I don't mean to offend, but there is no way you would be able to decipher this thing if I posted it on this forum. you cannot see the physical system, there is so much about it that is spaghetti and (mistakes), they have tried to use multiple PIDs to compensate for different conditions that occur, It is in conjunction with a 43 screen Panel view project...If you could physically be here and see it, and have access to everything , it would take you a couple of days to figure out what inputs and outputs are for what, and where, and how many places the variables are being modified.... so there is no point in suggesting that I post it... I sincerely apologize if I am offensive , but this is not a simple project.... and it was not very well done or documented.
This site is the best resource anyone could have and I really appreciate all your help, but if you would be so kind as to answer the questions I ask (or don't if you don't want to... I know I am not entitled to anything from any of you and I am appreciative of anyone who is willing to help).
The questions I ask are what I want to know. I can't post the trends or the program files ... I don't think either one would tell you anything. I have been looking at them for a long time.
I truly do not mean to be offensive or un appreciative I do appreciate everything that everyone is saying, thank you thank you thank you
and I mean that sincerely...but I am the one here with this mess. I want to know some things about the allen bradley PID operation in a Micrologix 1500 PLC. The help and the documentation is not particularly helpful IMO. If it were, I would be reading it right now. I have and I will read it again... In conjunction with Mr. Beauford's fine articles, some of it makes more sense than it did before.
I come here because you are the best help I know of. Being demanding of all this info from me is not what I need. I can't supply it, and it wouldn't do any good if I did. Anyone who has worked with Factory talk and multiple screens and being unfamiliar with, and blind to the physical system, should understand that that is just not practically useful.
I will be back when I formulate my questions and answers to the questions some of you asked, as this post is already too long.... I hope I didn't offend everyone and someone will still be willing to help. I just want to know what I want to know. I don't expect you to solve the whole problem for me.
again thank you.
Try not to take this the wrong way, but imagine you went to the doctor because of a nagging back pain. You tell the doctor that you have no idea what is causing the pain, and that nothing you do seems to give you any relief. You are out of ideas and need some help! So the doctor tells you, "Well, we don't have enough information to go on here. I'm going to have to get an X-ray before I can begin to help you out." To which you respond, "Sorry doc, my back is a mess, and looking at it, won't help you. Besides, I've already tried." Seem a little off to you?
It's ok if you are not able to post program files or trends, you are not obligated to do so, but you shouldn't take the requests personally. Ron has posted some good info for you to answer your specific question about % of scale. But to be able to begin to suggest what you need for a faster response, or to suggest gains for the system, someone would have to have these things. At least a better description of what the system is, and what it is trying to control may give some hint as to what may be the problem. It sounds like it may be a combination of a lot of things! It sounds like a lot of people have tried to "throw PIDs" at a system that was poorly designed in the first place.
Are you able to try running the system with the CV in manual, and observe PV response? If you can get the system stable in manual control, then "bump" the CV up a certain amount, then observe how much the PV changes, then you can get a rudimentary estimate of the process gain. For "moderate" tuning purposes, you can start with a P (Proportional) gain of (1/Kp) * ((PVmax-PVmin)/100) (Kp = Process Gain, or PV/CV). Now keep in mind that this is a VERY simplified version of a scientific way to arrive at a P gain, but it can get you started.
Now let's look at the I (Integral) gain. For a micrologix plc, the typical unit for the I gain is min/repeat. For practical purposes, this means how many minutes you want the PID instruction to take to "repeat" or double the effect of the P gains contribution to the CV if there is no change in error. So if you are using a typical reverse-acting error equation (E=SP-PV) and you are 20 degrees below SP, then your error = 20. If you have a P gain of 2.5, then the P gain will add 2.5*20 = 50% to your current CV. If you have an I gain of 2.0 min/repeat, then after 2 mins, if your error has remained at 20 degrees, then your P gain will still be contributing 50%, and your I gain will also be contributing 50% so you will be pushing full scale 100% CV.
Now knowing what to set your I gain to requires some empirical testing. This is where a trend comes in extremely handy. Remember the CV "bump" test that gave us Process Gain? Well we would also use this test to give us
Process Time, which is basically the time (in minutes) that it takes the PV to complete 63.2% of the delta from the moment it begins to change. For simple tuning, you will set your I gain = Process Time. Bear in mind, none of this takes into account
Deadtime, which from the trend, would be the amount of time elapsed between a CV bump and the moment that the PV begins to change. The greater your deadtime, the more the simple tuning described above falls apart, and the more complicated the gain equations become to compensate. For a more detailed explanation of this tuning process, check out
www.controlguru.com.
All the tuning knowledge in the world won't help you, though, if your mechanical design is flawed. I wish you all the luck in the world with your project, and be sure to post back if you have more questions, or if you make any headway, let us know how it goes.
Cheers,
Dustin
