PID for Web Tension Control

I have asked similar before, and I know there is plenty of PID info on this forum, what I am asking though is there anyone out there that is done much with Web tension control systems? We have an old wallpaper machine that we have replaced the DC motors and control systems with a servo and plc. There is a standard indiction motor on the lead nip that runs at a fixed speed, and then we adjust the torque on the trailing nip to adjust the tension.

I have the pid tuned pretty well for tension SP changes when running at a fixed speed, the problem is that if we give a large SP on initial startup, then it seems to take to long to acheive the requested tension.

I know this sounds like a tuning issue, I haven't been back to try and tune the loop even faster. The biggest issue is that there cannot be any hunting, and since is it paper, any overshooting can cause paper to break and possibly cause fires

The main reason for me posting this is to see if there is anyone that has had any experience with tension control system?

I've had limited experience with this type of system. It was a wallpaper factory and I've worked on their equipment a few times.

The systems I worked on generally had a specific Siemens card for the job because it has to calculate torque to turn the main roll, plus the diameter of the main roll. Obviously, as it unrolls and feeds the machine, the diameter is reduced and so the speed needs to increase to get the same amount of paper through.

If your system is slow to respond, you might want to check the encoder on the feed and make sure its not slipping or that its not got dust or grime in it. This can cause the fault you describe.

This is a wallpaper machine as well, actually it is an old english machine, i think it was second hand when it came down under.

I have pretty good tension response when adjusting the SP. The exercise we are trying to acheive is registration control, having been to a wallpaper factory you may be familiar with the principle. It as a Hot embossing machine that embossers a texture over a printed pattern.
So when the operator sets up the machine, it is in such a way that the pattern is pretty much on check , then the tension is adjusted to get the pattern in register pretty quickly, this is where the problem with my PID loop comes in, i am going to try and tune it even faster, responding, but need to avoing any hunting and overshoot especially.

I have a few other things that i am going to try, but am just after any suggestions from anyone that may have achieved the same thing with the hardware we have.

We have a closed loop system, there is a tension load cell feeding back to our PLC, into a PID with a Digital SP, controlling a torque reference to the servo drive.

Yep, I'm familiar with hot and cold emboss. Has the machine been working before or is this part of the initial set up?

If the machine has been running at your factory for a while and then suddenly started playing up, then I would check the encoders (there's normally one at the gathering end and one at the feed end). Also check you are getting the full swing from the tensioner (ie. 0 - 10 V or 4 - 20mA).

If the machine has never worked at your factory and this is the first time its run. Still check the encoders (it only takes on spec of dust in there to mess up the count enough to cause problems). Also check the ramp up/down time of the inverter and also the scaling parameters of the inverter. Double check all the wiring to make sure something isn't in the wrong place.

It might be worth giving the factory where this machine originated a phone and speak to one of their engineers. It could be a problem that they have had many times in the past and they might be able to suggest a fix?

Scott,

To eliminate the overshoot in our tension systems, I tend to dampen the proportional gain and rely more on integral control. This will tend to give you a slower response to a large SP change though and is probably what you have already done.

Two things to look at:

1. Can you set your drive up so that the drive parameter for the proportional gain is set externally (for example with an analog signal)? Or switched by an external digital signal? You could then have two control behaviors - one, dominated by proportional control if you were radically low on FB relative to SP, and the other as you approach setpoint and/or run . . .

2. If you can't do this, can you disable the PID loop on startup (or when radically low FB relative to SP), and then enable it as you approach setpoint and/or run? This effectively does the same thing as above . . .


Good luck,

Marc

Johnny,

What we are doing is actually upgrading the machine, so that the dc technology is redundant, and we use parts available off the shelf, and that service agents ( us ) are a lot closer than the UK, it means win win for the company, but for us given the project, a big learning curve in web tension control systems.

I am happy enough with the encoder at the moment, we had earlier issues with a dodgy coupling. There is about 4 p/r difference which equates to about 0.1 or less of a mm, depending on drum size, so that is workable.

Marc,
I have quite a fine proportional band, and a fast intergral time, with a bit of Derivative in their also, and can a response within 10 or so seconds.
What you suggested about the disabling the PID loop on startup is what we are going to try next, i am thinking about adjusting the torque directly and then adding the PID loop in when happy, just need to be careful on how sensitive this makes the tesnion response.

Thanks for you input, anything else is greatly appreciated.

Scott

scottmurphy said:

I have the pid tuned pretty well for tension SP changes when running at a fixed speed, the problem is that if we give a large SP on initial startup, then it seems to take to long to acheive the requested tension.

Click to expand...

