Greg: The future of the automation profession depends upon replenishing and revitalizing the expertise in process control. Nearly all of my protégés have retired without an opportunity to mentor and develop their sequels. There was a dearth of hiring automation professionals for 30 years. While there has been a recent increase, many if not most of these engineers new to the profession are not going to work for the plants or associated engineering departments of the process industry manufacturers. They are instead being hired by service providers, either in the application departments of system suppliers or their business partners, or by independent engineering firms. The challenge is how these new engineers go from what they learned in universities to what they have to do on the job. Since they are not starting out gaining extensive plant experience by being stationed at a plant and/or spending significant time checking out, commissioning and starting up the systems throughout their career like Stan and I did, they are challenged to learn what is really going on in the field. Further complicating matters is that university degrees in process automation are rare and the courses offered in control are oriented toward frequency response, emphasizing Laplace and Z transforms and Bode and Nyquist plots. While this knowledge is useful for advanced degrees and a deeper understanding for someone in research and development, the approach has little practical use for the automation engineer in the process industry. The courses in control theory largely reflect an original intent to educate potential aerospace engineers in the 1970s.
The Missouri University of Science and Technology (Missouri S&T) is making a difference principally as a result of the efforts of Dr. Kelvin Erickson. Kelvin worked at Fisher Controls in Marshalltown, Iowa, which was at the time owned by Monsanto. Among other projects, Kelvin helped develop the programmable logic controller (PLC) interface and a large database console for a distributed control system (DCS).
]Kelvin came to Missouri S&T in 1986 where he developed courses and labs that emphasize practical applications in automation. One of the courses is based on the book Plantwide Process Control, which he coauthored with a leading industry expert John Hedrick and focuses on a complete solution to maximize the scope and success of automation in both batch and continuous processes. Kelvin also teaches full and abbreviated lab-based courses onsite and online on using PLCs for automation in and beyond the process industry.
Stan: Kelvin, what is your approach?
Kelvin: Since 80% of the graduates end up providing services to the process industry (e.g., chemical, beverage, and wastewater) and to some extent in packaging, filling, assembly and small parts manufacturing, we developed labs to provide a practical and hands-on education of implementing automation in these industries. The focus is more on practice than the control theory that they can get with traditional courses in process control, which originated based on aerospace. Since PLCs are often used in these manufacturing processes, the labs extensively use PLCs.
Greg: I think this approach is critical because most will not have the advantage that Stan and I had of spending a lot a time onsite in labs and manufacturing plants as we progressed from university graduates full of theory to being specialists and eventually Fellows in process control. Also, the process industry is fraught with nonlinearities and deadtime not typically seen in aerospace.
Kelvin: The feedback we get from graduates and the companies that hire them is that they greatly appreciate the hands-on practical approach providing so many benefits including understanding the terminology and implementation details and seeing the reality of automation in action.
Stan: What do you cover in the labs?
Kelvin: The labs deal more with the operational aspects—controlling the loop mode, setting up the loops and getting the automation system to work. There are a lot of parameters you need to get right. A third of the students do this lab before I cover PID control in the lecture, so I don’t focus on loop tuning.
We have labs on ventilation system pressure/temperature control, pH control, and heat exchanger temperature control. These are all continuous operations. I use only one of these labs in a particular semester. Other lab exercises are discrete sequential in nature.
To support the PID control labs, I spend a few weeks on PID tuning in the second half of the course using a simulator so we can look at more conditions and types of loops. Because it is referenced extensively, I start out with Ziegler Nichols (ZN) tuning but later show how tuning rules developed by H.A. Fertik and described in his paper, “Tuning Controllers for Noisy Processes,” in ISA Transactions Vol 14, No 4, provides more robustness and a compromise between the load and setpoint response.
Greg: ZN tuning is widely recognized as being too aggressive. We have learned to provide robustness with other tuning methods and to minimize setpoint overshoot courtesy of Greg Shinskey, using good load rejection tuning by the simple addition of a setpoint lead/lag with the lag time equal to the reset time and a lead equal to ¼ the lag time. Alternatively, we have learned to set the beta and gamma factors equal to 0.5 and 0.25, respectively, in a two-degrees-of-freedom (2DOF) PID structure.
