Uncoiler strip guiding with PLC – application issue

[jim767]

Are you varying the update time as a function of line speed?
If so how? QUOTE]

I'm using a CPT to send (1/LineSpeedFPM)*60 to the PID.UPD
also using a CPT to send (1/LineSpeedFPM)*6000 to the FIFOtimer.pre - Fifo array has 10 elements - sensor is 10 foot downstream.

this should get me a PID update every foot of travel and load and unload the Fifo every foot. At this point the parameters I'm changing will be the .bias gain as discussed earlier, the integral time which is expressed in min per repeat, the PID blocks proportional gain. The min/repeat setting and what I should use is throwing me for a loop at this time as one of many things but I am reading the control guru site trying to understand the statement quote" Best practice, Sample time should be 10 times per process time constant or faster (T< or = 0.1 Tp" Does this go with the integral rate or how sampling for the FIFO.

 

[kamenges]

If you look at the way webs respond to cross-machine forces, I don't think this is so much a deadtime thing as it is a varying time constant thing. As soon as you put a cross-machine stress into the web the web will start to move in a direction to take that stress out. How long it takes to get the stress substantively removed (approximately 5 web spans) is based on the web span and the speed.


Keith

 

[Peter Nachtwey]

Are you varying the update time as a function of line speed?
If so how? QUOTE]

I'm using a CPT to send (1/LineSpeedFPM)*60 to the PID.UPD

Does the PID.UPD change the interval at which the PID is running or just the value the PID uses to compute the true gains? Are you executing the PID at constant time periods or as a function of distance. You have not been clear about this and I don't like playing 20 questions. Is there any reason why you can't trigger the PID for every foot of travel?

The integrator term is

COi=Kc*T/Ti*(SP-PV) substituting 1/LineSpeedFPM*60 for T.

COi=Kc*(LineSpeedFPM*60*Ti)*(SP-PV)

As the speed gets slower you say you changing the update time. Is that the time T in the equation above? If so what you are doing IS NOT RIGHT. As the line speed slows down your equation shows that PID.UPD gets larger. This would cause T to get larger if I interpret what you are saying right. This would make the control output due to the integrator higher, not lower.

The PID should be executed as a function of distance and how fast that distance is changing.

 

[jim767]

The PID.UPD appears to change the interval at which the PID is running as I’ve tried both manipulating the PID with the .UPD setting to give an update every foot and another

method where I trigger the PID block with a bit set to go true every foot of travel. There appears to be no difference in the results. I am now just triggering with a bit so I can be sure I’m executing and not just computing new gains. With the (1/FPM)*60 I am executing it as a function of distance and see that the execution slows as line speed slows. I’m just using the line speed as a value to convert it to update every foot of travel.



In looking at the COi equation for the CO of the integrator term, I appear to be using (1/FPM)*60 to substitute for T (sample time) as you said and you would be correct and my action would be reversed for the CO due to the integrator term. The integrator term COi equation tells me this is what’s happening, I’m executing the PID every foot which is basically my sample time and runs the PID but since the COi is part of the PID block I don’t see how to reverse the integrator action. One run I did had the following settings:



Kc= 0.2

Ti = 0.5 (reset time, min per repeat)

T= (1/FPM)*60



Or do I want to reverse it? We’re thinking that if you had a constant error the slope of the integrator would be the same since at low line speeds you are taking small steps over short periods of time whereas at slow line speeds you are taking large steps over long periods of time.



fast speed – higher execution rate



























































































slow speed – slower execution



























































































I’ll keep researching, Thanks

 

[jim767]

I see my graphics did not come through.