Dancer web tension control

Hy sirs, we have the following special application:
A web is passing through a dancer which is mounted on two linear bearings.
One unwind motor is used to control the position of the dancer (keep it in the middle) by changing the web unwinding speed.
Also a servomotor, which is mechanically linked to the dancer, should provide a constant force on the dancer (constant tension on the web) by moving the dancer forward or backward, the actual force is measured by a load cell.
The web is subject to fluctuations (tension variation) from the main machine and for other reasons. After checking the linear bearings characteristic, we found that it has a variable friction coefficient which depends on many things (status of the parts, greasing, aging,…).
So this made our task very difficult. Any ideas on how the servomotor should react to compensate the frictions and provide a constant force on the dancer?
Appreciate your help.
Thank you

Is this a dancer position control trimmed with a load cell? or an accumulator type control? Seems a bit fuzzy.

If it's as you describe, then I'm thinking the dancer position loop is closed by controlling the unwind speed. So, if you then want to control web tension separately, you would use a different control loop, with a tension setpoint, load-cell feedback, and apply the result to the torque command of the servo motor, as long as it is not at a limit.

We have two motors. The dancer position is controlled by the unwind motor (with potentiometer feedback). On the other hand, the web tension is controlled by the dancer servomotor separately as you described. This is not a problem.
The problem is how to compensate for the fluctuations in the linear bearings friction which will lead to high instability in the tension loop (the dancer servomotor will oscillate continuously). This is caused by the characteristics of the linear bearing which produce frictions dependent on the dancer linear speed (which could be stationary) and by the aging of the mechanical system (greasing,…).

There really is nothing you can do to 'compensate' for the variable forces in the mechanism. As you already stated the forces aren't consistent and aren't readily measurable. You can't really compensate for somehting you don't know about. You are very likely running into a stick-slip type situation. The resulting displacement will inevitably cause tension fluctuation.

Unless the mechanical system is modified to produce less friction you are stuck. All you can do is try to maintain the control system stability by operating with lower tension correction gains and allowing greater instantaneous tension variability.

Keith

Kataeb said:

We have two motors. The dancer position is controlled by the unwind motor (with potentiometer feedback). On the other hand, the web tension is controlled by the dancer servomotor separately as you described. This is not a problem.
The problem is how to compensate for the fluctuations in the linear bearings friction which will lead to high instability in the tension loop (the dancer servomotor will oscillate continuously). This is caused by the characteristics of the linear bearing which produce frictions dependent on the dancer linear speed (which could be stationary) and by the aging of the mechanical system (greasing,…).

Click to expand...

Right... So, again, why not control the torque on the servomotor (which is loading the dancer), with feedback from the loadcell?.

You really don't care at all about frictional losses in the linear bearings, or the actual force generated by the servo, you only care about web tension, which the servo-loaded dancer should be able to regulate.

If you are suffering from 'stiction', which is that the dancer tension roll actually binds in place due to bearing problems, you might try adding a small, medium frequency dither signal to the torque command input to the servo drive. (Dithering is just a square wave of some amplitude to introduce a very slight vibration into a system in order to try and prevent stiction).

Not trying to start a war. Why not do PM on a regular basis and RR the bearings when the end of life is up.

We had an oldtimer that always pointed to mechanical only because he couldn't do any of that new fangled electrical stuff. He could change bearings, gears, ect. This guy was a great asset, most of his replacements always blame the eletrical stuff.

Originally posted by rdrast:

Right... So, again, why not control the torque on the servomotor (which is loading the dancer), with feedback from the loadcell?.

Click to expand...

The way I read the post the OP is already doing that. This is where the 'instability' is coming from. If the OP was just loading up the servo with a fixed torque the dancer would be stable but the tension would have error.

Originally posted by rdrast:

If you are suffering from 'stiction', which is that the dancer tension roll actually binds in place due to bearing problems, you might try adding a small, medium frequency dither signal to the torque command input to the servo drive.

Click to expand...

This is pretty easy to add so I would certainly give it a try. But generally speaking you need displacement to make dither work well. Bouncing a valve spool around at high frequency with a low amplitude is one thing. Getting a dancer assembly to do that but not influence tension is a little different story.

We have tried linear dancers with linear bearings in the past and have largely gone away from them. The design intend of a linear bearing (accurate motion with little influence from load) doesn't fit well with the design intend of a linear dancer guide (as near zero friction as you can get). Manufacturers build linear bearings way too tight (to get the accuracy) to provide low friction motion. We quite often go with rolling element support with rack and pinion alignment. Or we go with pivoting arm systems if it is just a dancer and not also an accumulator. The pivoting arm definitely provides the lowest friction but either is better than a linear bearing system.

Keith

Keith on you pivot arms do you add mass to the counter to decrease the effective loading caused by the rol weight? If so how do you overcome the added mass to the system?

Originally posted by JeffKiper:

Keith on you pivot arms do you add mass to the counter to decrease the effective loading caused by the rol weight? If so how do you overcome the added mass to the system?