Stan: How do students get to see what is happening in the process and complete the lab within the limited time available?
Kelvin: The response of the ventilation system is a few seconds and the pH system is about two minutes. The temperature response of the heat exchanger is much slower.
The ventilation system consists of two loops: pressure control by manipulating the variable-frequency drive (VFD) speed and temperature control by manipulating a heater coil through time-proportional control.
As a demonstration during the lecture, I tune the ventilation system pressure control by using a human-machine interface (HMI) to view the real-time response plots. The response is not smooth. There is a lot of fast oscillation and noise, not like what you see in textbooks. You have to go by your wits at times. Things can happen so fast, you have be extremely attentive and be able to look back in time on the trend charts with fast time scales to see what happened. Sampling at 0.1 seconds with remote access into the lab is barely adequate. I demonstrate moderated Ziegler Nichols (ZN) tuning, that is, after finding the PID gain for oscillation, you back off by cutting the gain in half and estimate the oscillation period for the reset time. I also demonstrate Fertik tuning, but estimating the process time constant and deadtime is a challenge since the trend plot does not stop.
In the pH control lab, they get to learn about level control and split-range control. Letting the controller output determine which reagent to use rather than preselecting an acid or base based on the pH is a preconception hurdle. The acid is dilute vinegar and the base is baking soda and water. The feed is spiked with reagents to create disturbances. The clear tank and clear plastic piping allow them to see the flows, which is not the case with the heat exchanger.
The heat exchanger system is a non-cascade control system where the exchanger outlet temperature is controlled by the position of regulatory valve in the hot water flow. Disturbances are made in cool water flow. Since you cannot see what is happening in the piping or exchanger, you have to rely on the measurements, which is instructive in that it shows that the measurement system is often your only window into the process.
Getting the lab completed in the allotted four lab periods is often a challenge at the beginning of the semester. By the end of the semester, they are able to work faster and most have no trouble completing the lab exercise.
Greg: It is a great idea for them to learn split range since many control problems tend to result from the nonlinearity and discontinuity of the split range point often causing oscillations back and forth across the point during low demand. I like the idea of using a weak acid and weak base, not only from a safety viewpoint, but also in terms of the titration curve slope being moderated and not posing excessive difficulty in control.
Stan: What is important for effective remote access to labs?
Kelvin: We have webcams and one or more teaching assistants (TA) in the lab when the lab is used either onsite or online. When used remotely, a TA can handle up to thtree groups at the same time. The partnering of one remote student with one present in the lab seems most effective. On-campus groups sometimes have a third member due to an odd number of students. All lab participants get to view and participate in Missouri S&T lectures. The remote access is also used for those students in Springfield, Missouri, enrolled in the EE program that is a partnership with Missouri State University.
Greg: Most of the employees at Mynah Technologies, where in my golden years I am a part-time consultant in modeling and control, are graduates of Missouri S&T. I am impressed at how quickly they become proficient in how process control systems are implemented and the critical role of process knowledge. I think dynamic simulation plays a big role in this learning experience, as it did with my career. My goal as I near the finish line is to show the synergy between modeling and control, enabling through virtual experimentation the ability to discover knowledge about the process and the automation system. This depends upon virtual plants being set up and made convenient and easy to use, and the recognition by management of the value of the time invested for development of their automation engineers. The Chemical Engineering Department at Missouri S&T has a state-of-the-art DeltaV virtual plant with dynamic Mimic models courtesy of Mynah. The hands-on access to hardware labs makes the solutions real, and the virtual plant broadens the horizons and flexibility of the learning experience.
The top 10 things you don’t want to hear in a process control lab
(10) I got all the loop dynamics from a single step test.
(9) The deadtime is zero.
(8) I simplified the response to be just a time constant.
(7) I have the constant for the time constant.
(6) I don’t see the need for control, just set the flows per the Process Flow Diagram.
(5) With feedforward, I don’t need feedback control.
(4) I don’t see anything to be gained from gain action.
(3) Why do we need pH electrodes when we have litmus paper?
(2) The best valves have least leakage and give you all the flow right upon opening.