The following tip is from the ISA book by Greg McMillan and Hunter Vegas titled 101 Tips for a Successful Automation Career, inspired by the ISA Mentor Program. This is Tip #92, and was written by Greg.
The reset time can be approximated as a factor of the deadtime, but this dimensionless factor can range from 0.5 to 40 depending upon the type of process. The control literature has focused on self-regulating processes with process time constants in the same range as the deadtime and with little treatment of deadtime dominant processes, almost no consideration of batch processes and other integrating processes (the exception is Lambda tuning), and essentially no analysis of runaway processes. This deficiency could merit a book in itself, but the point here is that engineers and operators have confusion and uncertainty with regard to reset time settings.
I found that for deadtime dominant processes (i.e., processes with a deadtime much larger than the open loop time constant) the approach back to setpoint would be slow for load changes and would falter (i.e., flatten or hesitate) for setpoint changes. The solution was to reduce the reset time to a low limit of 0.5 times the deadtime for essentially pure deadtime processes.
For integrating processes, I realized that the reset time factor depended upon the controller gain setting. If you use the controller gain for maximum load rejection, the reset time is four times the deadtime. However, if you decrease the controller gain, you must proportionally increase the reset time, a non-intuitive effect since we are taught that decreasing controller gain always helps to reduce oscillations.
Reactors in highly exothermic processes (e.g., polymerization reactors) can exhibit a runaway temperature response. The use of too much integral action can be dangerous. In some cases, a proportional-derivative mode is used. I found I had to increase the reset time to be 40 times the deadtime to prevent excessive overshoot in simulation tests.
Feedforward needs to be made smarter as well. Terry Tolliver, the world’s best distillation control expert in my estimation, found that for column temperature control, flow feedforward could make the effect of a disturbance worse. A feed change could provide a temperature change in the direction to compensate for a concentration change or some other unknown disturbance that was eliminated by flow feedforward. I found in simulation tests that, in general, for temperature and concentration control, an enhanced PID could correct the flow feedforward signal to reduce the feedback correction.
Concept: A smarter reset time can be computed online, based on a future PV value and ramp rate, to provide a balance between integral and proportional/derivative mode action to suppress overshoot and reduce faltering. The ramp rate can be used to compute a rise time to get the output to come off of an output limit at the right time to prevent overshoot. A future PV value and ramp rate can be used to prevent feedforward from doing more harm than good from unknowns. An enhanced PID can be set up to correct for significant errors in the feedforward signal to reduce the feedback correction.
Details: If the future PV value and ramp rate show that overshoot will occur, increase the reset time. If the future PV value and ramp rate show a lingering offset, faltering, or slowness in the approach to setpoint, decrease the reset time. Suspend the adjustment of reset time when the future PV is either too far away or within the allowable control error around setpoint. Continue the adjustment of reset time for each significant load and setpoint change. Set upper and lower reset time limits to prevent problems. Start out with tight limits until you gain confidence in the procedure. For processes where the analyzer cycle time is much larger than the process time constant, use an enhanced PID so the reset time does not depend upon analyzer cycle time. For large setpoint changes, abnormal operation, batch operations, start-ups, and surge control, compute the rise time from the ramp rate and the resulting reset time to get the PID output off an output limit so the PV ends up within an allowable error around setpoint.
If the future PV value and ramp rate based on both the actual response and the feedforward response are moving in the same direction, suspend feedforward action. Make a bumpless transition back to feedforward action when the future PV value shows a return to setpoint, indicating that the unknown disturbance has been accounted for by feedback action. Use an enhanced PID to improve the feedforward signal. This adaptive controller’s PV is current feedback correction, setpoint is zero feedback correction, and output is a flow feedforward signal bias or gain. Determine from an analysis of manipulated flow versus feedforward flow for various conditions whether the feedforward signal has an offset or slope that needs to be corrected. Set a threshold sensitivity limit to ignore insignificant corrections. Set directional setpoint rate limits in the analog output (AO) block used to adjust the bias or gain so that correction can be made slow enough not to upset the process controller. Set low and conservative high AO setpoint rate limits. Make sure the adaptive controller has external-reset feedback set up with the AO block so that the adaptive controller output does not change faster than the rate limited AO block response. The feedforward correction may need to be faster in one direction.
Watch-outs: Lambda tuning for self-regulating processes shows the reset time as a factor of the open time constant. However, if you use the Lambda tuning settings for integrating processes and near-integrating processes, the reset time is three times the deadtime giving better recovery from disturbances. Improper controller gains can cause excessive oscillations or too slow of a response that cannot be corrected by a smart reset or smart feedforward. The controller gain for best practical load rejection can be universally approximated as an 0.5 factor multiplied by the inverse of the product of the integrating process gain and deadtime. For self-regulating processes, the second equation in Tip #89 can be used to convert to the more familiar expression of controller gain in terms of a steady-state open loop gain and open loop time constant.
Exceptions: The future PV value calculation will not work for deadtime dominant processes, noisy processes, and valves with excessive deadband and stick-slip. For processes with the loop deadtime approaching the execution time of the PID, the calculations are too late to make any correction.
Insight: Smart reset can reduce overshoot and faltering, and smart feedforward can provide a more complete and accurate feedforward action.
Rule of Thumb: Use future PV value and ramp rate and an enhanced PID to make the reset time and feedforward action smarter.
Hunter Vegas, P.E., holds a B.S.E.E. degree from Tulane University and an M.B.A. from Wake Forest University. His job titles have included instrument engineer, production engineer, instrumentation group leader, principal automation engineer, and unit production manager. In 2001, he joined Avid Solutions, Inc., as an engineering manager and lead project engineer, where he works today. Hunter has executed nearly 2,000 instrumentation and control projects over his career, with budgets ranging from a few thousand to millions of dollars. He is proficient in field instrumentation sizing and selection, safety interlock design, electrical design, advanced control strategy, and numerous control system hardware and software platforms.