Interesting article on temperature PID auto tuning

[Peter Nachtwey]

http://www.controleng.com/article/CA6626810.html?nid=2361&rid=1263151521

You can see that not all auto tuning is created equal. What gets me is that these tests were conducted under lab conditions and still some of the 'auto tuners' had a lot of difficulty. This should be a slam dunk.

The auto tuning on-the-fly is a lot more tricky.

 

[Tom Jenkins]

Very interesting article, Peter. Thank you for this.

The article justifies my abandonment of PID years ago. I deal with process that have both time cnstants and dead time significantly greater than those tested in the article. To make it harder, the primary process I'm controlling is non-linear, not first order.

PID is a wonderful thing if both the application is right and the tunining is correct. This article illustrates that it can be over used. I was particularly interested by the first sentence - 30% are in manual, and 65% of those in auto would be better off in manual.

 

[jamesau]

This article doesn't seem to be of much value. It seems the author can create a nonlinear plant model with very large deadtime (relative to dominant time constant) to show how poorly an autotuner can work. I'm not sure anybody could duplicate his experiments given the information provided(how is "In these tests separate heat plant and cool plants are used, and each can have a different time constant" entered into the model).

Most loops that are run in manual are either
a) poorly tuned to begin with
b) use a poor choice of control structure (what input controls what output)
c) reflect the fact that a significant process change has occured which renders the loop inoperable. This includes a reconfiguration of the process, among other things.
d) an artifact of less-than-great process design.

This isn't a defense of auto-tuners; I've never used one.

Control Engineering has its place, but I'm generally quite wary of non-refereed technical articles which appear in trade-journals

 

[Peter Nachtwey]

Very interesting article, Peter. Thank you for this.

The article justifies my abandonment of PID years ago. I deal with process that have both time cnstants and dead time significantly greater than those tested in the article. To make it harder, the primary process I'm controlling is non-linear, not first order.

Your systems would be difficult or impossible without the model based control but others manage some how. However, just because a PID doesn't work for you does not mean it will not work for someone else.

PID is a wonderful thing if both the application is right and the tunining is correct. This article illustrates that it can be over used.

Auto tuning or PID? Assuming PID, I know there are some systems like Tom's that aren't impossible but extremely difficult.

I was particularly interested by the first sentence - 30% are in manual, and 65% of those in auto would be better off in manual.

This is something I can't understand. That would never be acceptable in motion control.

What I got from this article is that people are tuning on-the-fly and have adaptive PIDs. That impresses me even though temperature controllers don't need to run as quickly as motion control PIDs to they have more time to do the calculations. I also learned that many companies have garbage auto tuning algorithms. It is almost like the marketing guys wanted to say they had auto tuning too so asked the engineers to read their college text books and kludge something together.

Most loops that are run in manual are either
a) poorly tuned to begin with
b) use a poor choice of control structure (what input controls what output)
c) reflect the fact that a significant process change has occured which renders the loop inoperable. This includes a reconfiguration of the process, among other things.
d) an artifact of less-than-great process design.

LOL, I agree. That is kind of like my statement about a system is kludged, not designed, unless there is a transfer function stamped on it. If there is a transfer function then one can calculate the gains.

Tom has a transfer function or model of sorts.

 

[Binaural]

"It is surprising that more than 30% of all factory PID control loops are operating in manual rather than automatic mode. To make matters worse, of those loops operating in automatic mode, 65% produce less variance when in manual. "

This sort of blank statement of numbers without any accompanying link or at least reasoning makes me extremely suspicious of his data. How could he know this sort of information? The article is interesting otherwise, but my credibility detection sensors were sounding the alarm after the first paragraph.

 

[danw]

re: # of loops in manual vs automatic

There have been studies over the years by organizations like ARCX, and the numbers filter into generally accepted cultural vocabulary.

For example, huge numbers of liquid tanks are still filled manually.

When I started in this 25 years ago, the number of loops in manual was significantly higher than 30%.

A lot of the questions posed on this forum are not for new installations, but retrofit control on existing devices.

Dan