The following discussion is part of an occasional series, “Ask the Automation Pros,” authored by Greg McMillan, industry consultant, author of numerous process control books and 2010 ISA Life Achievement Award recipient. Program administrators will collect submitted questions and solicit responses from automation professionals. Past Q&A videos are available on the ISA YouTube channel; you can view the playlist here. You can read posts from this series here.

Greg McMillan’s Question

What have you seen and learned from the past, and what can we do to provide a better future for process control?

Michael Taube’s Perspective

Let’s first set the stage: I am reminded of a blog by Charlie Cutler (c. 2016, I think) that, sadly, no longer seems to exist (see link below), wherein he laments the lack of enthusiasm for computer-based control. In the blog, he comments on how conference sessions on “advanced” control (1980s-1990s) were standing-room-only events; today, it’s hard to get people to even SPEAK or present at such sessions, much less attend them! Someone who responded to Charlie’s blog theorized that prospective computer-control types were being pulled to more “sexy” jobs in Silicon Valley; that analysis also likely applies today as it relates to AI, “machine learning” (ML) and/or “deep learning” (DL).  

http://www.mpctoday.com/2016/04/19/what-has-happened-to-the-enthusiasm-for-computer-control-in-the-chemical-and-petroleum-industries

While looking for the reference to Charlie’s blog, I also found an article in my archives by Greg McMillan and Stan Weiner, which included Sigifredo Nino, titled “Invisibility of process control” (c. 2017) that echoes many of Charlie’s (and my) sentiments.

https://www.controlglobal.com/control/loop-control/article/11311827/invisibility-of-process-control

The fundamental issue, as I see it, is how process control is taught: Unlike heat or mass transfer, hydraulics and even thermodynamics, process control was (and, to my knowledge still is) based on esoteric mathematical theory that has little/no connection to the physical/real world. My process control professor at Texas A&M, Dr. Leo Durbin (whose nickname from previous students I later learned was “Mad Leo!”) had a standing offer for every class: He would give any student an “A” on a quiz if the student would put their fingers into an electrical socket and allow the rest of the class to watch them “oscillate at 60Hz!”  The students, of course, all thought that Dr. Durbin WAS absolutely crazy and no one, that I know of, ever took him up on the offer. It was several decades later that I realized that he was only trying to provide a little mirth and whimsy to an otherwise egregiously mind-numbing and impractical course that every chemical engineering student was required to take (and pass) and then never ever wanted to speak of again! 

Strangely enough, in spite of this situation, it was in the midst of taking that course that I had the epiphany that it was ONLY through the appropriate application of controls and, of course, the “Black Art” of PID Tuning, that one could get a process plant to perform as intended, and maybe even a little better than intended. Thus, as I approached graduation, I sought out roles and opportunities to pursue process control as, at least, part of my career. And now, 30+ years later, process control is my career! (Although I’m also now engaged in other aspects of the process industries, too, namely operational excellence and process safety.) 

Of course, what I was taught in the one required process control course failed to prepare me for the real world of an operating plant — having ONLY Ziegler-Nichols as the basis for PID tuning!  Thankfully, my first employer engaged Dr. Robert V. Bartman, who was formerly the head of Exxon’s Advanced Control Training, to provide a group of process control engineers from several of our plants his “Revelations” course — two weeks and 90 hours of class instruction! I was “drinking from a firehose,” but for me it “parted the veil” of process control: No longer was it a “Black Art,” but a practical science! The efficacy of his training (and the process identification and tuning software that our plant purchased) was proven out by the other control engineer at the plant where I worked: A production record was set from just ONE (1) PID application that used an online GC (with a 45 minutes, or longer, cycle time) that paid for EVERYTHING the company spent for ALL attendees on Dr. Bartman’s course and software.  I have been a strong advocate of Dr. Bartman’s approach to process control training, PID tuning, his course and software ever since.  Sadly, Dr. Bartman retired and his course and software are no longer available, but I do try to convey the essence and intended outcomes of his approach without violating any IP rights. 

Before attending Dr. Bartman’s course, I did take at least one AIChE short course on process control, but, paraphrasing Hitchhiker’s Guide to the Galaxy, it was… “mostly useless.”

Ever since I started speaking, presenting, as well as being published, in/about process control (starting roughly 10 years ago), I have always tried to highlight the fact that to be a good process control engineer, one must first be a good process engineer. In other words, one must understand HOW the process actually behaves, including hydraulics, thermodynamics, reaction kinetics, heat/mass transfer, etc. — first from a steady state perspective and then with the added components of dynamics. My pithy summary explanation is that there are two and ONLY TWO things that separates a process control engineer from a process engineer — lag time and dead time! 

Industry management (most of whom lack ANY process control, some not even process, background), however, seems to think/perceive otherwise: Process control is often perceived as a necessary evil — perhaps more evil than necessary! — and really exists only to ensure that the control system (hardware and software) is repaired and maintained and, perhaps occasionally implement a new control scheme (usually designed by someone else — often a process engineer with no appreciation for process control and dynamics). As a result, many chemical/process engineers, who would otherwise make excellent process control engineers, avoid pursuing such roles and opportunities as it amounts to career suicide! I experienced this very thing with my first employer, in spite of the fact that many management rising stars had process control backgrounds! And a quick look at many plant sites will confirm this assertion: Many people in “process control” positions are “long in the tooth,” have little/no (formal) engineering background (perhaps only some computer science or technology education) and have occupied that role for years, even decades! They are, however, very intimately knowledgeable of the control system, its capabilities and how to get the most from its limited resources. Thus, they are extremely difficult to replace once they (or nature) decide it’s time to leave. Case in point: I’ve seen a particular role advertised by several different recruiters looking for someone with intimate understanding of the VAX/VMS operating system! Digital Equipment Corp (DEC), the creator of VAX/VMS, hasn’t existed for decades and this site’s management expects to find an expert in that system today?! 

Hence, when speaking to people about “process control” roles they’re trying to fill, I probe to find out if it is, in fact, a process control role rather than a control systems role. For the process control engineer, the control system (DCS) is a tool; for the control systems engineer, it’s the job!

