Why industry choose PI rather than PID or PD?

ivanlch · Mar 18, 2013

Hi,

Why does the industry prefer to use PI rather than PID or PD? Is it because of the reasons below? Please let me know if my theory / finding is wrong.

- Easier to tune
- Easier to implement
- Save time
- D is difficult to find
- D slow down performance
- D creates slight oscillations.

Bit_Bucket_07 · Mar 19, 2013

I don't know which industry you work in, but it must involve processes that react relatively quickly to CV changes.

You might find this article informative. The author eschews getting bogged down in the calculus and instead speaks in terms of real-world, intuitive observation of processes under closed loop control.

danw · Mar 19, 2013

Proportional-only (P) inherently has 'droop', an offset from setpoint. Proportional-only can provide stable control, but with an offset from setpoint. Mechanical, proportional-only regulators exhibit this phenomena.

The integral (I) factor accounts for the droop (offset error from setpoint, over time) and corrects for it.

So when P is combined with I the combination of the two provides somewhat 'passable' control for a lot of applications.

"Passable" means fewer tech support calls.

Highland Controls · Mar 19, 2013

It all depends on how dynamic your system is. PI works well with processes that aren't very dynamic. Basic velocity control is one example. PID is needed for dynamic processes, like positioning.

Tom Jenkins · Mar 19, 2013

This begs the question "Why do so many engineers use P-I-D in any combination at all?"

PID is not a law of physics, and it is far from the only algorithm for analog or process control. It is 70 year old technology, developed as proportional, reset, and rate (so-called three mode control) back in the days when pneumatic control was the only game in town for analog devices. A small industry has grown up around tuning PID, and includes articles, books, software, proprietary "self-tuning" algorithms, and consultants. This should alert the thinking engineer that there are serious problems with the application of the PID algorithm.

And yet it is the first, and often only, choice for a great many engineers. I'm not sure if this is because of the neat equations, or the fact that they were exposed to it in school, or the inclusion of PID blocks in most PLC software. However, in my opinion, PID should be the last choice, not the first!

That isn't to say that PID doesn't work well for a great many applications, or that in the hands of an expert it can't be tuned to provide precise control. What I am saying is that simpler techniques more readily tuned by the average Joe should be explored first, and PID resorted to only if the simple techniques are proven inadequate. A great many posts on this site are for processes that would be served just as well, or better, by a simple deadband controller.

Don't assume that process control = PID control!

(Rant off)

JHarbin · Mar 19, 2013

Tom Jenkins said:
This begs the question "Why do so many engineers use P-I-D in any combination at all?"

PID is not a law of physics, and it is far from the only algorithm for analog or process control. It is 70 year old technology, developed as proportional, reset, and rate (so-called three mode control) back in the days when pneumatic control was the only game in town for analog devices. A small industry has grown up around tuning PID, and includes articles, books, software, proprietary "self-tuning" algorithms, and consultants. This should alert the thinking engineer that there are serious problems with the application of the PID algorithm.

And yet it is the first, and often only, choice for a great many engineers. I'm not sure if this is because of the neat equations, or the fact that they were exposed to it in school, or the inclusion of PID blocks in most PLC software. However, in my opinion, PID should be the last choice, not the first!

That isn't to say that PID doesn't work well for a great many applications, or that in the hands of an expert it can't be tuned to provide precise control. What I am saying is that simpler techniques more readily tuned by the average Joe should be explored first, and PID resorted to only if the simple techniques are proven inadequate. A great many posts on this site are for processes that would be served just as well, or better, by a simple deadband controller.

Don't assume that process control = PID control!

(Rant off)

I assume we are both old timers, but I wanted to provide an opposing view. I've been working on controls starting with Pneumatics, P only analog controls, P with Manual Reset (analog), and PID Digital controls.

I have seen your other rants against PID and I respect your opinion (since you are an old-timer) but I have thousands of PID control loops in the field (admittedly most PI) that are controlling quite well. As such, I see no need to look for an alternative to PID.

Granted there may be alternatives to PID that will work better in particular situations, but considering that PID works *very* well in 99.5% or more of the process applications I have done over the last 30 years, I don't think there is any need to abandon the technology.

Just a friendly opposing view...

Tom Jenkins · Mar 19, 2013

JHarbin said:
Just a friendly opposing view...

Understood - my view is certainly the minority one. You, being a fellow old ****, would qualify as one of the "experts" I made a point of mentioning.

I don't want to kill PID, and I have no financial interest in its use or non-use. I just get very frustrated with all of the questions posted here, and problem systems in the field, where the designer or programmer can't even concieve of an alternative approach.

In the interest of full disclosure, most of my work is in very non-linear processes with significant dead time and operators with low automation skill levels. This undoubtedly colors my experience!

Brijm · Mar 19, 2013

Tom In the interest of full disclosure said:
I will give you some support Tom, I come from an instrumentation background, so I am not afraid of tuning a PID. But they are not always the best answer. If the process is difficult to tune, and a simpler control will get the job done, then why complicate things.

ivanlch · Mar 19, 2013

Highland Controls said:
It all depends on how dynamic your system is. PI works well with processes that aren't very dynamic. Basic velocity control is one example. PID is needed for dynamic processes, like positioning.

