1

I'm learning control theory at university and I have to design PID controller for this object G(s) = 10/s(s+1); But my teacher requires settling time <10s and %overshoot <=2% Anyone can suggest me some methods to tuning PID parameters satisfying these requirements,please? Thanks

John Smith
  • 13
  • 3
  • Tuning is about getting the results you want, not it's accuracy. Any tuning method, properly applied, should get you a tuned system within your parameters. Even manual tuning will get you there (eventually). [There are a number of accepted methods](https://en.wikipedia.org/wiki/PID_controller#Loop_tuning) you need to pick the one that you (the tuner) are most comfortable with. – Ron Beyer Mar 30 '20 at 13:28
  • @RonBeyer Thanks for your advise. I have tried those methods but the result is not what i want ( maybe the reason is my exp) – John Smith Mar 30 '20 at 13:35
  • Yes, the tuning methods will get you a tuned system. Your instructor is giving you criteria of "when you can stop tuning and consider it "tuned"". The proper application of any of those methods will result in a stable system, tuning is one of the most difficult things to master and is especially difficult for people new to PID's. – Ron Beyer Mar 30 '20 at 13:38
  • In the real world, the plant parameters generally change at least slightly over time, due to air temperature and other environmental parameters. So there's rarely any point to trying to tune a controller to "high precision". It's more important to tune it so that there's enough margin for it to stay stable when the plant changes. – The Photon Mar 30 '20 at 15:05
  • What research have you done so far? And what does settling time <10% mean? – Chu Mar 30 '20 at 15:20
  • @ThePhoton Maybe I'm doing it in theory not real world – John Smith Mar 31 '20 at 23:22
  • @Chu my mistake. I mean 10s. Basically, if i use Z-N method the settling time is up to 40s – John Smith Mar 31 '20 at 23:24
  • Sounds like you need a critically damped system... – MadHatter Mar 31 '20 at 23:37
  • Also notice your G(s) already has an integrator term, so the integrator part might prove not so useful $$ G(s) = \frac{1}{s} \frac{10}{s+1}$$ – jDAQ Mar 31 '20 at 23:41
  • I suggest trying to use just the proportional part of the PID and trying to get a small rise time with overshoot of 50% or less, should be doable with some guesses (changing only Kp). Then, change the K_d to flatten that overshoot to less than 2%. – jDAQ Mar 31 '20 at 23:44

0 Answers0