Torque Control

Hi Guys;
I am trying to control the tension on a yarn. Doing so, I plan to design a PID system and implement via a PLC. The tension can be measured by a load cell which I can use as a feedback to my system. Tension on the yarn can be modified by a spring system, when you strecth the spring the tension increases and vice versa. I plan to strecth the spring via a servomotor, and control the torque on the motor so that in a sense I will control the tension on the yarn. My question is about the control loop. Since my reference and measured value will be the tension, my error will be in the form of tension. May I use this error to feed the control and the plant (motor)? Or do I have to convert the tension value into some other thing?

Well, the difference between your setpoint and your feedback is indeed your error, and in general, you control a loop based on error, so that should work.

What scares me here, is all of the 'springiness'... I sense that you may find lots of problems related to mechanical harmonics that may be tough to tune out.

do you mean the reaction of the spring can be different than what I expect?

Well, I don't know how linear it is, but more so, if you are using a load-cell for feedback, and are extending a spring to create tension (I'm assuming this 'spring' is attached to a rope-brake or something similar on the bobbin) you can develop an oscillation due to the lag in reaction times.

Try it, maybe there will be no troubles.

thank you rdrast. I hope I can find you in the future as I proceed in the project.

Load cell tensioning with purely torque based systems can be a little challenging to control. The problem is it is difficult to use any derivative gain in the tension loop as the tension changes so fast. But without derivative gain the system has no real damping and is prone to oscillation.


cakkostepen, you have an additional issue with the way you are applying tension. Between the motor and spring you have a spring-mass system with no damping. The motor-spring system will be very susceptible to oscillation.

One question I have is can you replace the band brake system and connect the motor directly to spindle? This would eliminate at least one point of low frequency oscillation.

Another option may be dual loop control. Close a tension loop around the load cell. However, send that output as a position command to the servo. Based on spring rate you should know how much force you will get for a given motor displacement. Then close the position loop around motor position. This will allow you to use some D-gain in the position loop t odecrease the chance of motor-spring oscillation. Then you can use a PI loop for tension, which tends to work better for load cell systems.

A third option might be to use a diaphram cylinder and an E/P transducer instead of the motor to generate spring force. The E/P-cylinder-spring system will tend to oscillate at a higher frequncy due to the low mass of the cylinder shaft and piston. Also, the E/P will provide a little bit of inherent damping due to the restriction in air flow.

Hope this helps.
Keith

I also agree that using spring tension for feedback would be rather squirrely. Another idea would be to create a drop loop in the yarn line by using a weighted roll to pass the yarn under. Devise a way to measure the height of the weighted roll and use the height as the feedback signal. As tension on the yarn increases so does roll height. Of course you will need to scale the feedback signal proportional to the speed control needed. This form of loop tension control is widely used in paper mills and steel rolling mills.

Tech7

the denim plant around here uses dancer rollers with rheostats to control tension on the dye range. USing DC motors, the armature is constant voltage & the + field lead breaks through the stat... a 'tried & true' method for tension control.

If this is for individual threads, then try to scale down that theory to match your application.

I cannot connect the spindle to a motor directly (that would be the best) because actually there are 2000 bobbins on the creel which makes it very expensive. On the creel there are 6 rows, and I am controlling every row not every bobbin. There is a reference bobbin on each row.
I cannot use dancer rollers here because of the space concerns.
And about the oscillation problems, I will have a force on the spring all the time, may that can stop some oscillations or will it?

Originally posted by cakkostepen:

And about the oscillation problems, I will have a force on the spring all the time, may that can stop some oscillations or will it?

Click to expand...

No, it won't. It may tend to influence the oscillation frequency but not the fact that it's there. Think of a weight hanging from a spring. If the weight is heavy enough and the displacement is small enough there is always a force on the spring. But it will still oscillate.
So it sounds like there are roughly 300 bobbins per row. Do you have a separate spring and brake for each bobbin all pulled together by a common motor? Or do you have one motor and one spring pushing on a bar or other assembly that rides on each bobbin? I'm just trying to understand they type of system that you have that can be controlled by one motor and spring through a friction brake but can't be controlled directly by one motor.

Keith

this is the system. As you pull from point 2 the brakes on point 1 makes bobbin to move harder by increasing the friction. The bar that the spring is connected goes through all the creel. So when you pull from point 2 all the springs that are connected to the bar are strecthed.

Well done with the picture. I understand now.

It doesn't look to me like the bar that the springs are attached to has any type of bearings or anythihg as it passes through the frame. I suspect that there is a pretty large amount of friction where the bar and frame meet. I also suspect that the level of friction is abbout as much as the sping force. This is good and bad. It is good in that the friction will remove energy from the system and tend to dampen oscillations. The bad thing is the friction in the system will require torque to overcome. That is torque that is not accounted for and that you believe is going to tension. Granted, the PID controller will ultimately add torque to overcome the friction. But you may have some problems as the bar transitions from static to dynamic friction and back again.

I would say try what you propose. It just may work. But if it doesn't, consider the position based system I mentioned above. That just might work for you, too.

Good luck.
Keith

I actualy know nothing about tensioners, but last week I saw a reel tension/braking setup that used a small PM DC motor. The two motor leads were connected to a rheostat that was used to adjust the force required to turn the armature. Kinda a "Rube Goldberg" setup.. but it seemed to be working.

Does anyone think you could filter the oscillations out, or calculate for them and account? Granted, I'm young, a student, and don't know alot about physics, but from what I've done in physics, springs are fairly predictable and good systems to model.

CroCop-

If cakkostepen is capable of defining the differential equations for his system and performing the LaPlace and z-transforms in those equations he could implement a digital filter that could account for the system dynamics. That is certainly a possibility. And if all he was worried about was motor position then he could do what he needs with a PID filter.

However, in his case he has a load cell based tension loop that is defining the torque command. So the torque command we would typically use to damp the motor oscillations based on motor position/velocity is actually based on the feedback from a device that is coupled through a pliable product. So, if I interpret this right, cakkostepen has a two pole system with very little plant side damping (the motor/spring/bar system) connected the a second two pole system (the product roll/yarn/loadcell/brake system). And I don't think any of the poles inherently cancel. That's why I don't think he can get away with using a single PID. He doesn't have enough controller side gains to account for all his system poles. He may be able to do it with either a custom digital filter or a cascaded PID depending on how things get set up.

Then again there may be much more natural damping in the plant model than I think, which would at least give him a shot at controlling the thing if he doesn't need to add too much integral gain to counteract friction (which just adds another pole to the system).

Keith