So, what does the future hold for process control? 

One thing is certain: If process control is to be perceived and treated as anything other than a “Black Art,” then the way process control is taught MUST change! So, those of you in such roles, PLEASE throw away the Laplace transforms and transfer functions (or, at a minimum, save them for graduate courses): We live, breath and exist in the time domain, not the frequency domain!  Students (and other practicing engineers) will find time-based concepts MUCH easier to comprehend and apply. 

Next, industry leadership must recognize (or be educated on the fact) that the foundation for ALL “advanced” control — of EVERY variety (i.e., DMC, RTO, AI/ML/DL, etc.) — relies and sits upon a foundation of well-designed and maintained regulatory control and instrumentation (in addition to many other layers below that — see below). If the foundation is weak or deficient, then no one should expect the “super structure” to remedy the deficiencies. Yet, that seems to be the perception, especially with the current fascination with “AI!” Which leads me to…

Finally, our organizational cultures must recognize the importance of (good) process control engineering. When I first started speaking on process control, I was sometimes accused of advocating that process control is the MOST important role in a plant. That was certainly not my intent and, while it’s not THE most important, I do believe that it’s definitely in the top 10, perhaps even top five, most important roles in an operating facility, accompanied by operational training and discipline, mechanical design, proper instrument selection and installation and maintenance of all the above. I also believe quite strongly, however, that process control is the MOST NEGLECTED of roles and professions that serve the process industries. Many plant/organizational cultures (as described above) illustrate this assertion quite well. 

Thus, my “Why” for what I do (cf. Simon Sinek, “Start with Why”) is to help affect an industry and a profession such that process control receives its due credit and attention. I hope that others will join me in this crusade. 😉

Russ Rhinehart’s Perspective

Adjust Engineering Education: The process control course is taught by professors who rarely have either any industrial or process control experience. They choose textbooks to represent their academic values (these include mathematical complexity that differentiates engineering from technology, isolated idealizations, exercises that are easy to grade by teaching assistants, authored by high-ranking academics and such). Successful textbook authors know this, so the course continues to be math-oriented and about Laplace and z-transforms to solve linear ODE systems and frequency analysis, all applied to trivialized situations. But students did not choose engineering to get a job in mathematical abstraction, and as a result of the control course, are more apt to reject a control engineering employment offer than accept it. The course needs to be reshaped to make control jobs attractive and to provide the breadth of understanding that would make students workforce-ready. The course should focus on the practical side of this essential topic and also provide a comprehensive view of the role of devices, calibration, communication networks, final elements, startup/shutdown, safety, batch, HMI, standards and such. Academe cannot do it. Faculty do not understand what is needed to prepare workforce ready graduates, and with their key performance metrics based on their research they have no incentive to do the work to change the course. To implement relevant change, the course requires substantial industrial practice involvement. If we want a better future for process control, industry needs to facilitate a substantial turnaround of the college course.

Nonlinear Model-Based Control: Our standard control concepts (SISO, PID, FOPDT models, Feedforward, RGA, etc.) are all based on linear (constant gain) and stationary (constant time-constants and delays) concepts. We inherited linear techniques from pneumatic and electronic analog devices in the pre-computer era.

But real processes are nonlinear and non-stationary. One of the problems of applying linear techniques to a nonlinear application is, that for good control, we often design processes to be compatible with the control math. We use vertical tanks, not horizontal, so that the level controllers see a constant gain process. We use large tanks and oversize distillation columns to dampen upsets. We also use gain scheduling so that linear controllers can adapt to nonlinear and non-stationary aspects. Although 2x2 interaction is not difficult with conventional ratio or decouplers, many processes have higher-order interactions, and although conceptually possible, even a 3x3 decoupler is impractical to implement. We limit the complexity of energy integration to satisfy what controllers can cope with. And, when the CVs or MVs hit constraints our standard practice is to add override structure, which also means anti-windup modifications. This works, but all these issues can be handled — and much more simply — by a controller grounded in first-principles process models.

A first-principles model-based controller has one tuning coefficient for each CV (not 2 or 3 as a PID requires) and has natural decoupling (not requiring several more coefficients to be adjusted for each decoupler) nor needing gain scheduling (of all the many tuning factors). It is much easier to tune. And being nonlinear, it does not need re-tuning when the process throughput, setpoints, tank levels, etc. change.

If controllers are grounded in first-principles models, the control engineer must learn the process phenomena (“to be a good control engineer, first understand your process”). To me, the control engineer should not be distracted by learning the mathematics and representation of linear ODEs, Laplace, z, FIR, NN or the diverse other models.

We use supervisory steady state phenomenological models for real-time optimization, but control is grounded in alternate models, such as FOPDT. We also use first-principles models for process analysis, and update such models to match the process to forecast constraints. The models in all these functions are different. They don’t exactly match, and they require different skills and procedures to develop them. I think it would be wonderful to have the “One Model to Rule Them All.”

I think that there are many other benefits of using first-principles process models in control and auxiliary automation functions, and that the computing power now in place in plants makes the transition possible. One of these benefits might be stopping the practice of operators placing controllers in manual because the process gains or dynamics changed making the controller dysfunctional.

If we want a better future for process control, we should move toward algorithms that use first-principles process models.

Computer-Enabled Techniques: The computational power associated with Industry 4.0 provides methods to take control and automation to the next level. This includes optimization of the process, scheduling, diagnosis, advising operators, safety, forecasting constrained conditions, digital twins, stochastic simulation, etc. The economic benefits of improved process management are much larger than the “control” objective of keeping CVs at a set point. Let’s move from control to a more synoptic process enhancement benefit that company managers will embrace.  

