1

I'm trying to use a PID to control temperature, but I have 2 problems: the system cannot oscillate +/- 0.005C; and I need little to no overshoot. I'm testing/learning trying to control temperature using a light bulb and a class a pt-100 because the system I want to control is a little slow and I want to understand a method. Currently, I'm using the Ziegler-Nichols method, but it has some overshoot.

melpomene
  • 84,125
  • 8
  • 85
  • 148
user169808
  • 503
  • 1
  • 6
  • 27
  • You should use very low "Ki" coefficient; a fair amount of Kd can help in stabilizing the system. But temperature is hard to control because it has a high inertia. – linuxfan says Reinstate Monica Jul 01 '17 at 17:06
  • +/-0.005C seems pretty precise. How does that resolution compare to your output control system and your measurement system. Driving a 60W lightbulb with a 16 bit timer would give you milliwatt resolution for power in, and a 10bit ADC on a 5C full-scale range would only get you to 0.005C resolution. If your dT, measurement and control resolutions aren't about 1/10 of your physical system, you'll end up chasing noise. – Dave X Jan 10 '20 at 17:55

1 Answers1

0

Ziegler-Nichols is not a very robust for the nonlinearities you find in heat exchange and it also does not fit well slow processes.

For temperature control, I recommend a relay-feedback (square wave) and a step to find a second-order model for your system. Then I suggest the AMIGO tuning rule if your system deadtime is not too large (otherwise I recommend Shinskey formula for deadtime-dominant processes).

You find how to implement the relay-feedback and the AMIGO tunbing in the book "Advanced PID Control" by Astrom and Hagglund.

The formula on deadtime-dominant processes you find in "Feedback Controllers for the Process Industries" by Greg Shinskey.

I do not have experience with light-bulb element, but the pt-100 won't have enough accuracy for your requirement. The best pt100 won't have better than 0.05C. If your span is narrow and depending what you are measuring, a thermistor might be useful for better accuracy (but becareful with its calibration).

Prox
  • 41
  • 1
  • 4