Why industry choose PI rather than PID or PD?

[Heartnet 0]

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.

[panic mode 246]

i don't know about one being the favourite, it is application driven and there is plenty of PID controllers and PID instructions in PLCs.

[Sleepy Wombat 27]

Like every other question you seem to be posing on this forum lately, they appear straight out of a text book / class teacher notebook.... You are in Oz, are you going to TAFE or UNI... if so .... don;t you think that the learning process would require more effort on your part other than posting a question and simply assuming that forum participants would simply pass on their hard earned knowledge. We all volunteer our help here, and will generally help where we can, espeically if the person has shown some effort in trying to understand the problem and made a genuine effort... you have more tools available to you to learn about everything then most of us ever did......if you don;t put in the effort then you will start getting the response with another acronym that you can look up.. RTFM...

[The Turkey Slayer 5]

The PID loop is a calculus equation, and the vast majority of people who tune loops don't have much of a working knowledge of calculus (me included). I remember when I first started getting into programming and I asked someone how a PID loop worked. His answer was "it stands for Proportional, Integral, Derivative," which of course was no help at all. Unfortunately, I've run into few people who truly understand how it works and pretty much just adopt a "throw numbers in it until it works" approach (again, me included). And people do that because, generally, it will eventually work. The P and I terms are relatively easy to understand. Proportional is how much to respond based on how far off target you are. Integral is how much to add to that response based on long you've been off target. Make P go up and the amount the CV changes in response to error goes up. Make I go up and CV ramps up faster over time. D is more complicated in that it tries to compensate for where you're going to be, and (I think) uses the derivative of your PV slope at given intervals to do this. Tuners generally don't like messing with the Derivative term because: A) It can make the loop behave in "unpredictable" ways, and B) In many cases, just tweaking the P and I terms can get the loop to a state that's "good enough" for the application Generally, I don't mess with the D term unless there's a critical component to the control that needs to strictly avoid under or over-shoot. Edited 4 Apr 2013 by The Turkey Slayer

[BobB 58]

I work with generators a lot - diesel only requires PI - gas engines require PID with normally a lot of D - has to do with response of the gas engines. PID is quite often overkill. Depends on the process. Edited 5 Apr 2013 by BobB

[Peter Nachtwey 15]

The statements above are correct. It depends on the process and it is a calculus problem BUT one doesn't need to know calculus if you know Laplace transforms. Laplace transforms are a way of doing calculus and differential equations using algebra. Basically there are two general type of processes. 1. Type 0 or non-integrating process. These are processes where the process value will return to steady state or ambient if control power is turned off. These are processes like temperature control or speed control. These are usually simpler to control and don't need a derivative gain. 2. Type 1 or integrating processes. These are processes that integrate the control output. Tank level control integrates flow and position control integrates velocity. In theory these processes don't require an integrator but do require a derivative gain. The reality is that these processes are never perfect so an integrator gain is required along with the derivative gain. There is a general rule. There should be one controller gain for every open loop pole in the process. The integrator gains doesn't count because it comes with its own pole. I know I just lost everyone with this statement but it is true and if you want know the truth before you die you should look into this. ( add it to your bucket list ). If you want to learn more you should look up topics such a pole placement or lambda tuning. I deal with calculus and differential equations just about every day so the math is no strain on my brain.