Dancer Systems

[OkiePC]

Sending the dancer signal directly to the drive will work, but as the roll grows, the position of the dancer will change too. It should start out at the lower (fuller) part of the dancer travel and end up at the higher (emptier) end when the roll is finished.

And it will react AFTER the line speed changes, so it will "top out" or go near empty when the line stops, and must wait for the dancer to fill up some when the line re-starts before it will resume.

I am shocked that your calculated winder speed produced so much instability. If you base it on line speed and then trim it with the ratio "learned" or calculated, then trim it with the dancer, it should be very stable.

I had an old honeywell winder program that was an offshoot of just sending the dancer to the drive directly and it worked quite well, but the dancer would get out of position when the line stopped or started. It took the dancer signal and changed the slope of the linear device to put the full range of the final value in a narrow band of the dancer's physical travel range, multiplied it by line speed, and then sent that value to the drive. It worked like a charm, causing the dancer to only "range" a very small amount as the winder diameter increased.

Just watching the line run as you have it, you can monitor the line speed versus winder speed at core size and max size of the wound roll. That will give you the upper and lower limits of your range of ideal ratios.

Then you can use a PID to "find" this ratio with very low gain settings. That eliminates keeping track of how much stuff is currently in the roll, and will take care of variations in product thickness and elasticity. Use the full range of PID CV then, external to the PID, rescale it between your known upper and lower limits (plus a little padding on each end of the scale)

Multiply the ratio by the line speed and send it to the drive.

This works well for me on systems where I can not tell when a new roll is started and therefore can't track the length of the winding roll.

I use a very low P-gain setting and an even lower I gain, 0 derivative. When a new roll begins, the winder will run too slow for the first one or two seconds and then ramp up to the line speed as the PID nails down the gear ratio between the two. If the gains are set too high, there will be oscillations. You have to compromise between responsiveness and stability, but there is usually a happy medium. If not, integrate the wound length into the formula, and give the PID less influence, or set the PID output to the known starting value when the winder begins a new roll.

Whatever you do, an accurate line speed signal should be integrated into the logic especially if stops and starts or frequent speed changes occur.

Hope this helps.
Paul

 

[scottmurphy]

I am shocked that your calculated winder speed produced so much instability. If you base it on line speed and then trim it with the ratio "learned" or calculated, then trim it with the dancer, it should be very stable.

I need to go back to this method, but the reason I went away from it. was beacuse I couldn't get the accurate speed control that I wanted, and I put this down to not having feedback from the motor.

Whatever you do, an accurate line speed signal should be integrated into the logic especially if stops and starts or frequent speed changes occur.

The Line speed signal is fairly accurate.

The other thing that I am probably focusing too much on, is trying to maintain tension ( position )at zero speed, as this is how it functions at present, and the tension is needed, especially if the first print unit is needed.

When I did have this working ok, it was the acc/dec that was instable, this was using the calculations based on reel size etc.

I may get another chance this afternoon to try again.

 

[OkiePC]

So, your dancer position also controls tension? Mechanically?

Okay, at zero speed, you may be able to switch to torque mode, but that depends on the drive. Some drives don't allow switching on the fly.

But you may be able to cheat, and use a "2nd current limit" digital input to select a lower current setting (when the line is stopped), and then you can send a low speed reference to keep the motor stalled against the dancer. It will take experimentation to get the values right.

Paul

 

[OkiePC]

I just scanned over your previous posts on this thread. You have DC drives, so the torque mode is more likely to be available than the 2nd current limit for zero speed tension control.

Back to the theory of control...You have a variety of thicknesses, so I recommend that you don't make the operator responsible for entering them.

Use a PID based ratio detector. I have several of these running on AB PLC5s and a few SLCs too. I have stock hanging freely in a loop between the ReWinder and the previous line station. My loops are 13" tall and read with an analog light curtain. The loop position stays within 1" of the setpoint at all times, including fast acceleration and deceleration, so I can tell you, that this sytem works really well. Here it is in a nutshell:

Find out what the minimum and maximum ratio between the line speed signal and the winder speed signal can be.

Example:
RW_Smallest_Core : Line_Speed = 4.1
RW_Largest_OD : Line_Speed = 0.7

Create a PID instruction with the dancer position as the PV. Rescale it if necessary to use the full range of the PID instruction. Put the SP at the value you want the dancer to stay at. The CV is allowed full raw units and no limiting also.

The PID must be disabled, frozen, switched to manual, whatever you prefer to call it, when the line stops. Be careful to ensure you use the correct method for your hardware. The reason is: If the line is stopped, the winder speed reference will be at or near zero, so the dancer position may not be able to change. If the PID is in AUTO, it will saturate at one extreme in a futile attempt to get the dancer signal to match the setpoint.

Make sure you set it up correctly for direct or reverse acting.

Next, take the raw CV and rescale it to another floating point value that can range from a little bit less than the minimum calculated ratio, to a little bit more than the maximum required ratio. (0.5 and 4.5 for our example)

So if your PID CV goes from 0-4095, and you want 0.5 to 4.5, the RW_Ratio = 0.5 + PID1.CV*(4/4095)

Now multiply your line speed by this ratio and send it to the drive. Set the PID I and D parameters to 0, and start out with a moderate P gain value.