Michael Taube’s Follow-Up

LOVE the line in the middle of the first paragraph, Russ: “The course needs to be reshaped to make control jobs attractive and to provide the breadth of understanding that would make students workforce-ready.”  That should be BOLD, UNDERLINED and blinking! 😃

You also mention instructors lacking industry experience. While I was in New Zealand — lived there for about three years on a project for their ONLY refinery, now converted into an import terminal! — the department head of the University of Auckland’s chemical and materials engineering department (who is also one of the coauthors of my book, A Real-Time Approach to Distillation Process Control) opened the door for me to provide occasional guest lectures to both the process control and process design classes (there was only one session of each per semester) with the obvious emphasis on practical process control. When I repatriated back to America and found myself backfilling for an 11th hour cancellation at the Texas A&M Instrumentation & Automation Symposium, I approached the then chemical engineering department head with a similar proposal. While he was ostensibly supportive (after all, he gave me — a “nobody” — the “time of day” to meet with him in his office!), his response was that there were too many sessions to make my proposal practical. The “short version” — NOT INTERESTED. 

While I can only speak for myself, I’m sure that there are at least a few practicing process control engineers that would be willing to share their insights and expertise with ChemE Students and to “recruit” them into the profession.  All that’s needed is for ChemE departments to open the door!

Julie Smith’s Perspective

It’s easy to romanticize the past. Everyone loves to tell stories about the good old days when everything was awesome. Makes for fun bar conversation!

But the reality is the good old days had their struggles too. Technology was primitive and at times unreliable. Anyone else remember the strip chart pens running out of ink right when you began the heat up cycle for the much-awaited new product?

Today’s technology is much more computerized, with power increasing far beyond what we could have dreamed with the first DCS. Instruments and valves are getting smarter too, as are motors and other devices. We generate volumes of data about our processes and systems every day. With all this tech, how can we possibly fail?

The answer of course, is the same as it was 35 years ago. We need people to make any technology successful. The best tech will fail if not properly understood, designed, documented, operated and maintained. And as humans, we resist change. Just as a determined engineer could ensure the new product had its temperature recorded in ink back in the day, an equally determined engineer today can prevent the new system from making the right product. People need to buy in and believe in the benefits of the change.

The same is true for operations. Many advanced control schemes and other cool tools have fallen into disuse when the operations staff did not fully understand how to operate and maintain them. Or when we failed to change the associated work practices, the cultural shift needed for success is always far greater than the technology shift. Put the right team in place, and success will come.

So what tricks can we use to ensure this happens? Process modeling and simulation is one of the most powerful arrows in our quiver. A sound first principles model can enhance process understanding and form the basis for improved control design and operator training. The modeling platform has certainly changed over the years, but the benefits have not. More anecdotes can be found in ControlTalk and ProseraPod.

Greg McMillan’ Perspective

When I joined Monsanto in 1969, there was a 12-week course taught by company experts in instrumentation and process control. The course was shortened a week or more each year with the last offering in the late 1970s being two weeks. The corporate Engineering Technology section with an emphasis on steady state and dynamic modeling I worked in most of my career rapidly declined from 100 experts when the CEO was an engineer, to 5 and then 2 in the late 1980s. The chemical business was spun off into Solutia, Inc. A Process Control Improvement (PCI) group was formed with the ET remnants doing ground breaking Opportunity Assessments (OA) in many production units saving several hundred million dollars a year, but after eight years, we had addressed the biggest issues in most of the plants. Solutia sold many of its more productive plants to help pay for debt given in the spinoff. I retired in 2001 because some of my retirement fund would be gone.

ISA conferences had whole sessions devoted to process control. I chaired an Automation Week Conference in 2011 and had Charlie Cutler, Bela Liptak and Leo Staples give keynote talks and Greg Shinskey give a recorded interview. These days, there are next to no talks on process control at ISA conferences. The same thing happened more recently at the Texas A&M Instrumentation Symposium. In 2020, Russ Rhinehart and I gave keynote addresses for sessions filled with presentations on process control. In 2024, I was the only presentation on process control. We are fortunate that Mark Darby creates two sessions on process control at the AIChE Spring Meeting. I have noticed that magazines for the process industry rarely have had articles on process control since 2020. Fortunately, ISA supports the monthly posting of my “Ask the Automation Pros.” Control magazine still has significant content on process control by having Bela Liptak, Russ Rhinehart, Peter Morgan and I providing articles and monthly features such as Ask the Experts and Control Talk. Shinskey wrote many articles for Control magazine as well. Unfortunately, Greg Shinskey has passed and all his books are out of print, but you may be able to buy his last book Process Control Systems Fourth Edition by contacting the education department at Schneider Electric. Also, Control Global sponsors each year — since 2001 — inductees into the Process Automation Hall of Fame, and ISA has yearly inductees as Fellows and periodic Life Achievement Awards.

There seems to be an infatuation with the Industrial Internet of Things, and a resurgence in thinking that neural networks via machine learning and deep learning are the solution. While more information is helpful in possibly identifying unrecognized relationships, I think knowing cause and effects by understanding first principle relationships has much more value and prevents false conclusions from auto-correlations and cross-correlations. This is the route that Shinskey and I have taken, working in the time domain. I personally have greatly benefited by using first principle models. I think data analytics can help improve these models by identifying missing relationships. I think the incorporation of key performance indicators (KPIs) to show the increases in process efficiency and capacity in these models and then online can help management realize the benefits of process control improvement. Chapter 11 in my McGraw-Hill Process/Industrial Instruments and Control Systems Handbook, Sixth Edition has a lot of very practical information on how to make this happen by many “Ask the Automation Pros” experts.

On the bright side, measurements have gotten much smarter and more accurate and control valves more responsive. The PID controller that has been proven over decades including studies documented in “Linear Feedback vs. Time Optimal Control II - The Regulator Problem” by Alan H. Bohl and Thomas J. McAvoy published 1976 in Ind. Eng. Chem., Process Des. Dev., Vol. 15, No. 1, to be the best algorithm for unmeasured load disturbance rejection has many advanced features to address the challenges with override control, deadtime compensation, wireless update delays and analyzer cycle times as documented in the ISA-TR5.9-2023 technical report. Most notably, there is external-reset feedback and an enhanced PID that offer dramatic improvements with simple configuration additions. Also, the capability of a digital twin to explore, develop, prototype, test and demo process control improvements using the actual PID configuration offers incredible opportunities for expanding innovation and technical knowledge. Several universities have hands-on labs using digital twins and first principle models for key unit operations. Edin Rakovic does a great job of sharing knowledge, particularly showing how dynamic simulation is playing an increasingly critical role in sustaining and expanding innovation via his LinkedIn site.

