I would like to ask you to share with your experiences related to fuzzy supervisory control. I think about such a algorithm which is able to change parameters of a classical PID controller, based on a human experience due to plant's dynamics variations, disturbances etc.
I wonder if such a algorithm would be able to make a control process more robust.
Dear Slawomir,
I suggest you alinks and attached files in topics.
- Multivariable fuzzy supervisory control for the laminar cooling process ...
https://www.researchgate.net/.../3332263_Multivariabl...
- Perspectives of Fuzzy Control: Lights and Shadows
personales.upv.es/.../C49IS04IEEEXPL01344631.pdf
- INDUSTRIAL INTELLIGENT CONTROLLERS:FLC (FUZZY LOGIC ...
machineryequipmentonline.com/.../industrial-intellige...
Best regards
I would say it is normal practice for control of advanced machinery to switch PID controllers between parameter sets or filter settings, when system characteristics change. This can be manual or automatic.
Normally, this technology is developed in-house, though, and not shared.
A simple PID very quickly gets inadequate in real world control. Gain scheduling is the most basic automated parameter modification.
Dear Olof, thank you for your response. It is widely known that PID controller should be re-tuned in case of any significant dynamics variation. Gain scheduling it is one way which in practice we can do that. But this technique rely on a precisely developed model i.e. it is known how parameters will change their values when a working point will change. Correct me If I am wrong.
A fuzzy supervisor based rather on a behaviour of the whole control system (controller + plant). That behaviour could be presented as a e.g. settling time, oscillation index etc. It is a different way than a gain scheduling.
Anyway I would like to find people who confirm that fuzzy technique really works in practice as a supervisor of PID controllers, not only in publications.
This is an interesting question.
I have not seen fuzzy supervisory control used in practice.
I have found it more common to abandon the PID than to try to supervise it, when the system becomes too variable. Having a more general programming capability available in most controllers, it is simpler to program the control in some other manner, leaving the analytical solution of control problems to the theorists :-)
Olof
Dear Olof, my question was simply related to the fact that I am programmer of industrial several controllers and some of them e.g. controllers from Bernecker&Rainer provide a lot of opportunities for practitioners. I agree that sometimes faster and simpler is to find heuristic algorithm and implement it.
But sometimes as a scientist I would like to test some theoretically developed solution, simply to proof its efficiency or simply... throw it away as a useless for the industry (to sophisticated or not so nice than in a related publication).
I find several publications related to the fuzzy supervisors for PID e.g. Design and implementation of a fuzzy supervisor for on-line
compensation of nonlinearities: An instability avoidance module (Engineering Applications of Artificial Intelligence 19 (2006) 323–333) and after that I decided to do sth like information mining according to fuzzy supervisors.
- Badges
- Science topic
- Similar topics
- Engineering
- Control Theory
- Control Systems
- Control Systems Engineering