design of controls

[hooligan]

hi everyone,

i have a problem design a control system that i need help with. i have two webs that have holes punched in the material. the holes are both punched the same distance apart. the materials are then wound on two seperate rolls of material. i am feeding the two rolls of material into a machine. i want to align the holes of the two materials as they run into the machine. i have brakes on both of the unwind shafts so that i can control the drag on the materials. i also have sensors that can measure the registration of the two materials. i have a plc that can take the data from the sensors and manipulate the tension on the brakes. my trouble is coming up with a method to calculate how much brake i need to apply. for simplicity i was starting with manipulating one brake only, i am setting the other one manually for now. what i am doing now is i am taking the mismatch data and calculating a new tension value by the following: T new = [T old(MM old)]/ MM delta; MM delta would be the change in missmatch from the one previous to the most recent.
essentially i am using a proportonal method of changing the tension to try to minimize the mismatch (MM). my trouble arise when i have a large tension value and a small change in the mismatch value (there is a reaction time from the time you make a change to the time you actually see a difference) the above equation tends to blow up (divide by zero).

Any suggestion on how to better control the system? Any help is greatly appreciated!!

 

[RadarRay]

I would normally make anyone I work with give me a one line diagram of what you are describing before I would tender an answer, so if you can post a complete one line of the machine somewhere I can look at I will comment on what you can do.

I have done applications which are somewhat similar, so I will ask you some questions with the intent they will show you what you are up against with the way you are applying the machine now.

To align one set of holes with the other, you have to get one set to displace with respect to the other. The only way to do this in your approach is to make the controlled web slip with respect to the fixed web. Does your application allow this? Now while you are trying to get the web to slip, how much does it stretch? If you stretch the second web with respect to the first, your holes are no longer equidistant on both webs.

So your app now depends on how much a web will stretch and how much a machine will make it slip. And then you have to overcome what could be long time constants in you system response.

Hope this gives you an idea on what you are trying to overcome. I can't say it is impossible to do it the way it is but it is difficult.

When both holes line up, what do you do to the two webs?