M580 Autotuning PID - tutorial or guide

[Kevingalvez12]

Good Afternoon Community, 

I will like to know your experience with the autotuning function of the PID of M580. I have used autotuning with the M221 and I got mixed results. I have test order autotuning methods for example in siemens 1200 with great results. It will be great if you people have a tutorial, guide or recommendations on how to achieve it in the M580, so I can reach it´s full potential on this PLC.

Thank you very much.  

[john_m]

HI

I am not an expert on modelicon. But I know that the autotuning with PID is  complicated and not always reliable.  For examples It does not work on all the types of process and there are a lot of drawbacks as the difficulty to control the overshoot, to take into account to large dead time, saturation ( speed) , to get a good load disturbance rejection, etc  

But there is another solution is to use the new controller Relaxc. ( in the slide you have all the information link)  https://www.appedge.com/pdffile/relaxc_expertcontrol.pdf

The practical tuning guide of relaxc( see pdf)  shows how with only  five graphics rules of  auto tuning it is possible to tune  all ( almost) the types of process.  It is just necessary to identify the dead or gap time and the max speed. For that you can use the more suitable excitation for your system. ( step response, relay feedback, noise, armax  or identification of the transfer function ) to measure these 2 parameters and relaxc we will  give  us  the more secure and efficient control for our  process with the best load disturbance rejection.

 This methodology is only possible with relaxc, because its mathematical structure  allows  it  because   Relaxc is  “the plant”.  As the parameter of relaxc are a physical sense, a user can checks them without calculation and adjust directly them too.  

You can test and verify the behavior of relaxc with the free trail version of relaxc.  If you use relaxc I am thinking that many people will be interesting by your results and me too.

The autotuning of relaxc is independent of the modelicon device

https://shop.exchange.se.com/en-US/apps/55444/relaxc---critical-process-loop-control---trial-version

I can help you to use relaxc if you want or if you have some difficulties to download relaxc.  

J.M

ps Sorry for english

[Kevingalvez12]

Hi John,

I´m not a native english speaking person too so don´t worry your english is great. 

So reading a bit of the relaxc is basically a paid library for control expert or unity, but I´m not sure if the relaxc works like an autotuning or you have to do a lot of math because I saw like 5 steps of graphs method. The relaxc seems to work for 5 hours in the description, I´m not sure but does this mean that the relaxc calculates continously or is just my impression? When I released a little of the projects, I will dedicate some time for testing the relaxc. Thank you John. 

[john_m]

Hi Kevingalvez

For the moment relaxc is free but limited. ie it works five hours and after it is necessary to transfert your plc in the modicon via expert control and relaxc works again during five hours and so one. It is a evaluation version or for the students.

In the literature, we will see that auto tuning works generally /exclusively on the behavior of first/second/or more order +delay ( generally small). For the real world, the relaxc graphics rules for second order and more cover "70%" of the cases. Relaxc allows an extension at all process configurations and there are 5 fives rules for covering all the processes world.

(A) Effectively with relaxc it is necessary that you measure yourself the delay and the speed of your process and that you compute the coeff ks. But it is not to hard to identify automatically these 2 parameters. The speed of your process is given by the output Speed of relaxc. You identify and compute theses coefficients one time. If you want to calibrate your process again it is necessary to redo the measure. But with the simple structure of relaxc and the simple autotuning rules it is possible to identify continuously these coefficients where in the worst case to prevent the engineer that his process changes.

Now that you have relaxc, you can may be use it to test or to verify the quality/ performance of your auto tuning method that you use.

Regarding siemens device autotuning there is a paper ( 2017) Implementation of PID autotuning procedure in PLC controller (Marcin Daniun) which use the method of the moment+amigo plus “secret algorithm” which confirms that I say. And if you compare the amigo method to relaxc, you see that relaxc is will see to easy and efficient to use, because the coefficient of amigo are not physical and depend precisely of the mathematical structure of your process( no degree of freedom) . And if you apply the point (A) your autotuning will be better than siemens.

Keep me in touch for relaxc. and thank for your downloading.

john

Ps: the mathematic of relaxc are complicated to elaborate relaxc but not to use it. it is my point of view ;-).

[john_m]

Hi,

For a better understanding of Relaxc,  here are the slides

john

[john_m]

Hi

It exists also a trial version under simulink of relaxc ( Same structure that the PLC ). If you want to try it download

https://www.researchgate.net/publication/340788868_Relaxc_under_simulink_v1 

and ask me key1 for one month

thank

[pradiptasetio]

Hi Kevingalvez12,

Our team made a documentation of PI control simulation using Autotune function block. Unfortunately, it's not in English. From there, we've also successfully applied autotuning of PI control a few times in projects. However, we don't have comparison with other brand PLC, so I can't tell how great it is, but I can say it's useful. Let me share some key Autotuning parts of the simulation.

Below is how the FBD looks like:

The key here are the PARA input in both AUTOTUNE and PI_B blocks. In Autotune PARA, there are 3 variables: step_ampl, tmax, and perf. For step_ampl and tmax values in actual application, you can refer to below table (also can be found in the Help):

In order to set this values, you will need to create a DDT variable with Type: Para_AUTOTUNE. In this example, the variable name is "para_autotune"

Autotune PARA_C output is linked to PI_B PARA input. The variable is also a DDT: Para_PI_B. Or if you're using PIDFF block, the type is Para_PIDFF. The variable name in this example is "temp_parameters". Except Kp, Ti, (Td and Kd), other parameters must be manually set. Leave it blank for Kp, Ti, (Td and Kd).

And here is the simulation result:

note: Autotune was started few seconds after setpoint was set.

Hope this helps.

Cheers,

Pradipta

[john_m]

Hi

Interesting and good work.

If you have the  time, you can  compare the effectiveness of your tuning  with  relaxc trial version. ( compare time rise and time disturbance rejection). Your grapichs show that it is possible  to apply first order graphic rule of relaxc to get the coefficients of relaxc.  You can adapt your autotuning block at  these rules ( just to measure the speed and to set time response that you want by tg). I am interested  by the comparison.

best regards

ps: If you have not apure delay in you process put the relaxc parameter tre at 3*the sample time

[Kevingalvez12]

Thank you very much, that's a great document. 

[Kevingalvez12]

Hi pradiptasetio, 

It doesn't matter if the documentation isn't in english is always great to take a look regarding the language and thanks for explaining the use of the autotuning with the PI control it's really a great information. 

Cheers.