I know that Foxboro Exact controllers (760 family) use overshooting and dampening to get the Exact alogorithm tuned and I would like to know if somebody has worked with that kind of algorithms before. If so, could give me more information. If not, is it possible to tune a regular PID controller using overshooting and dampening? Thanks

 

Stephane, I am sure others wil reply to this as well. It is indeed practicable to use Overshooting and Dampening as a "method" of tuning contrrollers - in fact, this is the only one that you can use to get the system right. Another name for it is trial and error - "Overshooting and Dampening" just sounds better. Any theoretical calculation method you use can only give an approximate response, and the Ziegler-Nichols methods in particular are aimed at quarter-amplitude damping. Throw this at an operator and he'll tell you it's "cycling". So you have to judiciously tweak the settings to get a satisfactory result. NOTE: What is considered "SATISFACTORY"!!?? depends very much on the process and what the specific requirements are. Very simply - If the loop is too oscillatory, you need to reduce the gain or increase the integral action time/reduce reset rate. You have to decide which. If the loop is cycling slowly, there is probably too much integral action so reduce it. If the loop is cycling quickly, there is probably too much gain. With damping/overshoot as the only objective, you have one too many degrees of freedom in a PI controller - 2 extra with PID. A good start with a loop with an unsatisfactory response is to back off the integral action - increase IAT by 50% or reduce reset rate by 30%. Now adjust the gain until you get an acceptable amount of damping or overshoot. Once you have found this value of gain. set the gain to 80% of this value. So if a good response is found with a gain of 2, knock it back to 1.6. Now gradually increase integral action by increasing reset rate/reducing IAT Do this until an acceptable overshoot is again reached. If the required overshoot is reached but the tail of the response (time for the PV to stabilise at the SP) is tto long, reduce the gain a bit before increasing integral action. CAUTION: Some very clever people are proudly using "non-interactive" algorithms in their controllers. These DO NOT follow the normal tuning rules - in particular, it is possible to get a setup where a reduction in gain makes the control system less stable. Since you have to do any final adjustments on a live plant, any tuning exercise needs to be very carefully planned with limits set on process excursions. (Operators are very good at monitoring tuning operations very closely and evicting any over-keen technicians from the control room if things get out of hand.) Bruce.

 

Dear Bruce, i think Stephane means about the Expert Adaptive Control Tuning regulators from Foxboro (760 series). This controler work with adaptive algorithms (minium squares, distribuided parameters, etc) which identify the parameters of the controlers doing iterations with forgetting factor, this algorithm tunne the control parameters regardless if the system is unknown.

I've worked with this algorithm programming it in PIC's and using it for control unknown systems simulated by OPAMPS.

Stephane, look for the minium square algorithm and there thou shall see how it works.