Click to expand...

We just upsize the loading device (usually rolling diaphram air cylinders) to account for the assembly weight plus the force required to generate web tension. We stay away from counterbalances for exacly the reason you state. In fact we often will cut as many holes in the support arms as we can to keep the rotating mass down.

One thing we don't do much is inertial balancing. Generally speaking our parent rolls are round enough that we don't get much benefit.

By the way, you btring up a good point. pivoting dancers should be neutral in the horizontal position. Don't design dancers that hang in the neutral position. Horizontally neurtal dancers produce less change in loading as they swing (due to the change in position of the effect of gravity on the roll) than vertical dancers do.

Keith

kamenges said:

We just upsize the loading device (usually rolling diaphram air cylinders) to account for the assembly weight plus the force required to generate web tension. We stay away from counterbalances for exacly the reason you state. In fact we often will cut as many holes in the support arms as we can to keep the rotating mass down.

One thing we don't do much is inertial balancing. Generally speaking our parent rolls are round enough that we don't get much benefit.

By the way, you btring up a good point. pivoting dancers should be neutral in the horizontal position. Don't design dancers that hang in the neutral position. Horizontally neurtal dancers produce less change in loading as they swing (due to the change in position of the effect of gravity on the roll) than vertical dancers do.

Keith

Click to expand...

SO no Webex rolls for you? If you have good rolls (not welded pipe) the effect of the high dollar rolld are 0?
On a horizontal dancer once you get above 45° doesn't the loading from gravity cause you problems?

Originally posted by JeffKiper:

SO no Webex rolls for you? If you have good rolls (not welded pipe) the effect of the high dollar rolld are 0?

Click to expand...

I think you misunderstand. I didn't say we don't dynamically balance rolls. We do. However, we don't inertially balance dancer assemblies. Inertially balancing a dancer assembly means you make the web tension required to accelerate the dancer roll equal to the web tension required to accelerate the dancer assembly given an infeed web acceleration and no outfeed web acceleration. If you do that dancer web tension disturbances due to infeed web acceleration (like an out-of-round roll) won't be caused by accelerating the dancer assembly. But there is no free lunch. The tension variation you DON'T get on the outfeed side you DO see on the infeed side. And it all falls apart if you start to slip around the dancer roll. If you have ever seen single roll pivoting dancers with what look like donuts on the dancer roll, that is what they are there for.

As for the dancer stroke we don't try to operate that far out. We build them with about +/-30 degrees of rotation. We eat about 5 degrees of that on each end with dampers. So the mechanically active area is about +/- 25 degrees. That produces a +/-10% of assembly weight in loading variation. Unless we are having a bad day we shouod be able to stay within +/-5 degrees pretty easy. That produces a +/-0.5% variation. That same 5 degree variation on a vertically hanging dancer produces about a 9% variation. 25 degrees of variation produces about 42% of the assembly weight in loading variation.

By the way, did you know I work for Webex and were tweeking me a bit or was that just an example?

Keith

kamenges said:

By the way, did you know I work for Webex and were tweeking me a bit or was that just an example? Keith

Click to expand...

No I didn't know. My boss has gotten me to the point that I will only use Webex rolls. So if I do a project it will be a Webex period nothing else.

I am an electrical geek not a process guy (as you know from some of my other post) so I don't really understand all the webhandling stuff yet. I am trying to learn what and why we do these things.

What Load cell systems do you all use? We are looking at DFE and MagPower. It looks like they both have some great products. Everybody has thier pros and cons.

Originally posted by JeffKiper:

My boss has gotten me to the point that I will only use Webex rolls. So if I do a project it will be a Webex period nothing else.

Click to expand...

Well, at least we have that going for us. Granted, I'm partial. But I do think we do a pretty good roll. At the very least we will look at roll body and shaft dynamics and not just say it has to be somewhere near round and kind of smooth. Now whether you really need that close a look is always debatable.

Originally posted by JeffKiper:

What Load cell systems do you all use? We are looking at DFE and MagPower. It looks like they both have some great products. Everybody has thier pros and cons.

Click to expand...

We have been and still are all over the place on this. We do 99% of our tension control either in a drive or in a plc. So I really can't speak for anyone's closed-loop controllers. I haven't been able to nail down any huge technical differences between suppliers. We have been using alot of CMC in the past few years. They have been a good for us with delivery and the amplifiers have a decent form factor and are easy to adjust. I also like that the tension output goes bipolar. That helps eliminate offsets due to noise at low tension levels. They certainly aren't the cheapest out there but, being with Webex, I can't hardly hold that against them, can I. CMC is actually the only manufacturer that understood why I would want them to incorporate a high pole count low pass filter in their loadcell amplifiers. So I give them extra points for that.

Keith

If the dancer position is subject to varying friction/wear then use a position controller with the load cell trimming the position demand to maintain the tension at a constant value. The position of the dancer would also feed the "normal" control loop for the unwind roll speed control thus keeping the dancer operating about it's mid position.