Positioning with Cognex camera

aand74

Member
Join Date
Dec 2005
Location
Deinze
Posts
131
I am working on a continuous production line, more in specific where vynil is produced. At some point they have implemented a 'positioning frame', which is in fact a roller that can turn some degrees to the left or right of its central mounting position.By turning the roller the vynil can be guided to a requested position.
For the automatic positioning a cognex camera is used as sensor for the actual position of the vynil (the Cognex recognizes the edge of the vynil).
There is a drive with integrated position controller that positions the roller.
The position setpoint is generated in the central Siemens S7 plc.
There a PI controller has as setpoint of 0 and as actual value the distance from the edge of the vynil to the Cognex 'zero position'. The Cognex camera itself can also be positioned (independent of the guiding roller) to have another position setpoint for the vynil.
The position setpoint from the PI controller in the plc is sent to the drive of the guiding roller as a position setpoint of an absolute positioning.
The described system was already implemented before on another part of the production line. We have copied this system to our new part, but we have quite some overshoot when a new position is requested for the vynil.
We can off course tweak the PI values of the PLC PI controller,
but I wanted to get some opinions about this whole positioning system.
I am in fact more thinking about doing relative positioning in the drive and using only p-factor in the plc. We may not forget that there is not a fixed direct relationship between the angle of the guiding roll and the position the vynil will take. Moving the guiding roller to a certain extent will guide the vynil also in a certain direction, but it can not be exactly calculated. That is probably why the other company used PI controller in the plc. Need to say also that there is not an isochronous communication between plc and Cognex and drive. It is Profinet, should be rather fast, but not clocked.
So how about my idea of relative positioning in spite of the actual absolute positioning, and only proportional factor.
 
I like the idea of absolute positioning
You could use the technology controller in the drive for P or PI control
But the actual position of vinyl still comes from the plc (camera)
 
Last edited:
When a new SP is entered, have you thought about ramping the setpoint into the PID block. It means that the PID block doesnt see a massive change. It allows the PID block to make only minor corrections to keep tracking along with the setpoint change as it ramps to the final value.
This method allows you to have good disturbance rejection as you can have relatively sensitive gains, whilst still delivering finesse when changing setpoints.
 
Steering roll web guiding is a force differential process. Basically the steering action imposes a cross-machine force on the web that is ultimately counteracted by the webs desire to run in another position. People a whole lot smarter than me have determined through various methods that is takes roughly 5 free web spans for a web to establish position equilibrium after a change in cross-machine forces takes effect. By "free web span" I mean the distance from the steering roll to the controlling roll upstream of it.

What this means to you is that the rate of correction is not fixed. The web will respond consistently in position but it will not respond consistently in time. The faster the web moves the more quickly the web will establish its final position in the time domain. Since your PI controller is time-based you will need to adjust its gains based on how fast the web is moving. In your case this may be easy if you always produce vinyl at the same rate. However, if the vinyl productions rate changes you will either need to set your gains to those that work for the slowest vinyl speed or modify the gains based on speed.

If you evaluate the way the position command value changes, your idea of using proportional-only relative positioning is identical to integral -only absolute positioning. I don't think there is much to be gained there.

Keith
 
People a whole lot smarter than me have determined through various methods that is takes roughly 5 free web spans for a web to establish position equilibrium after a change in cross-machine forces takes effect. By "free web span" I mean the distance from the steering roll to the controlling roll upstream of it.
I need to think about this. I know what they are getting at and I will address that below.

What this means to you is that the rate of correction is not fixed. The web will respond consistently in position but it will not respond consistently in time. The faster the web moves the more quickly the web will establish its final position in the time domain.
That is the way it should work.

Since your PI controller is time-based you will need to adjust its gains based on how fast the web is moving.
Think about this. Usually we think of PID control operating every constant time interval. In this case the PID should update as the line moves a certain distance interval. This means the PID will update very slowly or not at all as the line slows or stops.

Keith's method would work but if the PID is accumulating error while stopped and then the line starts again the increased KI gain multiplied by the accumulated error will cause a huge change in output. The integrator should not be accumulating error or control output while stopped. If you update the PID every distance interval this will not happen.

Since the roll position is an integrating process the output should be scaled by the fraction of full speed the line is currently moving. This way the drive will not be trying to force the roll to turn while stopped due to the proportional gain alone.
 

Similar Topics

The system of position control is quite different in Beckhoff compared to Siemens. In a Beckhoff controller with Beckhoff drives on Ethercat, the...
Replies
3
Views
837
I'm working on a SCADA project in FTView SE where there are many instruments, each of them has a Faceplate that can be called by a "Go-To Button"...
Replies
1
Views
544
Hi, I am using P&F - WCS reader to read position of a trolley motion on a crane. Sometimes there is Ethernet communication failure error given by...
Replies
1
Views
1,029
We have a project with a couple different fairly simple positioning tasks. Speeds are fairly slow and positioning doesn't need to be super...
Replies
3
Views
2,013
Hi, I'm stumped with this. I'm attempting to perform some positioning using a Powerflex 755 series VFD. First time I've done something like this...
Replies
5
Views
3,454
Back
Top Bottom