There is an increased recognition started by Shinskey and by my efforts that the PID output must be driven past its final resting value for lag dominant, integrating and runaway processes and that PID controllers should be first tuned for load disturbance rejection and then a setpoint lead-lag or 2 degrees of freedom (2DOF) structure used to get the best setpoint response possibly complemented by a simple setpoint feedforward. The April Control Talk “Examining PID tuning essentials” with Michael Taube seeks to provide an up-to-date view on PID tuning including the recognition of mistakes and misconceptions and the use of better objectives, tests and procedures to meet application and safety issues. Also, there have been many advances in model predictive control, optimization and dynamic modeling. There are increasingly realistic opportunities for inferential measurements using first principle models to fill in the blanks and provide missing analytical measurements eliminating the deadtime from sample and cycle time. The inferential measurements provide fast results and are updated periodically by analyzer results. I have written a March Control “Fundamentals to Better Understand Process Dynamics” to promote the value of knowledge gained by first principle relationships. I hope to start working on an “Ask-Greg-AI” that can provide concise answers to questions and guidance to publications for deeper knowledge. I have started to make books free to download whose publication rights have been returned to me. Most notable is my book Tuning and Control Loop Performance Fourth Edition.

A university I was doing guest lectures for expressed no interest in the following ambitious curriculum for a Master’s Distance Degree in Process Control that I proposed four years ago even though I offered to find the industry experts who would be willing the teach the courses online.

1. Process Measurements and Final Control Elements (Sensors, Transmitters, Analyzers, Control Valves, Variable Frequency Drives, 5Rs: Resolution, Repeatability, Rangeability, Reliability, and Response Time). Books: Essentials of Modern Measurements and Final Elements in the Process Industry and Process/Industrial Instruments and Controls Handbook 6th Edition.
2. Batch and Continuous System Dynamics (Flow, Composition, Level, pH, Temperature). Books: Tuning and Control Loop Performance 4th Edition, Process Control Systems 4th Edition, and Process/Industrial Instruments and Controls Handbook 6th Edition.
3. PID Features, Feedforward, Loop Performance and Tuning (Adaptive Control, Anti-Reset Windup, External Reset Feedback, Form, Structure). Books: Tuning and Control Loop Performance 4th Edition, Process Control Systems 4th Edition, and Process/Industrial Instruments and Controls Handbook 6th Edition.
4. Regulatory and Advanced Regulatory Control Strategies and Key Performance Metrics (Online Process Efficiency and Capacity Metrics, Controlled and Manipulated Variable Pairing, Feedforward and Ratio Control, Open Loop Backup and Kicker, Adapted Setpoint Response, and Valve Position Control for Common Unit Operations). Books: Tuning and Control Loop Performance 4th Edition, Process Control Systems 4th Edition, and Process/Industrial Instruments and Controls Handbook 6th Edition.
5. Batch and Procedural Automation (Batch Sequencing and Continuous State Based Control).
6. Successful Automation Career Guidance (Best Practices, Concepts, Watch-Outs, Exceptions, Insights, Rules of Thumb, ISA Mentor Program). Books: Tuning and Control Loop Performance 4th Edition, Process Control Systems 4th Edition, and Process/Industrial Instruments and Controls Handbook 6th Edition.
7. Synergy between Modeling and Control (Digital Twin for Experimentation and Justification).
8. Operator Interface Graphics Design and Alarm Management (Philosophy, Identification, Rationalization, Detailed, Design, Implementation, Operation, Maintenance, Monitoring, Assessment, Management of Change). Books: Information-Rich Design: A Concept for Large-Screen Display Graphics, Alarm Management for Process Control, and Process/Industrial Instruments and Controls Handbook 6th Edition.
9. System Testing and Operator Training (Digital Twin for project execution).
10. Model Predictive Control and Model Based Nonlinear Control. Book: Process/Industrial Instruments and Controls Handbook 6th Edition.
11. Control Communications (Industrial Ethernet, Fieldbus, Actuator Sensor Interface, Wired HART, Wireless Technology, Wireless HART) Book: Process/Industrial Instruments and Controls Handbook” 6th Edition.
12. Data Analytics (Batch and Continuous Process Analysis and Prediction). Book: Process/Industrial Instruments and Controls Handbook 6th Edition.       (Elective).
13. Online Optimization (Linear Program, Real Time Optimization, Levenberg-Marquardt, Generalized Reduced Gradient, Hooke-Jeeves, Nelder-Mead, Leapfrogging) Book: Engineering Optimization: Applications, Methods, and Analysis. (Elective).
14. Safety Instrumented Systems (Analysis, Realization, and Operation: Reliability, Redundancy, Integrity Levels, Implementation, Testing, Maintenance). Book: Process/Industrial Instruments and Controls Handbook 6th Edition. (Elective).
15. Automation Project Execution and Management. Books: Successful Instrumentation and Control Systems Design 2nd Edition and Process/Industrial Instruments and Controls Handbook 6th Edition. (Elective).
16. Bioprocess Modeling and Control (Improving Fermenter and Bioreactor Performance). Book: New Directions in Bioprocess Modeling and Control 2nd Edition. (Elective).
17. pH Modeling and Control (Improving Process Performance and Environmental Protection). Book: Advanced pH Measurement and Control 4th Edition. (Elective).
18. Distillation Modeling and Control (Pairing of Variables, Decoupling, and Optimization). Books: Distillation Control - For Productivity and Energy Conservation 2nd Edition and Process/Industrial Instruments and Controls Handbook 6th Edition. (Elective).
19. Chemical Reactor Modeling and Control (Pairing of Variables, Inferential Measurements, Optimization). Book: Advances in Reactor Measurement and Control. (Elective).
20. Mining and Metals Modeling and Control (Pairing of Variables, Inferential Measurements, Optimization). Book: Mineral Processing Technology:   An Introduction to the Practical Aspects of Ore Treatment and Mineral Recovery 8th Edition and Process/Industrial Instruments and Controls Handbook 6th Edition. (Elective).
21. Compressor Modeling and Control (Surge Prevention and Recovery, Multi-Stage Decoupling, Optimization, Capacity Control) Books: Centrifugal and Axial Compressor Control and Process/Industrial Instruments and Controls Handbook 6th Edition. (Elective).
22. Food and Beverage Modeling and Control (FDA Regulations, Fermenters, Blenders, Extruders, Inferential Measurements, Drying). Books: New Directions in Bioprocess Modeling and Control 2nd Edition and Process/Industrial Instruments and Controls Handbook 6th Edition. (Elective).
23. Crystallizer, Evaporator and Dryer Modeling and Control (Particle Size Control, Solids Concentration Control, Moisture Control, Inferential Measurements). Book: Process/Industrial Instruments and Controls Handbook 6th Edition. (Elective).
24. Steam System Modeling and Control (Boiler Control, Header Control, Turbine Control). Books: The Control of Boilers 2nd Edition and Process/Industrial Instruments and Controls Handbook 6th Edition. (Elective).

