practical guidance for different applications. We desperately need help on how to encourage and advance innovation using existing tools and the PID controller. We need examples to include all of the automation system dynamics often overlooked, such as valve, variable speed drive, impulse lines, and sensor response considering the many things that can go wrong in terms of design, operating conditions (e.g., flashing and fouling) and installation. This need is distinctly different and challenging for key unit operations, such as biological and chemical reactors, columns, compressors, crystallizers, dryers, evaporators, kilns and neutralizers.

There is a wealth of untapped capability in today's PID algorithm not being put to use, including feedforward, override and valve position control methods, as well as powerful tools such as for adaptive control and external-reset feedback.

Russ: Yes. Omissions in control education and needed contributions of academics to the practice also include safety instrumented systems, dynamic simulation and testing, nonlinear compensation, and criteria for evaluating control systems. 

Greg: Why are we in this state?

Russ: I think that there are two major factors: curriculum time on control topics and faculty expertise.

In the United States, there are no control engineering programs. Within the major engineering programs, there is usually only one course related to control, and there is only so much that can be taught in one course. Further, the students are novices; they struggle to model and understand dynamic processes and solve differential equations or code simulators. So, the elements of instrument selection, dynamic modeling, PID control, and the mathematical language is about all that can be learned. Further, since many professors are working 55 hours per week to remain competitive in research, they need exercises and tests that can be graded by teaching assistants (simple, idealized) and are safe to use to defend a course grade (one right answer, no context). Consequently, the topical coverage is relatively shallow and unrelated to context.

Also, faculty usually do not have any control practice experience. They can understand the mathematics, but not the application context. They are not hired to prepare students for control careers, but to get research funding to support the university ambitions. 

Greg: You mention faculty research. Control engineering practice could use faculty and graduate student help. Critical is the details and guidance to increase confidence and justification for innovation. We need knowledge and examples of how to set up online metrics to monitor performance in real time, addressing the issues of synchronization, inverse response and noise for key unit operations. We need help in diagnosing problems, such as backlash, fouling, resolution, resonance and tuning. We then can move up to optimization—first on a unit operation, then a production unit considering plantwide control strategies such as those explored by Luyben and Skogstad dealing with recycle streams and heat integration. 

There seems to be a lack of recognition of the importance of closed-loop control. The success of PID and model predictive control (MPC) is in its practical, wide-spectrum capability to automatically affect the process. The decades of wasted effort on expert systems failed to realize that telling operators everything that is wrong just cause operational overload and, at best, a non-repeatable and greatly delayed response. Most of those systems were turned off by operators. The few focused successes were also turned off when the real expert left the production unit. PID and MPC are such success stories because they provide great, repeatable and timely response offering better analysis of the remaining sources of variability and optimization of setpoints. The Control Talk blog “The keys to successful process control technologies” explains why PID and MPC are so successful and the deficiencies in some more glamorous technologies.

Russ: It is all about funding. Most university faculty need to bring in research funding to support their career advancement and even summer salary. They follow the money. You can’t blame them. Around the 1970s, there was significant funding to develop a workforce capable of the control transition from analog to digital. But today, the funding for control pales in comparison to that heyday and to the bio-info-nano-cyber visions that are currently shaping academe.

Greg: There are examples of industry and university consortia that could meet this need. Why are there not many?

Russ: When the time was right, I had successful industry/university consortia at both Texas Tech and later at Oklahoma State. Industry provided some funding to support students, labs, student projects and some faculty expenses. I enjoyed the industrial perspective that helped define and shape student projects, and my students were eager and pleased to be hired by our industrial supporters. 

But even in those best of times, the industrial funding would not sustain the program. Consider an elementary cost analysis: Support of a single-faculty viable graduate program (about five students, materials, partial summer salary, overhead) is about $500k/year, and might produce one graduate every other year. It would take ten companies providing $50k each year for a sustainable academic program. The company expenses associated with a representative visiting the program on a semi-annual basis may make the total cost about $100k/yr. And their incentive would be a low chance of hiring a student, and answers to some questions they might have. 

So, to get industrial managers to agree to participate, the fee must be well below that needed for a sustainable program. This means the faculty must find alternate funding sources, and often control is of secondary importance to the primary funding focus. Except for a few remaining places in U.S. universities, there is little academic interest in control practice.

Greg: What about publications? These are important to academics. The big challenge is to make the methods and results useable and understandable by the average practitioner who has very limited understanding of process relationships and dynamics. The use of the time domain and implementation guidance together with an emphasis not on math but on explanation and example would help open people's minds and increase motivation. Academic interns could expand their horizons to undertake this role and publish what they learn.

Russ: You have published several excellent books for the practitioner. So have Wade, Luyben, Marlin, Shinskey, Trevathan, Murrill, Cecil Smith, and others. But, note, all of those authors came out of the practice environment, and understand what is practice-relevant. The International Society of Automation (ISA), a practitioner’s organization sells many of those books—in contrast to the big publishing companies who control the classroom textbook market. 

There are some books in the academic market that have a strong practice connection (Smith and Corripio, Riggs), but the more math-based books seem to be preferred by those teaching the courses. It is not that the material in the most popular academic books is wrong, it just underplays what is important to an application engineer and overplays what is important to graduate research. 

