Responding to “The truth about controlling processes…”

A while back I saw a Walt Boyes’ post entitled, The truth about controlling processes… in which he captures the response of Walter Drieger to a post on the Automation List at Control.com.

Walter’s response had some pointed statements like:

All process control loops are nonlinear. That is why the math you learned in school is useless.

I thought I’d run Walt’s post by Emerson Principal Technologist, Terry Blevins, co-author of Advanced Control Unleashed and recognized Automation Hall of Fame honoree for his thoughts on the subject. Here is Terry’s response:

I normally don’t take the time to respond to blogs if I think the topic is not been well framed. However, in this case I am making an exception since I find the comments to be misleading and thus should not be left unchallenged.

For many years I was responsible for the design, implementation and startup of advanced and regulator control strategies for control systems installed in the pulp and paper industry. Commissioning the control was especially challenging on faster processes, such as boiler combustion control, since there was often little time to establish the control tuning. In many cases there was the opportunity to make a small change in the controller output, observe the dynamic response of the process and then set the tuning before placing the control in automatic. What I quickly learned was that control tuning must be based on an understanding of the process. Specifically, to tune PID controller feedback or feedforward strategies in single loop of multi-loop configurations such as cascade or override control, it is necessary to understand the dynamic response of the process to changes in process inputs.

Often times the process response to a step change in the process input may be described in general terms such as the process has low gain i.e. little change in the process output for a change in the process input or there was little delay in the process response or the process was slow to respond. However, to establish control tuning setting, it is necessary to describe the process response in quantifiable terms. For self-regulating processes, the open loop response is often characterized in terms of process gain, deadtime, and time constant. If the control is associated with an integrating process, then the response may be characterized by the integrating gain and deadtime. Such characterization of the process dynamic response is commonly called the step response model. Similarly, the process dynamic response associated with feedback control may be characterized or modeled under closed loop conditions in terms of ultimate gain and ultimate period.

The techniques described by David St. Clair in Controller Tuning and Control Loop Performance and any number of references on this subject of tuning are fundamentally based on a knowledge of process dynamic response, the ability to characterize (model) the response, and to use this understanding in setting control tuning.

Take a read of Walt’s post and Terry’s response and join the conversation.

Posted Monday, August 28th, 2006 under Education.

2 comments

  1. Responding to “The truth about controlling processes…” – Part 2

    Adding to the conversation of the Truth about controlling processes… post and response by Emerson’s Terry Blevins, is this response from Greg McMillan. Greg is the author of quite a few books on tuning and control which you can see…

  2. Like Terry, I found the original questions to be well “framed”; however I don’t agree that my answers were misleading. They were directed at a very specific series of questions and were intended to lead the questioner away from the classic Laplace transform approach and towards more practical guides like David St. Clair’s book.
    From the nature of the questions and the specific words and phrases used, I concluded that the Questioner (Q) was a recent graduate of a controls theory course based on the usual assumption of linear systems. Q’s questions all tended towards finding a technique for using process data to pre-calculate the tuning constants, much as he had done in school.
    Terry very correctly points out that an understanding of the process is necessary for successful controller tuning. This, coupled with the techniques described Nichols & Ziegler and St. Clair will lead to a successful conclusion. To ask a process engineer, “What are the time constant, dead time and process gain so that I can calculate the tuning constants and determine the degree of loop stability?” will not lead to a successful conclusion. I tried it. They (the process engineers) laughed at me and said, “Wait until startup. Then we’ll tune the loops. You can’t pre-calculate them.” So I learned — but I had to undergo a significant paradigm shift.
    Or do you disagree?
    Feed forward was mentioned. That is a horse of a different colour. Here specific process values must be obtained so that valid calculations can be made. These calculations are made in advance of configuring the system and then refined based on actual trials. Since feedforward does not benefit from feedback, normal tuning methods have no direct application. That is definitely beyond the scope of Tuning 101 and was not part of the original question set.
    Walter.

Leave a Reply