ISA Technical Reports and Standards and A Guide to the Automation Body of Knowledge, Third Edition can be used to provide essential knowledge and guidance in all courses.

Potential instructors are contributors to Process/Industrial Instruments and Controls Handbook, Sixth Edition, key members of the ISA TR5.9-2023 Technical Report “PID Algorithms and Performance” and participants in Control Talk Columns and “Ask the Automation Pros.”

Brian Hrankowsky’s Perspective

Every time I started to write something it up, it seemed to go on and on. There’s so much to say on this topic. In reading the other responses, I tried to focus on where my observations might be a little different: I have recently become aware of some differences in today’s curriculum from what I had that alarm me. The courses or material that I rely on most seem to have disappeared or are watered down.

My background is likely a little unique. My B.S.E. and M.S.E. degrees are in systems and control engineering. I took A LOT of modeling, signal analysis and controls courses. The degrees are closest to electrical engineering, but had a good chunk of mechanical engineering and industrial engineering content — at least based on the electives I chose. I agree that there was a lot of time spent on linear time invariant systems and it was math-heavy — but I like math, and that’s part of what I liked about the major. I was very comfortable with Laplace, Fourier, discrete Fourier, windowed Fourier and Z transforms. Took a course on wavelets — very cool, but not sure I’d claim mastery on that one. I was a pro with MATLAB and Simulink. I find it very useful to be able to see physical systems behaviors in terms of what type of model fits them and use ideas from the tools that are designed for those models. My professors worked with industry — senior projects were often like mini-internships or co-ops as the problem would be an actual industry problem looking for a solution. The co-op program seems to be different that others too. They were six months, and we did actual work. I had one at North American Manufacturing Company, a combustion engineering company, where I continued R&D work started by other co-ops and engineers to eliminate resonance from their low emissions burner solution. I built on previous testing: I interviewed folks who’d worked on the problem before, read up on the existing reports and data, designed additional tests to fill out the data sets and prove out a hypothesis, wrote a mathematical model a genetic search algorithm in C to identify potential mechanical changes to reduce noise and had one of my designs implemented at a customer’s site. My second co-op was with the Foxboro Company. I wrote sequence code to control a chemical feed system deployed to a customer site in Mexico. My co-ops are where I got hands-on experience with industry tools. I’m not sure what I see of other school’s co-op programs gives them the same depth of experience.

I took me a long time to figure out why so much of what I learned in class didn’t seem to fit with the systems I was working with in biotech plants. As others have pointed out, process plants have significant deadtime and are non-linear. I will throw another item in the mix: The textbooks I have use the parallel form of PID. Seeing the transfer functions of the series and ideal forms and recognizing the relationship between the controller gain and process gain and between the integral and derivative and the system capacitance and deadtime unlocked the connection between the tuning parameters and the Bode plots. This made Fourier and Bode useful — not as an everyday tool, but as a basis for understanding the relationship between system dynamics, system stability and the tuning parameters. It is how I understand derivative as a tool that makes systems more stable. Continuous time root locus is useless for any system that has deadtime, BUT for me it is the perfect way to understand why tuning integrating processes can feel counterintuitive and gives me the confidence to crank up the gain to “get over the hump.” I’m not likely to do a lot with state space, but it was how I first thought of the decoupling strategies I read about for interacting loops. It's not perfect nor necessary to make that connection — but I did.

Where am I going with this:

The math matters. The theory matters. Could we drop Nyquist and Routh-Hurwitz? Probably. Should we drop state space? Maybe for ChE but not for all disciplines. I’d worry we did the wrong thing if we dropped root locus and Bode plots and the things that build up to them. These tools are how I understand instability and system dynamics. Are we missing tools like cascade, feedforward and external reset in the intro to controls curriculum? YES!

Here’s what I see that needs improvement in academia. Keep in mind this is from my perspective talking to new hires who are recent graduates and that I have taught our internal control theory refresher for 15+ years. Student backgrounds have noticeably shifted:

