Control Engineering for Chemical Engineers
The March 2017 Chemical Engineering issue features the article “Control Engineering for Chemical Engineers” [pp. 42–50]. The author has brought up an interesting topic that should concern everybody working at any chemical processing facility. I certainly agree with Mr. Heavner, the more the process engineers know about process control the better, as it will lead to better understanding of the process possibilities and limitations.
There are, however, some concepts presented in the article that need to be clarified:
“ Ziegler-Nichols tuning rules….But this kind of aggressive tuning results in some cycling”— Even if Ziegler-Nichols tuning rules are not the best, it must be understood that those rules were proposed to provide PID controlling lagtime-dominated processes with good capability to reject unmeasured disturbances, and that a change in the setpoint will invariably produce PV overshoot, this has been repeatedly misinterpreted as aggressiveness.
“… so most loops today should be tuned for a first-order response,…” In the chemical process industries the most infrequent change is to the setpoint, and the most important feature a controller must be provided with is load rejection capability. However, the author is suggesting that controllers’ response should follow a first-order response; this is the vice versa situation of the previous comment, a controller tuned for first-order response to setpoint change will invariably lack the required capacity to promptly return the PV to SP following the ubiquitous disturbances the chemical process are constantly subjected to.
“… controllers can be tuned on integrating processes to achieve a first-order response….Following a setpoint change, the PV will move to the new setpoint and overshoot slightly before turning around and settling back to setpoint…” What the author describes here is the use of a PI controller in an integrating process, which in general is a recipe for an oscillator. An integrating process should be controlled by a proportional only controller.
“ One guideline that is wisely favored is the ‘lambda’ tuning method.” Lambda tuning and aggressiveness don’t come together. It is a method that detunes the PID controller in favour of robustness — not a good compromise when lagtime-dominated process variables need to stay close to the setpoint following disturbances. This tuning method has been reformulated multiple times, as it is well known for its inability to provide reasonable disturbance rejection, and consequently unable to reduce variability.
Sigifredo Nino, P. Eng.
Process Control Consultant
Author’s Response
I am pleased that you took the time to read the article and offer such thoughtful comments.
It is important to understand and agree on what robustness and aggressiveness mean. When I refer to aggressiveness, perhaps a better phrase would be speed of response. However, the goal of process control should be to maximize the performance of the process, not necessarily the performance of individual loops. As such, the tuning methodology should allow the user to choose the individual loop responses as needed to fulfill this goal. In some cases, this means very strong emphasis on load regulation; in some cases it means a much slower response and in some cases it means coordination of the closed-loop response of several loops. A loop tuning methodology that only has “one speed” (such as Z-N) does not accommodate this goal.
Furthermore, a tuning methodology for integrating process that does not allow the choice of a closed-loop speed of response, will likely result in an oscillatory response if the controller gain is reduced to slow down the response. This often leads to the comment that “PI tuning for integrating process will oscillate.” This inaccurate statement has been made by other control experts. Lambda tuning uses a set of rules that produces a first-order (second-order in the case of an integrating process) response with a closed-loop time constant specified by the user. This allows the tuner to select a faster or slower response as is appropriate for the process. It assumes the process has been accurately identified and is linear, since PID controllers are linear. When there is doubt about the process model or the process is sufficiently nonlinear, the tuner must use judgment, regardless of the selected tuning methodology, to ensure process stability.
Sometimes adaptive techniques are helpful. And again, this is an area where chemical engineers can provide particular insight. As one of my colleagues is fond of saying, “Show me the data.” We have published the results of innumerable examples where lambda tuning made quantifiable economic improvements to a process. I have not seen a case that I can recall where properly applied lambda tuning produced a poor response, let alone made things worse and would appreciate seeing one if there is such a case. (refer to the following articles by James Beall: Loop tuning basics: Integrating processes— www.isa.org/intech/201604basics; Loop tuning basics: Self-regulating processes — www.isa.org/intech/201606basics; Loop tuning basics: Complex process responses — www.isa.org/intech/201610web. These articles do not cover all the 11 process response types we see in the field but it is a good primer on the topic.)
Lou Heavner
Emerson Automation Solutions
Editor’s note:The two letters above are excerpts. The full letters can be found online at www.chemengonline.com
Postscripts, corrections
April, 2017, “The Future of Safety Sensors is Here Now,” pp. 22–26. On p. 24, Upskill’s (Herndon, Va.; www.upskill.io) software platform is incorrectly referred to as “Skyline.” It should be “Skylight.” This has been corrected in the online version of the article at www.chemengonline.com/the-future-of-safety-sensors-is-here-now