torque winder vs. speed winder

This is not a PLC question. But I know there are many members who are experts in winding. So here it goes.
We are designing slitter-rewinder machine. This machine will be used to unwind various narrow (up to 17" wide) web materials like paper or films, slit these materials to multiple streams and rewind these multiple streams on the individual cores. Example is the web with 5-6 rows of printed labels across is converted to 5-6 individual rolls which have only one label across.
Rewind shaft of the machine is driven by the motor and VFD. Until now in these types of application we have configured drive in closed loop speed mode with external PID closed loop of tension. Tension feed back was coming from load cell. Signal from tension PID loop was trimming speed reference of the drive to achieve desired tension and produce good quality rolls. Max tension that machine is capable of is 20lb. Both speed loop and tension loop were residing in the drive software.
At this new machine someone is suggesting to configure rewind drive in torque mode still with tension feed back from the load cell, saying that in torque mode we'll have better tension control.
Why tension control is better, when drive is configured in the torque mode? In torque mode we'll have only two cascaded closed loops - torque loop and tension loop. In speed mode we'll have three cascaded closed loops - torque loop, speed loop, tension loop. Is it when less closed loops then better control?
I've read in few places that both speed mode and torque mode of the drive are successfully used in winder applications. But I couldn't find comparison of these two methods, their advantages and disadvantages, and recommendations when to use one or another.
Does anyone have some sources of information on this subject?
Thank you

PaulBr.

Torque is rotational force. By controlling the torque/force or pressure you maintain a smoother control on the actual tension. The speed of the rewinder will automaitcally increase as the roll builds. You will achieve tighter wraps and consistant wrapping too.

The speed of the material passing from unwind to rewind will always has the same pull exerted in torque control. In speed control, as long as the material is moving along it wraps. If for any reason the material or machine began to bind, the drive tries to make the material move at the set speed. Usually resulting the the material parting. In torque control, the material can actually stop if it binds and it will not have more tension put on it than desired.

paulB said:

Why tension control is better, when drive is configured in the torque mode?

Click to expand...

Torque is rotational force. By controlling the torque/force or pressure you maintain a smoother control on the actual tension.

Click to expand...

Maintaining a constant tension while accelerating and decelerating is the ultimate goal. Torque control is better than velocity control because rotational acceleration leads rotational velocity whereas positions control lag both the velocity and acceleration. Torque and tension control would be the same if the torque required to over come friction and to accelerate and decelerate are ignored. Don't ignore the torque required for accelreration and friction. A tension PID that just responds to error will see big errors during acceleration and deceleration if you don't use velocity and acceleration feed forwards. The tension PID should just be used as a trim for errors in the feed forwards.

Oh, the velocity loop should just limit the speed incase the system finds itself in a no load situation as in the material does separate.

Peter Nachtwey said:

Don't ignore the torque required for accelreration and friction. A tension PID that just responds to error will see big errors during acceleration and deceleration if you don't use velocity and acceleration feed forwards. The tension PID should just be used as a trim for errors in the feed forwards.

Click to expand...

Peter, does acceleration feed forward represent torque compensating for inertia? Does velocity feed forward represent torque compensating for friction loses?
Inertia of winding roll will depend of type (density) of winding material, which will be unknown to us. Friction loses will depend not only on type of material, but also on quality of slitting knives, and number of slitted streams, which are also unknown.
How close you need to be in determining amount of required feed forwards?
It seems that, when drive is in speed mode, you provide speed reference to the drive according to the surface speed of the web and building roll diameter, and you trim this speed reference with tension loop output. There still may be a need to provide feed forward for inertia compensation (acceleration/deceleration feed forward), but you may get by without guesing amount of friction loss compensation. Am I right?

PaulB

paulB said:

Peter, does acceleration feed forward represent torque compensating for inertia? Does velocity feed forward represent torque compensating for friction loses?

Click to expand...

Yes.

paulB said:

Inertia of winding roll will depend of type (density) of winding material, which will be unknown to us. Friction loses will depend not only on type of material, but also on quality of slitting knives, and number of slitted streams, which are also unknown.

Click to expand...

