Another Winder Question!

[Big_col]

Hi
I am currently writing software for a dual winder with dc drives in torque mode. I have used 1 pid loop to match the line speed (line speed is setpoint and tacho is the process value) and this works well. The torque when winding is fed back to the plc from load cells, then in the plc using a second pid loop to try to contol the tension. This dose not work. Am i write in saying the tension loop needs to be slow to react to changes so as to ride out errors? Dose the first loop have any part to play when it is winding or do we just use the second loop?

Thanks
Col

 

[allscott]

What kind of drive? Most new drives have winding macros in them which is the way to go versus using a PLC.

You cannot control speed and torque at the same time, something has got to give. In a winder you don't particularly care about speed, the drive will go to whatever speed it needs to to generate the desired torque.

 

[kamenges]

There are (at least) two things to keep in mind with torque mode loadcell feedback winders. The first point is that winders of this type typically have no control side damping. So it can be very hard to keep this type of winder from oscillating. The second point is that the transmission losses need to be a small percentage of the torque required to develop tension. If this is not the case it can be hard to maintain stability as the torque requirements not related to tension change so much.

However, as allscott said, generally torque mode winders don't require any particular speed information to work correctly. The exception would be if you are using velocity control with torque limiting.

Keith

 

[Big_col]

Thanks for the reply. The drives are mid 1980's analoge drives.kamenges, you hit the nail on the head. The problem i have is oscillation. The tension gose all then nothing and the pid loop dose not know if it is coming or going! The feed back changes by a lot very quickly. The old analog control still works on one of the winders and looking at the voltage out of the old card the signals are not stable but it dose work. So i need to dampen the load cell signal in the plc. Can this be done simply by increasing the intergral time in the pid loop so it slows the reaction to error? or is there a better way?
By the way i am using S7-314 PLC and FB41 pid loop.

Thanks
Colin

 

[zova]

May be it will be interesting for you.

 

[mangletech]

HI

Is this by chance an ATLAS/TITAN winder?

best regards

Neil B (ex Atlas/titan engineer)

 

[Peter Nachtwey]

These are master theses.

May be it will be interesting for you.

zova, can you explain those pdf files to us?

They are pretty much worthless unless you just need a hint as to what is required to model the system. The are written by students to impress their professors and the professor know nothing more than the students. Most of you that have a winder know more practical information.

The pdf files didn't really get into how to compute the gains given the model. In the past we have done some pole placement where we calculate the gains as a function of the characteristic equation of the system. This can be done symbolically. That is how I do it. Then one has the formulas for changing the gains and feed forwards as the inertia of winder changes. This can be done every loop if necessary.

I like the way the through around terms like H infinity, ooh.

All that **** in the 3 pdf files an it doesn't say how to change the gains as a function of the radius. Purist would say that you really don't know the radius of the rolls. I say that all you need to do is get close enough and the PID can easily conver for errors in the model.

Back to this problem.

The problem with running in torque mode is that the derivative term must be very large relative to what it is in speed mode. This is because the PLC must now do the velocity stabilization which is normally done by the drive in speed mode. The problem is that PLCs are not very good at this relative to drives or motion controllers. The derivative term is basically multiplying the error in velocity between the target and the actual by a gain. We all know how difficult it is to calculate velocities on a PLC with the aid of special hardware. A PLC PID does not have that 'special hardware'. If accurate velocity errors can't be calculated then you can't increase the derivative gain or time constant up as high as required to stabilize the winder. This is the problem. I know this to be true because I got my master degree from the school of hard knocks.

I would get a real motion controller or a drive that has the winder algorithm built in. A PLC is not up to the job on this. PLCs can work in voltage mode when dancers or idle pulleys are used.

Perhaps calculating the gains as a function of radius is in order for a Friday night quiz. We can start from where one of Keith's Friday night quizzes left off. However, I fear there would only be Pandiani and Keith playing. We can use the example from the CTM PID site. It models a DC motor.
http://www.engin.umich.edu/group/ctm/PID/PID.html

And the forum goes OH NO not again!!!??

 

[kamenges]

Originally posted by Peter Nachtwey:

The problem with running in torque mode is that the derivative term must be very large relative to what it is in speed mode.

