In the ISA Automation Week Mentor Program I am providing guidance for extremely talented individuals from Argentina, Brazil, Malaysia, Mexico, Saudi Arabia, and the USA. We will be sharing a question and the answers each week. If you would like to provide additional answers, please send them to Susan Colwell at ISA. This question is from Hector Torres of Mexico:How do you tune a temperature control loop when auto-tuning takes too long?
- I suggest the near-integrating method where you just identify the deadtime and initial ramp rate. This can reduce the test time by 94% as demonstrated in my Demo-Seminar (Deminar #6) http://modelingandcontrol.com/deminars/. The method can also be used to identify a feedforward deadtime and ramp rate for the estimation of the feedforward gain and dynamic compensation based on the difference in feedback and feedforward dead times and near-integrating process gains.
- You need to increase the integral time for an integrating process when the controller gain is set much less than what is allowed for maximum disturbance rejection. Most loops on integrating processes are using much less than the allowed controller gain because a higher controller gain causes rapid large movements of the output that scares operations or upsets other loops, Often a simple AO block or setpoint velocity limit with the dynamic reset limit option solves the actual or perceived problem. If the controller gain is decreased, the integral time must be proportionally increased to retain the optimum product of the controller gain and integral time to prevent slow rolling oscillations. Most control theory books and most people don’t realize that this counter intuitive relationship where increasing the controller gain allows you to decrease the integral time.
- From the control literature, we are familiar with the situation where too high of a controller gain (too much proportional action) or too low an integral time or reset time (too much integral or reset action) will cause oscillations and overshoot. In integrating responses, too large of an integral time or too small of a controller gain can cause oscillations and overshoot as well. While the oscillations are not as dramatic as going unstable, the effect is significant and confusing for integrating processes. For runaway processes as encountered with highly exothermic reactors (e.g. polymerization reactors), too low of a temperature controller gain can cause a dangerous runaway. Tests in manual for identification of dynamics and tuning are often not permitted because the temperature response can accelerate reaching a point of no return. Fortunately, the near integrator tuning method works well here with the controller in automatic for a setpoint change. The high controller gain and high integral times for these reactor temperature loops causes a step change in the controller output useful for the near-integrator method.
- The user must take into account the effect of engineering units on tuning settings. Reset settings have time units that may be different than what is used in the identification of the loop deadtime and ramp rate. The reset setting in repeats per second is the inverse of the reset setting in seconds. A controller gain should be dimensionless. If the units are per cent for the proportional mode tuning setting, the setting is likely proportional band that is 100% divided by a dimensionless controller gain. Nearly all PID algorithms work on percent inputs and outputs despite the operator trend, graphics, and faceplate being in engineering units. To compute the controller gain the PID process variable, setpoint, and output must be converted to percent signals based on scale ranges. Thus, scale ranges affect the controller gain setting. There have been some rare cases of a PID algorithm in a programmable logic controller (PLC) working in engineering units resulting in very bizarre controller gain settings.
For more information see the upcoming Jan-Feb article “PID Tuning Rules!”