1. Make the controls course required for all engineers. ABET accreditation seems to imply that it would be, but I know several recent graduates who did not have a controls course at all. They had never heard of a PID controller. One is from an Ivy. My colleagues and I can’t bridge this gap in our spare time with every new hire we get. There are definitely schools with great programs. It is strange to have new hires with the same degree but such differences in what they learned because of the school.
2. For ChE, add teaching common control strategies as part of the unit op course work. It seems like the natural place to introduce external reset feedback, P and or D on PV (2DOF!), feedforward, override, cascade, coordinated tuning and other common control techniques. Its seems more natural to start identifying inverse, runaway and integrating responses here too. There is a book or appendix for pretty much any process out there. On the third/last day of our inhouse class we have a group design problem. I have yet to have a group that I could not get to design and understand three element boiler level control. I do not understand why they don’t learn it in school.
3. Have students take the controls class right after differential equations. Too many students are taking it their last semester when they already have a job lined up and its been nearly two years since they did any Laplace. It is a struggle and they have senioritis — they are just trying to survive until graduation at that point.
4. Include a logic course. I took digital logic and still use Karnaugh maps to analyze complex logic expressions (you wouldn’t believe some of the interlock expressions …). It enables me to think in something other than if then else statements and extract truths where they weren’t obvious. I’m not sure if everyone needs digital logic or would benefit as much from the logic class in the philosophy department, but I see too many engineers with no tools to break down and describe complex behaviors in control logic.
5. Teaching Aspen and MATLAB is not providing transferable skills into automation. Yep, those are the languages that one recent grad learned in their programming class. The learning curve for them has been very steep to learn how to specify control system functionality and configure control or HMI applications. Python doesn’t seem to help much more. Its abstracted too far from the hardware. There is marked difference in uptake and onboarding rates with engineers who learned C, C++, Fortran or Pascal. I’d expect Rust or Java to be similar. I don’t have to teach them about memory, binary, floating point error, precision and a host of other things. I diagnosed several site issues in 2023 as floating point error related — they’d never heard of it. The programming course should prepare them to be able to use any tool they need to. AND change the programming classes to include how to plan/design a program. I was lucky enough to take a programming class for a few weeks in 7th grade. Before writing any Basic code (Apple IIe), we learned how to create a flow chart and how to write pseudo code. I rely on these skill/tools today. They have translated beyond just helping me code stuff into my ability to write good specifications.

It is hard to think about where to go next when it feels like we’ve backslid and need to get back to where we were.

Pat Dixon’s Perspective

History has been well-covered by my very esteemed colleagues, so my contribution addresses providing for a better future.

For the last four years, I have been part of an effort at Miami University (the original in Oxford, OH) to prepare students for process control internships. This could be a model for attracting more talent to industry. We have long known about the gray tsunami, where process control engineers retire without the talent pool to replace them. Exposing more students to opportunities and providing them practical support may help fill the talent pool.

At the same time, I am concerned that many facilities have become too lean. Industry increasingly relies upon instrumentation and automation, yet onsite electrical and instrumentation (E&I) and process control staffing can be very thin. This results in people being overworked and leaving while the quality of data depreciates. The firms that can better evaluate the onsite staffing investment, by considering the financial accounting impact with the efficiency and safety impact, will outperform their competitors.

Tools such as computerized maintenance management systems (CMMS), asset management systems (AMS) and loop diagnostics can greatly help. It is practically impossible to manually maintain and monitor every PID loop in a facility. Today we have applications that can identify problem loops and instruments and recommend tuning adjustments. CMMS and AMS can reduce the workload for E&I so they can prioritize work more efficiently. Adoption rate of these tools is rather low, but those firms that smartly implement the tools can reap large rewards.

There used to be dramatic differences between supervisory control and data acquisition (SCADA) and distributed control systems (DCS). Those differences are more subtle today. This can help ease the learning curve for process control engineers. Most modern SCADA and DCS have some standardization in programming, configuration and approach. In contrast, I may be one of the few people around that is proficient on Honeywell TDC 3000 because it was built in-house by Honeywell as a proprietary system. My system integration business today handles a variety of platforms where the talent base can more easily transition between systems. I can see a future of more standardization.

Regarding control technology, I share Greg’s sentiment that PID is well-proven and frankly is ideal for most single loop control. It does lack deadtime compensation, but there are well-established means for addressing this. It might be sexy to suggest PID is old and artificial intelligence (AI) is better, but those are people that don’t understand Michael’s comment about deadtime and lag. Our processes are not steady state; they are dynamic. Russ brought up the challenge of non-linearity, which makes the math really hairy. However, it needs to be recognized that nonlinear gains have constraints. It is unlikely to encounter process gains that cannot be modeled reasonably well with a third order polynomial. Overfitting a complex model results in gains that violate first principles and trains for noise instead of the actual process. As mentioned, online loop monitoring can go a long way toward maintaining loops that recognize first principles.

I will note that depending on your definition of AI, a pneumatic PID loop was AI.

In advanced control, we all know that MPC has had maintenance challenges. The 18-month model mismatch window cause some to abandon MPC. Today there are vendors including monitoring in their MPC package in the same way as loop monitoring is offered for PID loops. It is important that any real-time adaptation not learn noise. The pre-processing work done in model identification is most of the work, and if all data is assumed to be valid, you can end up with very bad results. I can see a renewal of MPC adoption with proper monitoring.

Lastly, the manufacturing execution system (MES) is the “final frontier.” In this fourth industrial era that is full of fluffy marketing buzzwords, some portray industry as if it has finally become digital. We have known that our facilities and corporate finance systems have been digital for a very long time. It is the bridge between operation technology in facilities and information technology at the corporate level where all the “digitalization” is happening today. This is the MES layer. The obvious challenge that has been introduced is cybersecurity. My system integration firm is focused on this, as are several others. A future of connectivity with security can yield real-time enterprise optimization.

In conclusion, I am optimistic about the future of automation but concerned that technologists without process understanding looking for sexy shortcuts and accountants without recognition of the actual return on investment stand in the way.

Mark Darby’s Perspective

The value of good regulatory control has been and continues to be underappreciated. In my career, it has been the newer technologies that generate the most excitement. Examples to name a few are:

• Model predictive control
• Real-time optimization
• Expert systems
• Neural networks (then and now)

Michael Taube correctly notes Charlie Cutler’s lament about the lack of enthusiasm for computer-based control, including primarily model predictive control (MPC), but which also included real-time optimization (RTO). Both were popular conference topics in the 1990s. Some conference topics reflect what is new or promising. I think it is noteworthy that regulatory (or “basic”) control even then then did not get the attention it should have. There continues to be significant potential for improving the regulatory controls, for example, through tuning, reconfiguration and adding functionality. Not due to Cutler, but many took away the mistaken notion that MPC would compensate for weaknesses in the regulatory control strategy and, as a result, insufficient attention was paid to the regulatory controls. It is notable that many papers and presentations on the application of MPC projects have stressed that significant benefits often accrue to improvements made to the regulatory controls. A key point: It shouldn’t take an advanced project to make this happen.