Hmm, I am trying to control and maintain the water level in a tank. The water flow is controlled by a servo motor. I am using Ziegler Nichols Tuning Method. I guess PI would be the best option for me?

Manglemender · Mar 20, 2013

If you could predict what the flow rate needs to be based on say a flow meter then you'd be more than half way there.

How about a proportional only system with a minimum gain? I use this quite a lot for rough postion control. How accurate does the tank level need to be maintained? With a proportional only system then you will end up with a level error that is proportional to the outflow rate and the P gain; if you could predict the outflow rate and generate a pump speed then that's probably all you'll need.

Nick

P.S. I'd be very surprised if D was required in your system, especially if you predict the pump speed fairly well.

ivanlch · Mar 20, 2013

Manglemender said:
If you could predict what the flow rate needs to be based on say a flow meter then you'd be more than half way there.

How about a proportional only system with a minimum gain? I use this quite a lot for rough postion control. How accurate does the tank level need to be maintained? With a proportional only system then you will end up with a level error that is proportional to the outflow rate and the P gain; if you could predict the outflow rate and generate a pump speed then that's probably all you'll need.

Nick

P.S. I'd be very surprised if D was required in your system, especially if you predict the pump speed fairly well.

It needs to be set dead on 50/100. I tried P only. It either overshoot/ undershoot. At time it oscillates

Peter Nachtwey · Mar 20, 2013

ivanlch said:
Hi,

Why does the industry prefer to use PI rather than PID or PD? Is it because of the reasons below? Please let me know if my theory / finding is wrong.

- Easier to tune
- Easier to implement
- Save time
- D is difficult to find
- D slow down performance
- D creates slight oscillations.

None of the above.

Bit_Bucket_07 said:
I don't know which industry you work in, but it must involve processes that react relatively quickly to CV changes.

You might find this article informative. The author eschews getting bogged down in the calculus and instead speaks in terms of real-world, intuitive observation of processes under closed loop control.

I am sure the author takes Dr Rice's comments out of context. Dr Rice is Dr Doug Cooper's, the control guru, partner in making Control Station which is auto tuning software. Dr Rice can calculate controller gains. I know Dr Rice can do the auto tuning and calculate controller gains. I would be very upset if I were a student paying a lot of money for tuition and all I was taught was to tweak gains or use ZN. I have seen the material the Dr Cooper and Dr Rice use to teach students.

Highland Controls said:
It all depends on how dynamic your system is. PI works well with processes that aren't very dynamic. Basic velocity control is one example. PID is needed for dynamic processes, like positioning.

No, it depends on the number of poles in the plant. If you want to place all the poles then one gain is required for every pole. Simple systems can be tuned with a PI controller. Under damped position systems should use a PID with a second derivative gain.

Tom Jenkins said:
Tom Jenkins said:

This begs the question "Why do so many engineers use P-I-D in any combination at all?"

Click to expand...

There are other alternatives. I like Sliding Mode Control for some applications.
One can control systems precisely with either, but you need to know how.

The problem is that few make the effort to understand PID control. There are only a few on this forum that have made that effort. Rytko, Pandianni and myself.

Tom, I defy you to control a mass on a spring. Pandiani and I went through this exercise a few years back.

Brijm said:

I will give you some support Tom, I come from an instrumentation background, so I am not afraid of tuning a PID. But they are not always the best answer. If the process is difficult to tune, and a simpler control will get the job done, then why complicate things.

Click to expand...

Some times a simpler control will get the job done. Do you know why? It can be proven mathematically that sometimes a PI is all that is needed. I provide an example below.

ivanlch said:

Hmm, I am trying to control and maintain the water level in a tank. The water flow is controlled by a servo motor. I am using Ziegler Nichols Tuning Method. I guess PI would be the best option for me?

Click to expand...

A PI will work just fine but again, do you or anybody else know why? On LinkedIn there were a couple of posts where the posters threatened to mark other posts that suggested ZN as inappropriate. ZN should be abolished. A few weeks ago I posted a link to a pdf file where the pump pumped water into a first tank that flowed into a second tank and the level in the second tank had to be controlled. No one made the effort to figure out what I did.

Mangelmender is right about getting by with just a P gain if you can tolerate some error. Does anybody know why? The D gain isn't required. Does anybody know why?

Here is a link to how to do simple tank level control.
http://deltamotion.com/peter/Mathcad/TwoTanks/Mathcad - t0p1 pi tc - Alin's tank.pdf
Only P and I gain are required. Notice that there is a difference in the response between having the P gain work on the error between the SP and PV or act on changes in the PV. Why?

I don't make a big deal about this most of the time but it bothers me a lot when people suggest using ZN. It bothers me a lot when people make suggestions that may or may not work because they don't understand why a PID works.

When I retire who is going to answer these questions?

I would like to save Tom from the dark side.

Why industry choose PI rather than PID or PD?

ivanlch

Member

Bit_Bucket_07

Member

danw

Lifetime Supporting Member

Highland Controls

Member

Tom Jenkins

Lifetime Supporting Member

JHarbin

Member

Tom Jenkins

Lifetime Supporting Member

Brijm

Lifetime Supporting Member

ivanlch

Member

Manglemender

Member

ivanlch

Member

Peter Nachtwey

Member

Similar Topics