Designing PID responses
There are many ways of setting PID strategy to make a controller respond the way you want it to. It all depends on what you think should happen.
In an earlier life, I was overseeing commissioning of a control system that our company had manufactured as part of a larger flue gas conditioning package. I was on the site with the system integrator that we had used to configure the controller. The customer wanted us to tune the main control loop but the integrator said that part of the process fell outside of his scope of supply. He did not want to do any loop tuning, no way, no how. Given that the whole project was not going well, I pressed the request and he told me that loop behavior needs to be tailored to the process and he wasn't going to do it. Ultimately, after much haggling, the customer relented and did the tuning.
At that time I didn't fully appreciate the substance of the discussion. Either a loop works or it doesn't, right? Eventually I came to understand that the way a system responds has to be appropriate for the application. When we write articles on the topic of PID control for Control Engineering , we have to approach these subtleties as best we can. Usually when Vance VanDoren discusses loop tuning, he does it in the context of setpoint changes rather than compensating for upsets. This is deliberate, and Vance and I have discussed that how you tune for setpoint changes can be much different than how you want your controller to respond to an upset. This topic will most certainly be the basis for a feature article one of these months.
Until then, here is a brief discussion of the topic that may get you thinking. Farther down the page there are three pairs of diagrams with a process variable and setpoint. Each pair (left and right) illustrates how a particular tuning strategy responds to a set point change (left) compared to an upset (right) of similar magnitude. The upper half of the graph shows what the variable is doing, and the lower half shows the controller output. Each scenario illustrates the same magnitude change, but with a different approach to compensating for it.
These charts were sent by Bob Rice, director of solutions engineering for Control Station. He includes the following explanation:
"The primary difference between a load (disturbance/upset) response and a set point response lies in the objective the end-user is typically trying to achieve. In most cases, users trying to get a good set point response are looking for a fast but smooth response which usually comes from the use of less aggressive tuning parameters. If the user is trying to reject a disturbance, then they are after a much faster response using more aggressive tunings. Therefore, they are at best conflicting control objectives. You cannot be both quick and responsive, while being smooth with little to no overshoot. There is always a trade-off when trying to tune for both set point tracking and load rejection using a typical PID-type controller. There is an‘I-PD' type controller, which is a ‘proportional on measurement' approach vs. ‘proportional on error,' which can improve the set-point tracking response under aggressive PI-control."
-Peter Welander, process industries editor, PWelander@cfemedia.com
Control Engineering Process & Advanced Control Monthly eNewsletter
Register here to select your choice of free eNewsletters.