Safety as a general topic and focus areas has increased significantly since the 1990s. The synergy with process control, including safety instrumented systems (SIS), is an obvious one for highlighting the importance of control, including improvements to better keep away from shut-down points.

Real-time optimization is an example of where what was once new and exciting may not endure. RTO has not lived up to its promise, with only a few companies willing to (continue to) invest in its application and the people needed to support it.

A few thoughts on the undergraduate process control course. I believe there have always been criticisms of the process control course. I believe complaints have gone up because faculty today generally have less industrial experience than before (but I don’t think it was ever great). I am still surprised to hear how the process control course has soured some on the prospect of actually working in control. To my way of thinking, students are aware that actual practice is different on the job, and further, they expect their employer will emphasize what is important for the job. I also put blame on industry for not valuing process control more. Companies will obviously have a big impact on choices /directions engineers will take with their career choices; I conclude that many companies are deficient on this.

In my interaction with universities ­— as a part-time lecturer and while serving on an advisory board — I have found that there are challenges in making changes to the core curriculum. One is that the reduced number of hours required to get a B.S. degree have decreased compared to say 25+ years ago, leading some universities, in the worst case, to drop the control course. In many schools, process modeling is also covered in the process control course, which necessarily squeezes out other control topics. In my view, the dynamic balances covered in the modeling is quite important; otherwise, there may not be the opportunity to be exposed to first principles dynamic modeling. Other topics also compete for coverage, e.g., MPC and optimization.

I had a discussion recently with a process engineer, who is about 12 years out of university. He said he did not mind the process control course, but he was missing the connection between what they were learning and how it fit into the bigger picture of process control and plant operation. I expect this view is widespread.

My recommendations:

• Provide the big picture view of process control and related technologies.
• Spend more time spent on PID tuning, which is one of the few gripes I hear from industry.
• Cover typical control loops.
• Hands-on labs. An option is the Arduino temperature lap, an actual electrical heater process that attaches to a laptop. A YouTube video: https://www.youtube.com/watch?v=zMBeL4HpFVY. Control of a realistic first principles model is another option.

Here are suggestions on providing a better future for process control, which would also improve the bottom line and deliver safer and more reliable operation.

1. Continue beating the drum on the importance of process control. Posts like this play a role.
2. If not done already, companies should make it easier to rotate in/out of the control group to expose more engineers. This would also help address the perception that control is a job that you can never leave.
3. Companies should assess their staff levels for process control — both number of staff and expertise. Alternatives to internal staff include utilization of part-time staff and outside consultants.
4. Companies should reach out more to universities. Possibilities include providing input into the control curriculum, assisting in a lecture or speaking to university organizations on topics of process control and the role of a control engineer. A benefit to companies is finding students interested in working in process control.
5. Use of control lecturers with industrial experience. The increased use of non-tenure track professors of practice is a positive step in this direction.

Ed Farmer’s Perspective

Here are my thoughts from life in the jungle. I used an old version of this dialog decades ago to help steer focus beyond the textbook example into the future that was still hiding over the horizon. I have also worked with non-engineering people on perception and vision in the cosmos and the ever-more-complex modern physics concerns. Here it goes:

In the piece of the cosmos we can know or imagine, Nature has spent at least the last 13.8 billion years finding its way from “nothing” to all the things that surround us. In our tiny location over its most recent history, the Nature we have come to know has arrived, adjusted and continued its evolution. We continually learn about life in this Nature and derive ways to optimize pieces of it for our diverse purposes. Our scope is limited and bound by “Nature’s rules.”

Back at the end of the 1960s, I started on my master’s degree in “modern physics.” My professor was an alumnus of the Manhattan Project. Among other things, he was artistic in guiding students through what we knew and didn’t know about “science.” Those of us who had worked our way through various forces, materials, motions, energies and compositions felt like we knew a lot, but as our studies evolved, it became increasingly apparent that the essence of nature involved a few key concepts — like “conservation,” applied to mass, momentum and energy. It involved just four forces. Schrodinger (and others) discovered “smaller” and others expanded the cosmos we knew beyond our perception. Einstein connected “energy” and “mass.” Our “big picture” simplicity gave way to spawning even more powerful realities. Evolution and discovery force “change,” good and bad. Sixty-five million years ago, life on earth involved dinosaurs, but suddenly a space-rock encountered earth and they were gone.

A half-dozen centuries ago, humans’ civilizations began adapting to the nature they knew. As their capabilities developed, they discovered ways to adapt some features of nature to “better” their experience. They also discovered “different” doesn’t always turn out to be “better.” Progress involves “change,” which involves learning, which motivates still more learning, more change, more desire, more hope.

In the ‘60s, automation concepts began changing process control. “What’s possible” has demonstrated tremendous progress which continually involves “change.” I can remember discussions about how pneumatic control was so much better than finding and adjusting the proper valve, and then how electronic controllers broadened capabilities, and then how integrated control systems involved massive changes in thinking, as well as the beginning of broadening capability.

I remember an effort to improve control of distillation columns. A change in feed temperature required considerable time to propagate through all the levels, producing a lot of off-specification output that would require recycling. The same kind of problems would occur when the tanker ship connecting to unload would be carrying significantly different crude oil than the one that just finished. Equipment automating adjustment to such situations improved operation and reduced loss. Designing these systems was a lot different than pneumatic control, and getting the available benefits involved thinking in terms of the capabilities of this equipment.

As one drives through the gate, past the office buildings and up a small hill, the acres of plant become visible. Across all the process piping, towers, tanks, reactors, heat exchangers, furnaces, control rooms and people, one can see a huge oil tanker ship at a dock unloading crude oil for processing, producing, packaging and delivering products to a diverse suite of users.

Plant-wide utilities service pretty much everything and can precipitate widespread difficulties. It’s nice to know enough about those “common elements” to evaluate how concerned to be. Questions like, “Is any process having utility issues?” can help develop feelings and understanding. Vague responses can motivate concern that can take time to investigate and resolve. When there’s no response to such a question, it can be useful to remember your college course in “operations management.”

