PID Ziegler and Nichols

[garg29]

hi friends,
A old problem, still trying to understand, I'm trying to design a AutoTune PID Temperature controller using PIC16f877A uC & SSR type output.
I have read theories regarding AutoTune PID on many sites but still can’t get clear.

Ziegler and Nichols method says :-

a. Reduce the integrator and derivative gains to 0.
b. Increase Kp from 0 to some critical value Kp=Kcr at which sustained oscillations occur. If it does not occur then another method has to be applied.
c. Note the value Kcr and the corresponding period of sustained oscillation, Pcr

Kp=0.6Kcr; Ti=Pcr/2; Td=Pcr/8;

Now, my doubts are as follows:
1. It says Increase Kp (i.e. Proportional gain) from 0 to some critical
But, Proportional Band = 100/ Proportional gain
& I think Proportional Band ranges from 0-100%
so will I increase Proportional Band value(in steps of 1% ?????) from 0 to a critical value at which sustained oscillations occur?
Then get back the Kp value from Proportional Band = 100/ Kp
2. I’m familiar with C coding but, How can I make a C routine which would conclude whether the oscillations are sustained or not
3. For this proportional controlling part will I require a Cycle Time; If so how to decide what value has to be set for it.
4. How can I make a C routine which would calculate corresponding period of sustained oscillation
5. After I get Kp, Ti & Td how do I use them

I’m not even able to understand how to start designing. I’m trying to understand Autotune PID Temperature since past 3 months but don’t know If I would ever be able to design a Autotune PID Temperature controller.

Please do reply.

With Best Regards,
Garg

 

[FvM]

I don't think, that Ziegler/Nichols tuning rules are particularly suited for autotune, it's more a tool for manual tuning. These days, autotune is a standard with industrial controllers, but I never saw a controller utilizing Ziegler/Nichols. Most controller are estimating the control process parameters and optimized controller parameters from one ore better several step responses, e.g. by placing the controller in on-off mode. This way you get a significant stimulation of the control process without previous knowledge about it's parameters. But sonsidering the possible huge variation of control process parameters, an universal autotune approach can't always work.

Regarding your question about applying Ziegler/Nichols, I suggest to evaluate the method with a real control process (or at least in a simulation). You'll see, how it works.

 

[garg29]

Thanks for replying FvM please refer this link
https://digital.ni.com/public.nsf/allkb/D2E364AAF4D74EE6862575840049BC6E

Here they say "The PID Autotuning VI uses the Ziegler-Nichols tuning method"

I'm confused, Can you please elaborate more on how to design code for Autotuning.

I request you to please help me out.

Thanks,

With Best Regards,
Garg

 

[micropar]

Hi garg,

You need to use many DSP algorithms for true Auto-tuning at every feedback cycle. For these, you also need to consider using 32 bit Processors something like ARM or PIC32 for raw processing power and lots of in-built memory, difficult to achieve using 8 bit PIC16F8xx processor.

Regards,
--micropar--

 

[FvM]

I didn't say, that Ziegler/Nichols method can't be used for autotune. I said, I haven't seen it with industrial controllers. In any case, all methods that temporarily change the controller parameters or apply large stimulus signals to the control process are only suited for single autotune action, e.g. during comissioning. Continuous auto tuning, as addressed by micropar, needs other methods, e.g. estimating the control process parameters from in- and output observed with "natural" stimuli.