Loop Control

Here's how to improve control loops

McMillan and Weiner talk with Jacques Smuts about improving control loops

By Greg McMillan, Stan Weiner

Greg: I had the pleasure of interviewing Jacques Smuts on improving control loops,. a subject near and dear to my heart. Jacques is the president of OptiControls Inc and the author of the book Process Control for Practitioners that provides a very understandable and practical approach to control loop optimization.

Stan: What do you do for plants?

Jacques: I do process control consulting and training to improve the performance of control loops. You would think after 20 years it would become old hat, but I am constantly surprised at how key fundamentals are not understood by people responsible for tuning.

Greg: In the midst of teaching a course on opportunities in process control at a plant I used to work at outside New Orleans, I realized the veterans were hanging on each word, but the young engineers were pretty much blowing me off. I found out most of their job was paper work in a Six Sigma program, and none had spent any time trying to tune or improve loops. As I lead the ISA Mentor Program, I am astonished how little time my protégés have to work on control loops.

Jacques: I am lucky that I can focus on tuning loops. This has unfortunately been reduced to a small fraction of what practitioners at plants are responsible for today. Tuning is a skill that requires practice. Many don't appreciate how much you really have to know beyond the basic tuning relationships. You need to know if the valve works properly. You need understand if a process needs to be tuned fast or slow and whether you need to schedule tuning. I enjoy the depth of the field. We lose sight of what the user needs to know and what the user does know. Increasingly, the user has more responsibility and less time for tuning and learning.

Greg: I was impressed with your chapter on valve problems where you describe tests for the various sources of limit cycling. Most other books don't make a distinction between dead band and stiction or the importance of valve problems in general. In university textbooks, the valve is perfect, and the whole emphasis is on a process time constant and gain and a setpoint response, reducing the loop to a servo problem. This may work for aerospace, but not for industrial processes with sloppy valves, slow measurements, analyzer sample and cycle times, large and variable dead time, interactions, unmeasured disturbances and non-self-regulating responses.

Stan: What has been the impact of digital positioners?

Jacques: Nowadays I find a lot less dead band, but also some erratic behavior from weird positioner tuning. I am amazed at how many tuning settings are on a modern positioner. Each manufacturer has a proprietary algorithm and tuning parameters. Tuning guidelines are slim to none. It helps to know the positioner is a closed loop controller that could have overshoot and limit cycles. Also, be aware that the response time depends on step size. I have seen a response time increase from 2 sec for a 2% change to 30 sec for 0.5% change. I look at past history to see the size of increments in the PID output to determine what the controller will have to deal with. I have also seen a positioner develop a limit cycle with a flow loop in manual.

Greg: The problem may be caused by the combination of stiction in the valve and integral action in the positioner. The solution would be to turn off integral action in the positioner or turn on integral dead band in the positioner. In a way, the old analog positioners were smarter than today's digital positioners by being proportional-only controllers. Controller offset is inversely proportional to controller gain and is very small because the positioner gain is so high (e.g., gain > 50). Offsets may be a cosmetic concern when manually stroking a valve, but in practice are inconsequential because the process controller will manipulate the signal as needed, eliminating the effect of the offset.

Stan: Are auto tuners the complete solution?

Jacques: People expect the software to be an engineer rather than a tool. Most people try, fail, get frustrated and stop using tuning software. Many try to tune from bad data, not realizing that more than minimal knowledge is needed. Loop tuning and performance monitoring software requires human skills to get the most out of them.

Greg: I found a production unit where all the gain settings were 1, and the reset settings were 1 repeat per minute. For the loops that were oscillating, the operator narrowed the output limits to achieve a sort of on-off control. When instructed the loops needed to be tuned, the operators were turned loose on tuning. Do you see operators tuning loops?

Jacques: Some plants have operators tuning loops. Some operators have had a class on tuning, but when I ask, they tell me the math was too much. If the Zeigler-Nichols tuning relationships are over their heads, you have problems. We do need to take advantage of their first-hand knowledge. Operators know important aspects of equipment performance, but sometimes have their own interpretations of the causes for poor control. You need to get operator observations rather than conclusions and explore the cause-and-effect relationships. In a liquid-gas separator, operators said you needed to close the outlet valve first and then open it to lower the level; otherwise the level would keep going up. I was tempted to discount this requirement, but found out under certain conditions a vortex formed. Closing the valve broke the vortex.

Stan: Is there a considerable amount of hand-holding needed?

