PID contollers

A PID controller is a device that compares a measurement to a setpoint and produces a control output signal that drives something that alters the condition being measured, such that the measurement is forced toward the setpoint.

I have written a non mathematical tutorial that explains the function of the P (proportional), I (integral) and D (derivative) terms generated in the controller, and how to tune them. It will give you a feel for what a PID controller does.

Good on you! :-)

We need more webpages like this, that we can refer people to.

Thanks! Rich

A controller that controls an output based on a feedback input using Proportion, Integral or Derivative (PID)control response

You caught me. I am an expert wibbler. They call me in after the analytical types have tried and failed. The real world is very messy and nonlinear, and wibbling is sometimes the only way to arrive at a reasonable solution in a finite amount of time. The method I describe is orderly and goal oriented wibbling based on years of practical experience with enough real world systems to fill a large industrial museum. I do not claim it is the last word on the subject, but an intuitive place to start in the process of understanding loop tuning.

The tutorial is a summary of my experience refined through several courses for operators and chemists I taught at a DuPont plant on the basics of PID control, which was used in conjunction with an elaborate process simulator I wrote for the Honeywell 3000 DCS system for them to play with.

I'm an opinionated drunkard but.....

When you see the term PID controller you would do well to steer clear of the general area where it was mentioned.

PID is an excuse for people to wibble about with things until something works.

As you can see from Mr Popelish's explanation.... the recommendation is that you wibble about until it works.

Having wibbled about with it you explain to others about your wibbles in terms of P, I and D.

Since they don't know what you are talking about and you don't either then you become an expert.

Obviously, since it is working, you must have been right.

Next Tuesday it blows up.... so, since you are the expert, they ask you back to wibble about some more.

If you are an experienced wibbler then your initial wibbles might be close and you might make a sensible wibble but you are still a wibbler.

Ask Tim Wescott, he knows all about it.

DNA

the

that

then

back

close

One day I'd like to put my bum where my gob is. Next time your analyticals fail give me a shout and we'll see just how shit I really am.

DNA

I am sure you would enjoy the educational experience.

I have been called in to "fix" the tuning on a process that would cost a million dollars a day if I shut it down. It is operating chaotically (not cycling sinusoidally, but as if it were a strange attractor), interacting with a half dozen or so other control loops in various ways, and giving everyone heartburn. They tell me that I can alter the output value, manually only plus or minus 2% (less than it is swinging in automatic or cascade mode), but they really don't trust me to put the loop in manual mode, at all. Essentially all I can do is watch it (and the other loops) work. And I can change the P, I and D constants, as long as I am willing to take the blame for any worsening performance, and especially any out of limits operation or shut downs.

Plug that into your analysis. ;-)

The thing that is neat about the graphical method I describe in my tutorial, is that it gives you some simple rules to determine (in a "you're getting warmer" or "you're getting colder" fashion) which term is most likely to be causing the trouble and which way it needs to change to back away, gracefully from that trouble, without having to pass through any mine fields. And it even works, most of the time, too. :-)