Greg: In recognition of my home in Texas, we thought we would round up the puzzlers that have been running wild this year.
Stan: I was going to put on the requisite hat, boots, chaps and spurs, but then I thought a T-shirt, sandals and shorts would be more comfortable sitting poolside and wouldn’t scare the neighbors as much here in Florida.
Greg: Truth be told, while I have been accused of being a cowboy, the closest I have been to the traditional paraphernalia has been when attending rodeos and local barbeques. Besides, on the “spur” of the moment, the dog lying at my feet might accidentally get a kick out of more than just this column.
Stan: The first Puzzler out of the chute in January was this: “For what situation would it be sufficient to simply set a fixed control valve position without any feedback correction?
Greg: Well if you didn’t need to be exactly at set point, and there were no set point changes or disturbances, you might be able by trial and error to come up with a controller output. However, this scenario is true only in the calculations and simulations taught in universities. Since these students become the process engineers who often write the control system definitions that the configuration engineers dutifully follow, the full power of feedback control ends up overlooked. This is particularly noticeable in batch processes where a simple properly tuned PID loop could take the place and do much better than a best-guess sequence of fixed-valve positions.
Stan: Paul Botzman lassoed the February puzzler that posed the age-old question, “What problems would you expect when a PID controller is used for concentration control of a process with a dead time of 120 minutes and a time constant of 60 minutes?”
Paul: Whenever dead time is longer than time constant, it is a major dilemma for PID control. It is a typical problem for water and wastewater applications. With such a long dead time, PID control will continue to increase its output due to integral actions. The system tends to end up with overshoot, then undershoot and a very oscillatory response. If you reduce or remove integral actions, the controller loses its ability to track the set points. The real solution for this type of problem is multiple inputs/feed-forward, model-based control. In model-based control, the dead time of the process is built into the controller. Hence, the controller will not react radically during dead time periods. Multiple inputs/feed-forward control allows one to include the effect of flow or chemical concentrations of interest. It has been proven in the field that model-based controllers can reduce long dead-time chemical variations by 80%.
Greg: Dead-time-dominant loops necessitate the controller’s rate time be set to zero and the controller gain detuned to a small fraction of the inverse of the product of the valve, process and measurement gain. Also, disturbances with time constants faster than one-quarter of the dead time will reach 98% of their full value before the controller can do anything. The slow integral action, the diminished proportional action, the omission of any anticipatory response, and the increase in the relative speed of upsets increase loop variability.
Stan: Not that you can choose the animal to ride in the rodeo, but it helps to recognize the meanest and wildest bucking bronco or bull. Greg Shinskey was right on top of the March puzzler, “What type of disturbance is too fast and repetitive for recovery by feedback control?”
Greg: Two types of disturbance too fast and repetitive for recovery by feedback control are
- Noise, defined as random variations in the controlled variable, that are beyond the bandwidth of the controller. If the controller responds to it, the result may be only the unnecessary exercise of the final element, possibly exacerbating the disturbance. Noise is typically injected at the controlled variable with no intervening dynamics.
- Sudden real-load changes, such as those caused by on-off devices, are too fast for the feedback controller to respond, owing to slow dynamics of the controlled variable, principally dead time. You can regulate in the face of these disturbances using feed-forward control, trimmed by the feedback controller. Real load changes are typically injected upstream of the dominant process dynamics.
Greg: Hunter Vegas reigned in the most exciting disturbance I have seen that has made a feedback loop helpless.
Hunter: My initial guess on this one would be the flow on a surging compressor. The cycles can be very quick (less than a second), and if you don’t do something very fast to restore normal flow through the compressor, you’ll have a very expensive paperweight.
Stan: In April, we were checking whether we had put our readers to sleep by asking, “What is the valve flow coefficient (CV) for one gpm of water at standard conditions and a 1-psi pressure drop across the valve?” We got the correct answer “1” from William Bloom, David Greene and Hunter Vegas.
Greg: In May we asked “What happens when you put a booster on the output of a valve positioner?” Based on my experience in applications to speed up large valves, the combination of a booster in series with a positioner will result in limit cycle unless some of the input to the booster is bypassed around the booster. Some boosters have a built-in bypass adjustment (e.g., tubing and a needle valve). If not, one must be added. New digital positioners may be able to be tuned to help deal with this situation so that not so much air needs to be bypassed. Bypassing air around the booster reduces its performance, so it is desirable to have it at the minimum needed to assure stability.
10. Any application cowboy—where have all the cowboys gone?
9. Retired experts—they know what doesn’t work.
8. Any systems troubleshooter who can actually troubleshoot.
7. Sufficiently tested software.
6. Installation detail—what a quaint idea!
5. Vendor who knows whether the instrument will work.
4. Your company’s stock (spin-offs excluded) no matter how little you think of it, especially if you were downsized, because the smaller you feel, the bigger the stock price. In fact, Wall Street has a new performance index—the profit-to-engineer ratio with a divisor of zero being the optimum.
3. Your clothes from the seventies—you never know when they will be a hit again or how far casual Friday will go.
2. 3D glasses—ask your friends to wear red and blue tie-dye clothes and hurl themselves at you.
1. Your dongle, even if it is a beta version.
This Month’s Puzzler: Is One of These Things Just Like the Other?
Are the dead times identified from a step change in the controller output and setpoint identical?
Send an e-mail with your answer to the Puzzler, CONTROL questions, or comments to firstname.lastname@example.org.
|About the Authors|