If you have auto tuning tools this is not a problem because the auto tuning tools adjust the feed forward gains.

How close you need to be in determining amount of required feed forwards?

Click to expand...

If you are within 10% of optimum that means the PID only needs the supply the remaining 10%. This will reduce your error or response time by about a factor of 10 because you need 1/10th the error to get the same control signal.

It seems that, when drive is in speed mode, you provide speed reference to the drive according to the surface speed of the web and building roll diameter, and you trim this speed reference with tension loop output. There still may be a need to provide feed forward for inertia compensation (acceleration/deceleration feed forward), but you may get by without guesing amount of friction loss compensation. Am I right?
PaulB

Click to expand...

Yes, the torque lost to friction should be small compared to the torque required to accelerate and decelerate. Plus frictional loss are a steady state loss. I would still use both the acceleration and velocity feed forwards.

In my experience with closed loop tension control I have found that I have needed both speed and torque control - the choice depends on the magnitude of the torque demand. If the torque demand is say below 3% of the drive rating, then most drives will not actually give you 3% - so you use speed control. Once the torque demand has built up to say 8%, you switch over to torque control (with a speed limit in case of web break)

Unless you have massive machine acceleration, inertia comp only really comes into use with diameters greater than 500mm.

Losses - the friction terms are usually temperature dependant so the amount of loss you need depends on how long ago the rewind was used.

If it worked before with speed control, stick with it.

A quick back to basics, is this a two drum winder, or a single shaft? If a single shaft, is the shaft directly driven, or does the windup core ride against a driven drum? Does it have individual motors on each drum, shaft and/or the unwind?

Typically torque control is the way to go in any event, and one way to compensate for the "big errors during acceleration and deceleration" is to use preset values for "thread mode tension" and "stall tension" which are triggered by speed. Yes Inertia of UN-winding roll is important for proper tension control during acceleration, inertia of the wound up rolls during deceleration, depending on how things are driven. If the unwind stand has a VFD motor (vs a less expensive brake system) the diameter of the UN-winding roll and an average density can be used to develop the appropriate tension. The load cells would be used only to check the deviation from calculated tension values...



Best roll structure starts off highest tension at the first wrap on the windup device, and ever decreasing tension as the roll gets larger and larger.

SimonGoldsworthy said:

In my experience with closed loop tension control I have found that I have needed both speed and torque control - the choice depends on the magnitude of the torque demand. If the torque demand is say below 3% of the drive rating, then most drives will not actually give you 3% - so you use speed control. Once the torque demand has built up to say 8%, you switch over to torque control (with a speed limit in case of web break)

Unless you have massive machine acceleration, inertia comp only really comes into use with diameters greater than 500mm.

Losses - the friction terms are usually temperature dependant so the amount of loss you need depends on how long ago the rewind was used.

If it worked before with speed control, stick with it.

Click to expand...

Simon,

Existing design is working well with speed control. We would like to see that machine perform better when run with lower tension setting. Some customers are running fanky materials and need "soft" rolls, which require low tension setting. From you post conlusion is when tension setting are low, torque demand is low and better to use speed mode than torque mode.
For this application friction losses not only losses in mechanical moving parts which depends on temperature and age. They are also losses during slitting web and depend on number of slits and quality of knives.

Thanks for the input.

paperboy4828 said:

A quick back to basics, is this a two drum winder, or a single shaft? If a single shaft, is the shaft directly driven, or does the windup core ride against a driven drum? Does it have individual motors on each drum, shaft and/or the unwind?

Typically torque control is the way to go in any event, and one way to compensate for the "big errors during acceleration and deceleration" is to use preset values for "thread mode tension" and "stall tension" which are triggered by speed. Yes Inertia of UN-winding roll is important for proper tension control during acceleration, inertia of the wound up rolls during deceleration, depending on how things are driven. If the unwind stand has a VFD motor (vs a less expensive brake system) the diameter of the UN-winding roll and an average density can be used to develop the appropriate tension. The load cells would be used only to check the deviation from calculated tension values...



Best roll structure starts off highest tension at the first wrap on the windup device, and ever decreasing tension as the roll gets larger and larger.

Click to expand...