Monitor your RW_Ratio variable, and adjust the P gain to get the rate of change in the RW_Ratio that you want. When it is responsive enough and not oscillating, add a very low I gain value. Now any remaining steady state position error should decrease. I am not a PID expert, but there are many here who can give you better recommendations on PID tuning and exactly how to program it with your PLC.

If you know when a new reel starts, you can preload the PID CV a high value, and set the gains even lower.

All you want from the PID is to slowly dial in the exact ratio.

Your dancer should ramp into position as the line runs, smoothly like a plot of a classic overdamped PID process variable. and stay there even when the line speed changes or stops.

If accel and decel cause disturbances in the dancer position, it usually means the accel and decel settings of the feeding line drive doesn't match well with the RW drive's settings. Suppose during decel, you observe that the dancer gets fuller. That means the RW drive is decelerating faster than the line. So reduce the drive decel rate a little and try it again.

You can tell which way to adjust them by observation and get them pretty close after the basic running speed control is steady. Each successive (downstream) drive may benefit from a slightly faster accel/decel time, but never slower. On digital drives, I set them all the same and tweak them if actual perforamce varies.

And, as mentioned before, you want to get the line speed from the nearest upstream driven part of the line. If that's a surface tach, there may be processing lag time to consider. Sometimes it actually works better to work with the line commanded speed reference, instead of a feedback signal due to these potential delays.

Basically, you have to build in software what that old electronics package was doing in analog electronics in order to match it's "coffee coaster" stable performance.

Good Luck

 

[scottmurphy]

So, your dancer position also controls tension? Mechanically?

Yes, the amount of tension against the web is controlled pneumatically. But this is only true when the dancer is not bottomed / topped out.

Okay, at zero speed, you may be able to switch to torque mode, but that depends on the drive. Some drives don't allow switching on the fly.

I have been trying to find out how to use the current control mode on the drive. Have not had much joy as yet.

What I was doing, to maintain tension at zero speed, was to use a 'holding ref', that was dependant on reel size. Obviously, it is dependant on what the RW is doing, which I have no control over at present.
This did work, but then when I went to acc/dec there was no consistency.

Just seen your latest post, will have read through that.

 

[scottmurphy]

I apologise if I am not clear, but I am trying to understand you ratio method. I am in brain saturation at the moment, too many things to do, and trying to get this done also.

Am also annoyed that my calculated method did not work as I had hoped. This is more so from the lack of feedback on the motor.

Also, bear in mind that I am working on UW, even still the principal is the same.

Find out what the minimum and maximum ratio between the line speed signal and the winder speed signal can be.

Example:
RW_Smallest_Core : Line_Speed = 4.1
RW_Largest_OD : Line_Speed = 0.7

For these ratios, I am slightly confused, do you mean the ID & OD core sizes in a measured unit, with the line speed referenced in the same measurement unit? Or do you mean the spindle ref required for the different diameters, based on line speed. As the line speed may vary, this ratio must be the same.

The PID must be disabled, frozen, switched to manual, whatever you prefer to call it, when the line stops

This is easy enough to acheive with the system that I am using.

So if your PID CV goes from 0-4095, and you want 0.5 to 4.5, the RW_Ratio = 0.5 + PID1.CV*(4/4095)
Now multiply your line speed by this ratio and send it to the drive

This will not work at zero speed though, I must need to enable this method when web is moving? That means I will have to use the current method that I had of holding tension at zero speed, on initial startup?

 

[OkiePC]

For these ratios, I am slightly confused, do you mean the ID & OD core sizes in a measured unit, with the line speed referenced in the same measurement unit? Or do you mean the spindle ref required for the different diameters, based on line speed. As the line speed may vary, this ratio must be the same.

The low number will be the ratio required to match line speed with a full reel. The center driven rewinder will have to turn much slower than the line to match it's surface speed. You can use the ratio in engineering units as calculated, or just measure the ratio in raw plc units monitoring the machine using the code you have working. Just add a line to divide them for you and store the result in a spare floating point address. Then, even with a signal going directly from the dancer to the drive, you will see the calculated ratio ramp down from it's maximum value to it's minimum value as the reel grows.

Next add a little cushion to what you have monitored, and use that as the limits of your rescaled PID CV. Don't clamp the PID action, rescale it like in my last post.

This will not work at zero speed though, I must need to enable this method when web is moving? That means I will have to use the current method that I had of holding tension at zero speed, on initial startup?

Yes. If you have a working method of maintaining dancer tension at zero speed, keep it, and switch between the two methods when the line stops.

At zero speed, I would multiply my dancer position error by a gain factor and another factor based on reel circumference, and send that to the drive speed command. Also, change the torque limit to best match the reel mass if possible. You need more torque for a heavier roll. I don't know if reversing when the dancer is too tight is allowed in your application, but any disturbance in dancer position incurred during deceleration could be eliminated that way.

Then you can adjust your gain factor until zero speed performace is optimized.