SLC 5/04 question for Peter Natchway, and thanks to all!

This question is relevant to the 'search' function capability.
After reading this post, which took a while, I've decided to ask a question, which is a SLC 5/04 PID question. I've never asked it before, as I figured it would not get answers, as many would consider it 'just another PID question without research'.

As far as the topic of search function, most of the answers I've ever needed, I had seen here in the past. If there was a way that I could save post here on the site, that would be awesome. But some times there are specific questions, about common topics, such as PID's, which are best solved by posting. Take care to not pass by a post, just because it is a common topic.

Here is the PID question. I apologize that I cannot post the code:

The PID is controlling speed, to maintain steady web tension.
After a ton of research, every thing I read said to put my I and D gains to 0.0, and adjust P gain, until oscillation, and set it at half of that value. The problem is, when I put the I to 0.0, and did a bumpless transfer, with my CV at 27%, from manual to auto, my CV dropped to 0% immediately, and then responded. In that time, it made one hell of a mess.

My bumpless transfer works now, and the PID controls, as by just educatedly guessing, I came up with P=5, I=.05, D=0.0.

My question #1 is, I believe with the 5/04, I is faster the lower the number, so for this test to find gains, I should've had a high number in the I gain?

My question #2 is, even with an I gain of 0.0, if my setpoint is equal to my process variable, why does the control variable drop to 0%, and then try to respond, when I switch from manual to auto?

Like I said, sometimes there are specific questions, to common topics, which require a post.

Update: Actually ended up with a P =.2, I =.1, D = 0, which gives good response, with no overshoot. The gains shown are with the RG not set. With the RG bit set, it would be by 2,1,0.

As I am on my own now, I only do automation work occasionally, and have no one I can go to with questions. This site is my 'go to' person. Without it, I could never have made this project work.

Thanks to all!

A quick answer to a well written question. Tidbits taken from refernce manual 1747-RM001-EN-P and the Knowledgebase.

PID Output = KC*[(E)+1/TI*∫(E)dt+TD*D(PV)/df]+Feed Forward/Bias

Reset Term Ti (word 4) is the Integral gain, ranging from 0 to 3276.7
(when RG = 0) minutes per repeat. Set the reset time equal to the natural period.

What can prevent a bumpless transfer from working?

1. Disabling the PID rung. The PID should be part of a continous rung of logic.
2. The Reset Term (1/Ti) is changed to 0 or to a large number. Since the Ti is inverse of one, Any larger number will make it approach zero.
3. Writing to the output image table of the analog output card (ex. O:1.0), Always write to the Control Variable when you are in Manual mode, ex. N10:29

For the bumpless transfer to work, the PID instruction has to back-calculate a value for the Integral Sum when the mode switches from manual to auto. It can not do this when the integral gain is disabled (equals 0) or too small (large denominator).

Originally posted by Brownhat:

It can not do this when the integral gain is disabled (equals 0) or too small (large denominator).

Click to expand...

I think you are half right here. I think you are right that if the integral is disabled (gain of 0) bumpless transfer will not work correctly. However, don't take the symbolic equation from the book too literally. It really isn't an "integral", it is a digital sum. The way the integral term is probably implemented is:

Int_Sum = Int_Sum + (1/T_i * Error * ScanTime)

The will be limit wrappers around it and it may do something to get a better idea of what the mean value is during the integration period. But the basic idea is ther.

So as long as you have any value for the integral time you should be fine for bumpless transfer.

Keith

I can see you guys figured this out before I woke up.

I am on the west coast. I can see the early birds or on the east coast or mid US.

When going from manual to auto mode I use the current manual control signal and copy it into the INT_SUM in Keith's equation. I also set the SP to the PV so there is no error. This should be bumpless. Then I ramp the SP to the desired value.

If the SP can't be set to the PV then some calculations are required. Since
MANUAL_OUTPUT = INT_SUM+K*(SP-PV)
The INT_SUM must be set to
INT_SUM=MANUAL_OUTPUT-K*(SP-PV)
Now when you go from manual to auto the INT_SUM + proportional term will equal the MANUAL_OUTPUT and there is little or no change in output and the system should close on the SP nicely if it is tuned correctly.