I’ll also note that PID came out of industry, not academic research. So did Cutler’s Dynamic Matrix Control (DMC), now generically called MPC, and as well Richalet’s predictive functional control (PFC). The major advances that have been industrially accepted have been industrially grounded. I think that education of practitioners needs to be by practitioners. 

Faculty need to publish to survive. But since most universities are driven by a research agenda (analysis and possibility) their publications need to relate to their research—not the practice, and not even a development agenda. So, it would be unusual for active faculty to write a how-to book.

Greg: One of the many challenges is that practitioners are overloaded from cutbacks and retirements and emphasis on budget; their mind or schedule is not free enough for innovation. Copy jobs are being done on system upgrades. Also, managers tend to have a business focus and little knowledge or experience in process control. Dynamic examples with metrics are essential to address these and the many other issues. The Control Talk blog “How to motivate management and millennials (M&Ms)” and the Control Talk columns “Invisibility of process control” and “Getting innovation back into control” try to address some of the issues from an industry perspective.

Russ: Yes, and this also makes it difficult for academe to access industrial funding to support building faculty expertise and workforce development. There is also no national agenda to develop advances in control. Those of us with practical control interests are leaving the university, to be replaced by others who can seek the bio-info-cyber-nano funding. 

I do not think that we can restore strong control programs to U.S. engineering education. I don’t think that either industry or government is willing to finance building faculty expertise, graduating prepared students or even creating control engineering degrees. By contrast, many other countries have strong control engineering (or similarly named) degrees. 

Greg: What can be done?

Russ: I think the practice community can shape the list of fundamentals that application engineers need to know. It would not be all of the hundreds of topics such as listed in Trevathan’s “A Guide to the Automation Body of Knowledge”—we cannot expect one course, taught for and by novices, to develop career expertise. We need students to understand the essentials so that they are prepared to self-teach on the job. To do so, practice leaders need to coauthor papers with high ranking academics. Published in the academic community, these papers would provide the visibility and establish the authority so that practice leaders can influence and guide course topics and textbook choices. Also, the practice community needs to accept its residual role—to teach control, and complete the graduate’s preparation.

Greg: I had a wild idea a couple of years ago after creating some digital twin labs and giving some guest lectures at the Missouri University of Science and Technology. I came up with a program for a virtually taught master’s degree in process control that had 12 required and 12 elective courses based on what is needed in the process industry to increase innovation and realize the benefits of basic and advanced process control. The idea was that key resources in the ISA Mentor program, and contributors to the latest McGraw Hill Process/Industrial Instruments and Controls Handbook, 6th Edition, would be the instructors. I realize these people are too busy, but they could provide guidance to recent university graduates with some practical experience—possibly by internships—to become adjunct professors who would actually teach the lectures. Download a proposed list of courses in the online document “Proposed Curriculum for a Master’s Distance Degree in Process Control.” The program would need sponsorship by industry.

Russ: We had a masters of control systems engineering (MS CSE) program here at OSU. I started it about the year 2000, along with several practice-oriented faculty members from chemical, electrical, industrial and mechanical engineering. It was created for distance education, but we also had on-campus students (about a 50/50 mix). 

I taught courses in it and enjoyed the interaction (questions, creativity, application feedback) with the professionals. I also taught a process control lab course. The on-campus students used the ChE Unit Operations Lab pilot-scale facilities, and the off-campus students used whatever they could convince people to let them use at their location (pilot plants, labs, production, significant simulators). 

After a few start-up years, it became financially self-sufficient and was meeting degree production targets. Students greatly affirmed the value of the program, and it was attracting new students and industrial affirmation to OSU. OSU-campus students used the networking as a path to employment. Faculty enjoyed the insight that professional experience brought in, and also enjoyed the extra salary for dealing with distance students (who paid a premium tuition to cover extra OSU expenses). Every indication pointed to success.

But, in about 2009, administration closed the program. It took faculty effort to manage the new program—write reports, recruit, advise, validate, etc. But this faculty effort was not aimed at bringing in research funds (which fed university overhead), getting research publications or generating PhDs (essential for university reputation). The administrators wanted faculty effort focused on those essential things, not service to society. 

We do not survive pursuing our education mission. Presently at OSU, the state funding is less than 10% of that needed to cover operating costs. Tuition and fees only cover about 30% of the cost of education. We lose money on every student we graduate. Imagine the pushback if we tripled tuition costs to the students! We remain alive because of alumni donations (driven by athletic prowess, research excellence, and other brag-able achievements) and faculty research funding. We have to focus on what sustains us. You might also know that NTU, the National Technological University failed. It was absorbed by Walden University, which now does not provide any engineering programs. 

I think that there is a need for a distance education MS CSE program. I also think that today’s funding reality will not make it possible.

Top 10 Signs a Control Course is not the Real Deal

(10) Disturbances are on the process output

(9) Overshoot of controller output is prohibited for all processes

(8) Smooth setpoint response is the primary goal

(7) Near-integrating, true integrating, and runaway processes not addressed

(6) Universal replacement of PID for single loop control

(5) All the dynamics are in the process

(4) All oscillations are caused by too high a PID gain

(3) Performance comparison is for PID tuned too aggressively or sluggishly

(2) Performance comparison is for PID without taking advantage of structures or setpoint lead-lag

(1) Lessons not learned from expert system craze of 80s and 90s

About the author: Greg McMillan