Some processes, when unintentionally disturbed, work backwards from the usual expectations. In many processes, turning off steam or electric supply simply shuts things down, but suppose there is a chemical unit that becomes increasingly exothermic when the flow isn’t moving fast enough. That was an exciting day! Then there was a place, back in the pneumatic days, where process equipment air supply ran near an old fired heater that started a fire that burned through the pneumatic tubes that controlled automatic sprinkler equipment. It’s often important to learn and know about situations that produce such events.

In engineering college, we learned about process dynamics using mathematics from a class (or two) in “multi-variable partial differential equations.” Process controllers provide programmable equations like that to interact with and control the dynamics inherent in the process. Adjusting the controller’s parameters to interact in a way that produces the desired changes in the process is all about good design and appropriate tuning. The guy who tunes the controller probably is not the same guy who does valve maintenance. Suppose a valve gets changed and the new one looks the same but has grossly different trim. That builds in a change in loop gain that result in needing to change the controller’s parameter’s “gain” setting. Sometimes controlled output goes to a valve manifold that effectively changes loop gain expected by the controller by opening and closing valves. Getting better control requires thinking through what is supposed to happen with those manifold valves and what the control loop really needs to, or can, do.

Dead time is the interval between when a change occurs and when it becomes visible in a pertinent measured variable. Conventional single-loop controllers only know when they “saw” the change, which might be much different than when dead-time-corrupted data arrives. The inherent assumption is that tuning is based on “real time,” but in this case, that isn’t so. A solution will involve more than conventional tuning.

Sometimes, a perfectly good feature is added in a control loop (or system) that negatively impacts others. Perhaps this can be solved with tuning, or maybe reconfiguration. Maybe the ripple-out from the addition requires rethinking how several related “other things” work and the situations they create.

Somewhere in the ‘90s, the available control equipment included capability to consider more process characteristics and situations in economical equipment. That helped, and capabilities continue to develop. It was often attractive to evaluate the value of new capabilities, preferably with some thought about where the future of a loop, process, plant or company might be going.

Don’t get lost in the jungle. Conditions will be different from “here” over “there.” As time runs along, what is important evolves, and “what’s expensive” must be freshly evaluated. Look (carefully!) for opportunities, and look insightfully back for things that could motivate them. Understand what your plant is about in its marketplace, what sort of “things” are coming and the capabilities that grow with “new” or “better.”

About the Authors

Michael Taube is a principal consultant at S&D Consulting, Inc. Serving the greater process industries as an independent consultant since 2002, he pursues his passion to make things better than they were yesterday by identifying the problems no one else sees or is willing to admit to and willingly “gets his hands dirty” to solve the problems no one else can. Due to the continued occurrence of individual injuries and fatalities as well as large-scale industrial incidents, he collaborates with operational excellence and safety culture experts to promote a real and lasting cultural shift in the process industries to help make zero incidents a reality. He graduated from Texas A&M University in 1988 with a bachelor of science degree in chemical engineering. https://www.linkedin.com/in/michaeltaube/

Russell (Russ) Rhinehart has experience in both the process industry (13 years) and academe (31 years). He is a fellow of ISA and AIChE and a CONTROL Automation Hall of Fame inductee. He served as president of the American Automatic Control Council and editor-in-chief of ISA Transactions. Now “retired,” Russ is working to disseminate engineering techniques with his web site (www.r3eda.com), short courses, books and monthly articles. His 1968 B.S. in ChE and M.S. in NucE are both from the U. of Maryland. His 1985 Ph.D. in ChE is from North Carolina State U.

Julie F. Smith is the global automation and process control leader for DuPont. She has 35 years of experience in the process industry, having been part of numerous engineering and operations activities across the globe. She has written several papers and columns highlighting the value of modeling and simulation. Julie has a BS in chemical engineering from Rensselaer Polytechnic Institute and an MChE from the University of Delaware.

Gregory K. McMillan retired as a Senior Fellow from Solutia Inc in 2002 and retired as a senior principal software engineer in Emerson Process Systems and Solutions simulation R&D in 2023. Greg is an ISA Fellow and the author of more than 200 articles and papers, 100 Q&A posts, 80 blogs, 200 columns and 20 books. He was one of the first inductees into the Control Global Process Automation Hall of Fame in 2001, and received the ISA Lifetime Achievement Award in 2010, ISA Mentor Award in 2020 and ISA Standards Achievement Award in 2023. His LinkedIn profile is: www.linkedin.com/in/greg-mcmillan-5b256514/

Brian Hrankowsky is a senior advisor of engineering with 23+ years process control experience in the pharmaceutical and animal health industries. Brian has experience in large and small molecule synthesis and purification, continuous utility, discrete assembly and packaging automation with various DCS, PLC, vision and single loop control platforms.

Pat Dixon, PE, PMP is president of www.DPAS-INC.com, a system integrator with expertise in data analytics and advanced control.  Pat has experience in industrial automation beginning in 1984, having worked for SD Warren Paper, Honeywell, Pavilion Technologies, DuPont and Emerson as well as system integrators.  Pat is a professional engineer in four states and a certified project manager.  His LinkedIn profile is: www.linkedin.com/in/dixonpatrick/

Mark Darby is an independent consultant with CMiD Solutions. He provides process control-related services to the petrochemical, refining and mid/upstream industries in the design and implementation of advanced regulatory and multivariable predictive controls. Mark is an ISA Senior Member. He served on the TR5.9 committee that produced the PID technical report and has presented at ISA technical conferences. Mark frequently publishes and presents on topics related to process control and real-time optimization. He is a contributing author to the McGraw-Hill Process/Industrial Instruments and Controls Handbook, Sixth Edition. His LinkedIn profile is: www.linkedin.com/in/mark-darby-5210921

Ed Farmer completed a BSEE and a physics master's degree at California State University - Chico. He retired in 2018 after 50 years of electrical and control systems engineering. Much of his work involved oil industry automation projects around the world, and application of the LeakNet pipeline leak detection and location system he patented. His publications include three ISA books, short courses, numerous periodical articles and blogs. He is an ISA Fellow and Mentor.