One of our processes requires water at 210Â°F, which is heated by a steam exchanger. Demand varies and can change stepwise. We have a small, continuous bypass flow so we can maintain temperature even when there is no flow of water. The temperature controller is tuned for this minimum-flow condition so that temperature will be stable and on set point whenever there is a demand for water.
Our problem is that temperature doesn’;t stay on set point when demand increases, especially when it changes rapidly. It seems to control below set point on rising loads and above it on falling loads. But if we tighten the PID settings during high load, the temperature cycles badly when we return to minimum flow. Is this a place for adaptive or self-tuning control? Is there another solution?
Feed-forward will take the heat off
Think of a heat exchanger as a controller for which the gain varies inversely with flow. At minimum water flows, a small increase in steam flow will produce a certain steady-state rise in water temperature. When the water flow is doubled, the same size increase in steam rate results in only half the water temperature rise. This varying gain can be compensated for by using an equal-percentage steam valve for which the gain varies directly with its delivered flow.
The exchanger also has a time constant and dead time, which vary inversely with flow. Since optimum integral and derivative settings are related to process dead time, they can only be optimized at one flow, and the worst case (slowest dynamics) is at minimum flow. If your controller has only one value for its proportional-integral-derivative (PID) settings, optimize the integral for minimum flow and the derivative for maximum flow. Some controllers have multiple sets of PID parameters that can be invoked as a function of another variable, such as water flow or controller output — this would allow optimum tuning at two or three loads. Ideally, integral and derivative settings should be programmed to vary inversely with measured water flow.
Feed-forward control will help immensely. Measured water flow is passed through a lead-lag unit and multiplied by the output of the temperature controller to set steam flow in cascade or to position a linear steam valve. In this case, the multiplier compensates for the variable process gain. Feed-forward is fast and minimizes the need for integral and derivative action.
F. Greg Shinskey, process control consultant
North Sandwich, N.H.
Correct the derivative
It appears that the derivative is not set correctly and that there is some dead time in your process. Adaptive or self-tuning control may not solve this problem. An advanced controller with more than simple PID could best solve this.
Joel Heidbreder, principal engineer
Monsanto, St. Louis
Install bypass valve
A complex controller may require technical training and time to install while incurring costly downtime. I would suggest replacing the continuous bypass flow with an automated bypass/divert valve that provides full flow to and from the process depending on load. Insert a check valve at the process discharge to prevent bypassed water from backfilling.
To ensure temperature control, modify the bypass piping to achieve a flow rate that is equivalent to the flow that would pass through the process. This would also prevent water hammer when the divert valve shifts (regulating the discharge air from the solenoid also helps). The control RTD should be located at the exchanger discharge and provide feedback to the steam valve to control the water temperature.
Ron Johnson, process engineer
Alpharma Pharmaceuticals, Baltimore
Increase size of exchanger
One of the problems associated with intermittent flow is the temperature lag between low- and high-flow conditions. Adaptive control may be useful if the upsets are small and regular.
Some other ideas you might try include:
1. Increase the size of the heat exchanger so it can better handle the heat requirement during high-flow conditions. Also increase the bypass flow to minimize overshooting the temperature.
2. Create a fluid reservoir to handle the upsets. Draw from the reservoir during high-flow periods, which will tame some of the temperature variations and allow your control system to react.
Jim Stahl, engineering manager
Dekoron-Unitherm, Cape Coral, Fla.
Use a power trap
We’;ve had success with a power trap system. In this system, the inlet control valve sets the steam flow to the steam exchanger. The condensate from the exchanger drains by gravity to a reservoir drum, below which the power traps either pump or trap the condensate to the return header.
This system works well when piped correctly; the reservoir and power traps need to be equalized to the exchanger at the condensate outlet high point with air vent and check valve to the pad; properly sized check valves on the power trap inlet and outlet; and the motive steam should come from a constant steam-supply source to help pressurize the trap and discharge condensate when the inlet control valve is throttling down. However, this system requires routine maintenance about every 18 months. Otherwise, excessive water hammering will damage the exchanger, power traps and check valves.
Miguel Y. Chen, project engineer
Chevron Phillips Chemical Co., Baytown, Texas
Implement gain scheduling
Since there are fluctuations in the water flow rate, the controller will need to be retuned for the complete range of load conditions. Gain scheduling may assist in this case. When the process variable (PV) is, say 0%-30% of range, the controller parameters are tuned for this region. When the PV moves to the next region, perhaps 30%-70%, the controller tuning parameters are switched to another set (previously tested and determined). When the PV moves to the highest region, above 70%, the tuning parameters are switched to new settings.
These ranges are just examples and can be changed as needed and should be verified with testing. This method also assumes that the process is relatively linear between each region.
David G. Rolfe, P.E., process controls engineer
Ashland Chemical, Columbus, Ohio
Switch between two controllers
You can use two controllers in parallel, but only use one at a time. Add a switch to sense when the flow exceeds what the controller with the low-flow PID settings can handle. Use the switch to change to the other controller that is tuned for the high-flow conditions. You can also put a limit switch on the temperature control valve and set it at a point just above the low-flow cycle range to perform the same function as the flow switch.
Terry V. Molloy, P.E., president
CMES Inc., Novato, Calif.
Opt for feed-forward control
Use feed-forward control. Install flowmeters on each water stream and use them to calculate the percentage of the total flow. Use the data as a feed-forward input to the temperature controller. We did just that and reduced variability in a three-user system from Â± 30 degrees F to about Â± 5 degrees F.
Curt Caveney, principal process development engineer
Eastman Chemical Co., Kingsport, Tenn.
Base feed-forward on water rate
Add feed-forward control based on water flow rate. Reset the feedback when the flow rate is stable. Many hotel hot water heaters successfully use such a strategy.
Andy Schwartz, chairman
Keeshan & Bost Chemical Co., Manvel, Texas
Try controller with two modes
The process dynamics are very different when operating at minimum water flow compared with normal load conditions. Effective values for both the process gain and the process time constant tend to increase inversely with the water flow rate. The temperature offset in rising (or falling) loads is inversely proportional to the controller gain. This means that if the temperature controller is tuned for the minimum flow condition, the controller gain and reset rate will be too small when the load increases, resulting in excessive offset during load ramping. On the other hand, if the temperature controller is tuned for the normal flow condition, the control action will be too aggressive at minimum flow, resulting in oscillation.
You should try a gain schedule controller or a controller with two control modes. We do not believe that feed-forward control by itself will handle the problem because it will not compensate for the change in loop dynamics as the load changes from zero to normal.
Irving Lefkowitz, vice president of engineering
ControlSoft Inc., Cleveland
An electric cable bundle was supported by cable hangers. The hooks at the ends of the cable hangers were hooked over the top of a metal strip (see top figure). The electric cables had to be lowered to the ground to provide access to whatever lay behind them. When the cables were returned to their supports, they were put back as shown in the lower part of the figure. This increased the load on the upper hooks. One hook eventually failed, increasing the load on the adjacent hooks. They then failed. A 200-ft. (60 m) length of cables fell, injuring one man. How can we avoid this situation in the future?
Send us your comments, suggestions or solutions for this question by Nov. 29. We’;ll include as many of them as possible in the January 2005 issue. Send visuals, too — a sketch is fine. E-mail us at ProcessPuzzler@putman.net or mail to ProcessPuzzler, Chemical Processing, 555 W. Pierce Rd., Suite 301, Itasca, IL 60143. Fax: (630) 467-1120. Please include your name, title, location and company affiliation in the response.