I'm familiar with tissue rewinders but not wallpaper winders.
The various sections in the machine will have different inertias and therefore have different acceleration limits. Probably the embosser is the highest inertia. The line speed setpoint should ramp up at a rate suited to the embosser. Tissue paper stretches under tension - I assume wallpaper does as well, though probably not as much. The web takes on some spring-like properties. To avoid tension oscillations during speed changes, jerk should be minimised - use an S-curve ramp.

Thanks for you reply Gerry, good to see someone else from our corner out here.

Yes, wallpaper does stretch, but not as much as tissue i would imagine. Acutally, we stretch to acheive registration, and vice versa depending on our offsets.

During a speed change, I slow the PID loop right down so as to minimise response, then when at req speed, speed PID loop up again. All that happens here is that the tension increase's, then settles down again.

I am unfamiliar with the S-Curve ramp that you mentioned, I realise it is the charecteristic of the curve, but is it specific to a particular application, how is it calculated?

I use ControlLogix which has a function block to generate S-curve output from a step input. Other inputs control accel, decel, and jerk.
You said you were using a servo - your motion controller should be able to generate an S-curve.
What equipment are you using?

We are using an Omron PLC, with an Omron Servo drive. I have a tension feedback into a PID, and am controlling the torque of the servo drive with the PID OP. We use + - 10 v as the reference. At 0 reference, we get a tension of 32 ( approx for info )to increase the tension, we give a - reference ( torque against web travel ) and to decrease tension we give a positive reference ( torque with web travel ). This works quite well and reasonably smooth control, the main issue at present is the speed of the PID. eg if we start at 0 reference, and want to go to 38 tension SP, with the line running at low speed, then it just takes to long. I am going to keep trying to tune the loop so that it responds faster, but was just after any other suggestions from people that have acheived the same thing.

Scott; Is this the Pacific Wall Coverings Porirua site? If so you already have a ControlLogix implementation of what you are asking for on the existing vinyl coater.

Scott,
PID refresh time from Omrom is slower and why difficult to maintain tension stability during speed change. I used controllogix with PowerFlex 700S for winder application with good result. As gerry said that we used ramp function to get accel, deccel and jerk. Might be you consider to replace omrom PLC with Controllogix to get best result.
James

Are you using true torque control or speed control with torque limit or some other type of speed/torque combination? If you have a web break does the torque control axis run away (which it should in pure torque control)?

If you are sure you are running in pure torque control you probably need to account for some torque feed forward terms. Torque control is nice because you are directly controlling the attribute that affects tension. However, it brings with it the baggage of having to account for many more load components on your own. You need to account for friction, both static and dynamic, the torque offset for your desired tension, accel and decel torques, etc. Asking the tension loop PID to do all this makes for a tough tuning challange. If you get the loop tuned to respond to all the above load changes it may be too aggressive during normal run.

Check on your drive mode first and we will go from there.

Keith

Keith,
We are using pure torque control to control tension. I am not using speed limit as we ran out of analogue outputs, so what i do instead is if the web breaks, or the nip opens, then set the torque reference value to zero to prevent run away, this actually works quite well for this application. ( i am not sure how it would work with other applications? )

Thie biggest problems we have with this mode, is that any small changes in product quality ( such as a join ) can have amazing effects on the tension feedback, which when we have the loop responding very fast, can cause massive hunting to occur
These are all things that the more time I spend with this machine, the more I am learning how it responds under different conditions.

The loop is actually responding pretty well, and as I said above, the times it gets really aggressive is when it see's a massive change in feedback. I am trying to educate the operators in what to do if they see something in the product that will affect the tension fb, and they are getting better, and we are acheiving results slowly, and with less web breaks and fires!

Scott

Scott-

That's the other issue with load cell tension systems in torque mode. These systems have no inherent damping to them. You can't use derivative gain on the tension loop since the tension changes so quickly you become oscillatory very quickly. Also, the driving element (driven roll) is not directly coupled to the sensor (load cell). They are coupled through the web, which changes dynamics with product grade and, worse yet, does not always provide reliable coupling, like when the web is loose.

This is why alot of people like velocity based systems in cases like this. The velocity loop provides inherent system damping (prevents oscillation). But just as important, because the drive is controlling velocity, if something comes through that tries to slow the motor down (like a join) the drive very quickly increases torque to power through it and then very quickly drops the torque again when the disturbance passes, all without the tension loop knowing about it.

If your product has any stretch at all you might want to consider a velocity based system. Since you are going nip to nip you don't have to worry about changing diameters likemyou would with a center driven winder. So most of your tension could be developed with simple draw. Then add a small tweak from the load cells to account for system errors and you would have it. I have done this with machines running the base paper for wax paper very successfully.

Keith