Winding technology.

[Peter Nachtwey]

[/LIST]QUESTIONS
However I think you are going to have control issues with keeping constant peripheral speed on drum to match input string speed and prevent slacking or stretching on string.

If the feed speed is known then it should be easy, with a little trig and calculus, to compute the winder speed. Ok, it isn't so easy, but I can do it with a little thought. I am waiting to see how Knut does this. I am also interested in how this works in a PLC. A real motion controller would be much better for this this application. I would use ST or SCL for the programming. Most motion controllers would use a cam table but I bet I can do the calculations and updates on-the-fly. I think I can even take into account the angle changes at the wire wraps across the winder.

 

[Peter Nachtwey]

The calculations will be fun.

[/LIST]QUESTIONS
However I think you are going to have control issues with keeping constant peripheral speed on drum to match input string speed and prevent slacking or stretching on string.

If the feed speed is known then it should be easy, with a little trig and calculus, to compute the winder speed. Ok, it isn't so easy, but I can do it with a little thought. I am waiting to see how Knut does this. I am also interested in how this works in a PLC. A real motion controller would be much better for this this application. I would use ST or SCL for the programming. Most motion controllers would use a cam table but I bet I can do the calculations and updates on-the-fly. I think I can even take into account the angle changes at the wire wraps across the winder.

 

[Paul B]

Is Peter multitasking, I bet so.

 

[PLC]

Peter
Since the S7 T cpu is a real motion controller, he
Can build cams , create cams on the fly use virtual masters
etc......

 

[Honga]

Apparent diameter x:

x = d/sqrt(2) * cos(alpha)

Where alpha = 45+ smallest angle from drum centerline to thread.
d = drum width

given w = V/x
V_barrel = w*x

Therefore w's function to normalise velocity to zero should be
V_barrel = 1/x*V_thread

 

[leitmotif]

If the feed speed is known then it should be easy, with a little trig and calculus, to compute the winder speed. Ok, it isn't so easy, but I can do it with a little thought. I am waiting to see how Knut does this. I am also interested in how this works in a PLC. A real motion controller would be much better for this this application. I would use ST or SCL for the programming. Most motion controllers would use a cam table but I bet I can do the calculations and updates on-the-fly. I think I can even take into account the angle changes at the wire wraps across the winder.

He did mention 1 RPM on winder so it is a much easier control issue. Sorry I missed the number.

What really intrigues me is what is he going to do with only one wrap on it?

Dan Bentler

 

[knutabru]

He did mention 1 RPM on winder so it is a much easier control issue. Sorry I missed the number.

What really intrigues me is what is he going to do with only one wrap on it?

Dan Bentler

Hello Mr. Bentler.

The machine will have two drums, so when the first one are wound up, the traversing ball-screw, will advance to the next and the operator will attach the string to the next drum, cut it from the first and then remove the drum for further processing.

If I land on a solution that uses a CAM-table the drums must have a keyed spline for remounting so I can avoid a re-sync for every new winding. Since the rotation are so slow, the operators will have no problem doing this operation without stoping the extruder. -It's easier to waste some meters of thread than to stop the extruder.

Knut.

 

[Peter Nachtwey]

Apparent diameter x:

x = d/sqrt(2) * cos(alpha)

Where alpha = 45+ smallest angle from drum centerline to thread.
d = drum width

given w = V/x
V_barrel = w*x

Therefore w's function to normalise velocity to zero should be
V_barrel = 1/x*V_thread

It is more complicated than that. I have done a previous problem that is similar.

 

[504bloke]

You could just use SSD 890 Drives which have winder functions built into them, plug in a redlion G3 and away you go !

 

[Peter Nachtwey]

Ah, but that is the problem!!!

You could just use SSD 890 Drives which have winder functions built into them, plug in a redlion G3 and away you go !

What are the winder functions or formulas? Cam tables are canned routines but the right information must be put into them.

Honga got the radius to the corner right but there is more.
The radius to the corner is not the effective radius at all times This is really a triangle geometry problem with calculus to calculate the rate of change in one angle as a function of the growth in one side of the triangle.

I have given a big hint. Lets see who is up to the challenge.

 

[kamenges]

Originally posted bu knutabru:

The thread can be thight but does not tolerate beeing stretched.

That statement contains incongruous elements. It can't be tight if it didn't stretch. You need to put limits on either or both of tension and stetch.

Originally posted by 504bloke:

You could just use SSD 890 Drives...

Originally posted by Peter Nachtwey:

What are the winder functions or formulas?

There are two different flavors; a speed programmed winder and a current programmed winder. The SPW uses a base PID with (optional)gain scheduling, a speed ratio diameter estimator and (optional) accel feed forward to handle center driven winding. I really don't think this would work here. The basic premise of the SPW is that the diameter is changing slowly relative to incoming web speed. In this case, while the change in diameter is very deterministic it is not changing slowly relative to web speed.

The CPW is effectively a dynamic current limiter that will limit the motor torque based on tension command and calculated diameter. This would have a better shot. Both methods have some handy aded functions that make implementing a typical unwind/winder realatively easy.

I might take a crack at this thing tonight if no one else posts a solution.

Keith

 

[Honga]

Take my previous formula.

To account for winding angle - Substitute d for d_apparent:

d_apparent = sqrt(d^2+W^2)/4

Where d is your drum side width and W is the width of the thread.

Working off your challenge Peter!

EDIT: Second thoughts. When thread like this winds up, does it wind up neatly, or do you normally run something to guide it?

 

[Peter Nachtwey]

??????

d_apparent = sqrt(d^2+W^2)/4

No. There is a simple trig formula that needs to be used. That is a hint.

Does anybody know what the cam profile will look like in general? My engineers intuition says there are going to be discontinuities in the derivatives when the edge of the winder hits the thread. I think the motion profile will look like waves but not sine waves but more like scallops where the points are up. This discontinuity will make it hard for third order cam table interpolators to follow.

There are four steps to this problem
1. Finding the right equation.
2. Converting it to a rate as a function of another rate.
3. integrating to convert the rate back to positions for the cam. The cam table could be done as rates but I better the errors would build up over time. Maybe it doesn't make any difference over just one level of winding.
4. Finding where the transition points are where the wire hits the edge of the winder.

I have solved similar problems in 3D. This problem has a twist in that there is a discontinuity. That discontinuity has got me thinking about the best way to handle it.

I love problems like this.

 

[Honga]

Sorry I meant winding angle in the direction of the axis of rotation.

The plot of the first function was as you described the cam profile Peter. Without allowing for any of the other factors I suppose it could be used as a hint.

Just started working on the equation to include the lengthening and shortening of the extrude to the barrel. I overlooked that in the originial, assume the extruder and barrel are fixed.

 

[Honga]

Tell me if I'm not on the right track lol