Jacques: Patience, explanations and time are essential, particularly for systems with significant dead time. Many years ago I worked at a production plant for whiskey. The dry distiller grain solids (DDGS) is a valuable by-product for cattle feed, and was dried in a kiln with a dead time of about 9 minutes. The main disturbance came from the level control of a syrup tank that feeds into the wet grain and was manually set by the operators because level control was deemed not possible. I tuned the level controller and added feed-forward control to the kiln. Operators said it wouldn't work. I had a difficult time getting the operators to leave the moisture controller in automatic because the control was not smooth because of clumps. I sat with the operators and showed them the controller had taken all the action they would take (e.g. 40%), and there was nothing more to do than wait 9 minutes.

Greg: Looks like what the operators needed was a moisture calculation one dead time into the future on a trend chart described in Control Talk Blog "Future PV Values Are the Future."  Dead time is the most difficult thing for humans for deal with. Tuning can take the dead time into account avoiding overreaction and providing a consistent response. Does tuning ever become boring?

Jacques: After working on thousands of loops, you would think you have seen everything, but almost every plant I work at I see something new. I found a controller with a derivative time of 300 minutes. The low output limit was 75%, and the high output limit was 78%, so the loop was essentially in manual. Recently I found loops on an oil rig that were never tuned since they were commissioned in 2004. No attempt was made to rectify severely oscillating loops. They just lived with them unless they caused a shutdown.

Greg: What about configuration problems?

Jacques: Configuration can prevent a loop from ever performing properly. Configuration engineers often haven't had time to learn tuning. I had an override control system that was not working. The initialization was done in the wrong blocks. I had to wait till shutdown to have it fixed. I don't need to know how to do configuration for each of the many types of DCS, but I do need to recognize when the configuration is a problem.

Stan: Before we retired, we saw a dramatic decline in technical training. A 12-week instrument and control course for new employees was decreased to 8 weeks, 5 weeks, 2 weeks, and finally no weeks. What do you see happening with training for your customers?

Jacques: One technical training company says they cannot fill their classes in the United States. Yet the need is greater than ever. Jobs are so spread out. The question is how do you make tuning a hot topic? If I can just get into a plant and tune a problematic loop or two, the results often convince the customer of the need for a consulting/training program.

Greg: What do you recommend as the best training program?

Jacques: The best results are gained from a class covering topics in my book and containing a hands-on component. A program must include users working hands-on with the loops, because if the training is just a class, the attendees go back to work and may not try out what they have learned for weeks or ever.

Stan: What do you say about the use of derivative?

Jacques: The benefit of derivative diminishes as the total loop dead time becomes greater than the open loop time constant. For derivative to be effective, the process response must predictably be in the same direction for at least as long as the derivative time. The improvement offered by derivative is evident, but not dramatic for most processes.

Greg: For slow integrating and, especially, runaway process responses with secondary time constants (e.g., heat transfer surface lags and thermowell lags in batch and exothermic reactor temperature loops), the use of derivative becomes more important. What do you do about noise, and what do you think about notch filters?

Jacques: I use filtering only when really necessary. Actually, I am more likely to take a filter out than to add one. Notch and higher-order filters may have some advantages in instruments where the signal noise is definitive. Process noise is too unpredictable for these more sophisticated filters to be generally applicable.

Greg: What did you find out from your tests of the tuning rules from the H.A. Fertik paper "Tuning Controllers for Noisy Processes" (ISA Transactions, Vol. 14, No. 4, 1975)?

Jacques: The response of his PID tuning rules on first-order plus dead time systems was very oscillatory. It turns out the Fertik tuning rules were applicable for systems with two or more equal time constants. As the secondary time constant became closer to the primary time constant, the response became much smoother, although the return to setpoint for a load disturbance was slow and had some hesitation.

Stan: How would you sum up tuning?

Jacques: You need to understand how to adjust a PID controller's response to match the process dynamics, know your process' control objective and recognize when configuration, equipment and valve problems need to be addressed.

Top Ten Signs Loops Need to be Tuned

(10) There is no process control engineer in sight or on-site.
(9) When asked what the controller gain setting is, the answer is given in %.
(8) When asked what the controller reset time setting is, the answer is given in repeats/min.
(7) The data historian compression setting is 25%.
(6) There is more recycle than product.
(5) Valves are wearing out.
(4) Tempers are wearing thin.
(3) Operators are placing bets on what loop will cause the next shutdown.
(2) The output limits are set to keep the valve from moving.
(1) The gain, reset and rate settings are all 1.0.