This is single shaft directly driven winder. It has motors on rewind shaft and on draw roll. Unwind is not driven. It is brake system. We are not very concerned with unwind tension. We are able to control it to provide web running without bagging or stretching. But rewind tension is important because it determines quality of the finished roll.
Thanks for the input.

I would tend to stay with speed control. Torque control can work but, as Simon and Peter have stated, it gets very difficult if the system losses are a significant portion of the total torque required. In any case feed forward values for tension setpoint, velocity and acceleration are very important.

I would only attempt true closed loop torque control if the mechanical drive ratio is very low, say 3:1 or so. If you go significantly above that it becomes harder and harder to tell the difference between tension torque and system loss torque.

The biggest headache with speed based center driven rewinds is getting the diameter right. The majority of the speed reference is a velocity term based on line speed and roll diameter. If the roll diameter is wrong the baseline spindle speed will be wrong and you are in for a tough time. The good thing about speed based center winders is they can be tuned relatively tightly and still stay stable. The velocity loop will tend to keep the roll from bouncing around too much.

Torque based winders are comparitively insensitive to diameter. However, they have very little (or no) damping. This means you can very quickly get into oscillation issues at relatively low gain settings. You can't effectively use the differential of a load cell signal because of the inherent mechanical noise and the relatively high rate of change of tension as the web is pulled against the load cell roll. Using spindle velocity for damping is an issue of it's own as you would need to compare that to a feed forward term based on dimaeter and you are right back to the issue you have with a speed based winder; dependence on an accurate diameter.

I think a torque based winder with an acceleration damping term might be the ticket. But I haven't done a system yet with small enough interrupt jitter and high enough feedback resolution to produce a stable second derivative of position. One of these days I might bite the bullet and talk to Peter about that.

I can't argue that a torque based system produces more direct control over the variable you are trying to control. It just comes along with alot of baggage you need to manage.

Keith

In the end you are always doing torque control

You may think you are only controlling the velocity loop but somehow the drive is sending current through the armature that creates torque. If you aren't controlling the current then who or what is? Velocity mode will work in many cases where the acceleration or deceleration is slow.

Jitter? What jitter? The sampling is all done in a FPGA. No software required for the sampling. I bet all the better motion control manufacturers use FPGAs to do as much as possible. It is the encoder resolution that is the big problem.

I do most of my encoder testing in torque mode. In fact I haven't had a drive with an internal velocity loop for years. I need to get a torque demo going.

Originally posted by Peter Nachtwey:

You may think you are only controlling the velocity loop but somehow the drive is sending current through the armature that creates torque.

Click to expand...

That is true. However, you do need a minimum loop closure rate to effectively control velocity. Which brings me to my next point:

Originally posted by Peter Nachtwey:

I bet all the better motion control manufacturers use FPGAs to do as much as possible.

Click to expand...

You are making an assumption, that being that these applications are accomplished with products from 'the better motion control manufacturers'. This is usually not the case. Many end users have preferences for drive and control platforms. Those preferences don't often include 'the better motion control manufacturers'. So guys like me are stuck making silk purses from sows ears (I think that's the saying) on a regular basis. Do me a favor and start banging on my customers. Then I might finally be able to use the good stuff.

Originally posted by Peter Nachtwey:

Velocity mode will work in many cases where the acceleration or deceleration is slow.

Click to expand...

Accel rate really isn't a limitation outside of the obvious 'thou shalt not violate maximum torque' thing. The drives I have used for winding applications have a torque offset field that is added to the velocity loop output. Assuming a relatively stiff mechanical system and some jerk on the main webspeed reference one can develop an accel feed forward torque value that bypasses the velocity loop and is sent directly to the current loop. This largely eliminates the tension upsets due to accel.

Originally posted by Peter Nachtwey:

In fact I haven't had a drive with an internal velocity loop for years.

Click to expand...

That is certainly the way to go if you have the platform to pull it off. Ahh, maybe someday...

Keith

We have to do the automation for the same machine will be used to wind and unwind , with motor on winding and motor on unwinding . The current (Amps) read from the winding inverter could be used to control the tension in film? We have only a sensor for diameter and VFD.