Help me debug a code for auto tune pid

[lats]

auto tune pid?

Hi all,

i am new to pid temperature controller. i have studied many links and basics i got a link on this forum were a person has explained pid very basic .the link is


Simple PID

PID = GainP * actual error + GainI * SUM(previous errors) + GainD * (actual error - last error)

error = sp(set point) - pv(process value)


float pid (float sp, float pv)
{

err_old = err;
err = sp - pv;

// note
P_err = err;
I_err += err_old;
D_err = err - err_old;

return 0.1*P_err + 0.3*I_err + 0.02*D_err;

}

now here he has set the values "0.1 ,0.3, 0.02" as constants .
this doesnt seem to be a auto tune code. what changes will i have to make and how.

 

[JoKKeR]

Re: auto tune pid?

I like to see a well formed answer for this question, as im also interested.

The auto-tune means. amplification constants in pid algorithm are change during work flow of device... Changed based on ? ? ? on... time and error, because tease 2 parameters are reason why ur using algorithm. tease parameters are changes in a way that quality of your regulated process improves.

How ?
if you got stable and pattern like process then you can do it easy way (better of ... result) if no then it will be long and complicated task.

 

[sg962]

Re: auto tune pid?

I would recommend you modify PID formula a bit:
PID = GainP * ( actual error + SUM(previous errors) / Ti + (actual error - last error) / Td
where Ti - integral time, Td - derivative time.
In this case all terms will be normalized.
PID tuning process is a complicated procedure based on controlled object behavior.
Brief explanation of PID params influence on PID behavior you can find in this doc: https://termexlab.ru/upload/filearchive/363_M03 manual.pdf.
See chapter "PID controller parameters".

 

[FvM]

auto tune pid?

Td must be a factor, not a divisor in your PID formula.

 

[sg962]

Re: auto tune pid?

It's up to you, tastes differ. You may use a factor Kd instead of Td in the case if Kp=1/Td, but Td has a physical sense as a time constant of differential stage in PID controller. I wrote that multiplication of integral & diff terms by proportional gain makes PID more robust and easier tuned.

 

[FvM]

Re: auto tune pid?

It's up to you, tastes differ. You may use a factor Kd instead.

I don't mean changing to a factor Kd. If Td has the physical sense of a time constant, you have to correct the differential term to

(actual error - last error) *Td/Ts

Ts is the controllers sampling time interval. In other words, decreasing the Td time constant decreases the "D" gain, while decreasing
Ti increases "I" gain. Check with your actual PID code or a text book.

 

[sg962]

Re: auto tune pid?

Oh, sorry, you are right for sure! I've been very negligent. Thanx.

 

[tarzan2404]

auto tune pid?

who has ever tried GA for tuning pid parameters?

 

[deperkin]

Re: auto tune pid?

There are better control algorithms for a tuning mechanism than PID.

I would look into programming an estimator.

PID is a good cheap controller, but if you dont understand fully the system it is not the best to use.

I built a full-state feedback system awhile back and it worked much better then PID.

Also you could take a look at some of Kalmans work on the topic.

Are you modeling the system first of just 'over-engineering' it?