You are typically left with a case where you have no derivative term at all unless to bring the velocity feedback into the plc and close the loop around that. As Peter said this is difficult to impossible due to the limitations in a plc. When using a PLC in a loadcell system the tension changes too quickly to calculate an effective processs derivative term. So in effect you are left with no damping. You are relying any damping in the physical system to keep you out of oscillation. This can be a dicey proposition.

We have done torque based driven winders. But you need to go heavy on the torque feed forward and gentle on the corrections. You basically run it open loop with just enough correction to keep the system on track. However, we only use this type of system if we can't generate or receive a valid master speed reference or if something acts on the web outside of the influence of the master speed reference. In all other cases we go with velocity mode center winders for the reasons Peter stated. We may still develop the velocity command out of the plc but the drive is operating in velocity mode and is closing it's own velocity loop.

Keith

BTW, it would be fun to play with a Friday Night Quiz but I am out of the running this weekend. Go get 'em, Pandiani!!

 

[Peter Nachtwey]

BTW, it would be fun to play with a Friday Night Quiz but I am out of the running this weekend. Go get 'em, Pandiani!!

I will wait until someone is interested. Pandiani is 10 hours ahead so he is sleeping.

 

[Big_col]

Thanks all for the replies. I have got it working and holding tension with no oscillation! Could not have done it without zova
pdf files


Thanks
Colin

 

[Peter Nachtwey]

So what did you do?

 

[Big_col]

Ok this is where i am told i have done it totaly wrong!!!(but it works)

I created 1 pid loop and used the line speed as the setpoint and the tacho as the process value, the output value to the drive and then tuned the loop. Then created a second pid loop and used the tension pot on the machine as the setopint and the load cell as the process value. allscott was to a point correct when he said you cannot control speed and tension at the same time, but my concern was if the web was to brake the winder would run off as fast as it could, you need to be able to stop that from happening so you do need to control tension but keep an eye on the speed too. so when winding i noticed the winding speed was allways less that the line speed due to the build up of the product. keeping the first PID loop enables i feed the output from the first PID loop to the output limit of the second PID loop so if it saw a loss of tension the first PID loop would limit the seconds output to the first's output hence limit the speed and prevent uncontrolled run off. Another problem i encounted was as you enable the second PID and went to tension control the motor lost tension(speed) as the second pid loop was initally giving a 0 value output and had to ramp up. To get round this I eneabled both PIDs first pid was allways running but the output was fed into the manual value of the second PID loop. When tension was not selected the second PID loop was in manual so what ever value the first loop was giving the second loops output would follow. When tension is selected the second pid loop is switched out of manual to pid control the output is allready at the line speed so no loss of motor speed can be seen. When tuning the load cells i found that they where very unstable and the P gain was only about 0.4 and the intergral was 1 second. The deriviteve was disabled.

Col

 

[kamenges]

Your first PID loop is doing what the drive would do on it's own in velocity mode. Does the drive you are using support variable torque limit? That may ultimately be an easier way to go. Just run the drive in velocity mode and run the ooutput of what is now your second PID loop into the torque limit slot.

Depending on your speed and take-up ratio (ratio of core diameter to largest roll diameter) you may still have a web break issue. It sounds like you are simply controlling to line speed with your first PID. This means the winder spindle will go up to the speed asociated with a minimum core before it starts to limit. If you have a 10:1 or 20:1 take-up ratio and you are going 300 or 400 MPM you will have a large roll spinning at a REALLY high speed. Your limit speed really should be related to the roll diameter. Be careful. If you lose control of a large roll at high speed it can take itself right out of the winder.

Keith

 

[Big_col]

You are correct kamenges, the first pid is just controlling velocity. On the machine there is a contact bar which dependat on where it is scales the line speed back the to the plc so acts as diameter sensing and limits the speed. Max speed of the machine is only 50m/m. The drive is an old analoge drive and can only work in torque mode without been hardware changed.

Col

 

[Peter Nachtwey]

I missed th poit about the tachometer instead of encoder

Your first PID loop is doing what the drive would do on it's own in velocity mode.

Tachometers provide velocity instead of position feed back so the PLC doesn't need to differentiate encoder positions.

The proportional gain in position mode does the same job as the integrator gain in velocity mode. Integrated velocity errors are position errors. The derivative gain in position mode does the same thing as the proportional gain in velocity mode.