Powerflex 700 Torque Control Tuning

[MaxK]

I filtered the "armature amps on the main motor" curve and scaled the curves #45 graph.

"motor current" orange curve confuses me:
Graph #5 has spikes.
Graph #8 does not show any reaction to the change of other parameters.
Graph #45 has spikes again.
I don't have the slightest idea how to interpret this curve (dependence on other parameters). Perhaps an increase in resolution will clarify.

Take a look at "armature amps on the main motor" green curve and “I.Feedback from the main drive” purple curve.
Graph "Oscillation" - the curves are quite pronounced in antiphase.
Graph " Stabilization" - as the antiphase (high frequency) decreases, the system stabilizes.
I am not saying that the main drive is the cause of the oscillations. Actually I don't think at all that there is "the only" source of oscillation - I consider that interaction of few elements of the system generates resonance.

I would suggest changing the main drive (speed control drive) PID settings, for example reduce the gain (Kc) a little bit and see if the system behavior changes.

 

[Laminar]

I filtered the "armature amps on the main motor" curve and scaled the curves #45 graph.

Wow! Thank you for putting this together!

"motor current" orange curve confuses me:
Graph #5 has spikes.
Graph #8 does not show any reaction to the change of other parameters.
Graph #45 has spikes again.
I don't have the slightest idea how to interpret this curve (dependence on other parameters). Perhaps an increase in resolution will clarify.

Yes, I believe it is purely a resolution/rounding issue. The motor current is coming from the drive via a datalink and only has resolution to the 0.1A, though actual control is finer than that. I assume the spikes are when the value is fluctuating around X.X5 and the logger ends up continually rounding up or down to the nearest tenth.

edit: Now that I'm looking at my more recent image below, there are clearly steadier areas and more active areas, indicating some kind of control attempt from the drive. It is still wild to me that the Torque Reg Kp and Ki parameters make no impact on this behavior.

Take a look at "armature amps on the main motor" green curve and “I.Feedback from the main drive” purple curve.
Graph "Oscillation" - the curves are quite pronounced in antiphase.
Graph " Stabilization" - as the antiphase (high frequency) decreases, the system stabilizes.
I am not saying that the main drive is the cause of the oscillations. Actually I don't think at all that there is "the only" source of oscillation - I consider that interaction of few elements of the system generates resonance.

I would suggest changing the main drive (speed control drive) PID settings, for example reduce the gain (Kc) a little bit and see if the system behavior changes.

I gave this a shot. The "spd reg Kp" (par 87) is set to 10.88% in the main drive. I reduced it to 6.0%. I noticed no specific improvement in behavior. In the graph below at the red line, I returned the value to 10.88%. I didn't see any change in oscillation behavior, but overall stability of the main line speed did seem to improve after returning the parameter to its original value.

 

[MaxK]

I was speaking about
Kc * (1 + Ki / s + Kd * s)
I suppose your PID representation is
Kp + Ki / s + Kd * s

Anyway, I please you to publish the graphs of the system response and "mark" the PID controller settings changes "points"

And one more. If you "play" with main drive PID settings, I please you to show with main drive curves (armature amps on the main motor)

 

[kamenges]

Originally posted by Laminar:

It is still wild to me that the Torque Reg Kp and Ki parameters make no impact on this behavior.

Why should that surprise you? Consider what you are controlling with the torque regulator. You are giving the drive a fairly static torque reference. Just because the motor SPEED is oscillation doesn't mean the motor generated TORQUE is changing at all. The torque regulator doesn't really care at all what the motor shaft is doing. All it cares is that the motor generated torque is held at the command level. In that context changes in the torque regulator gains SHOULDN'T have an effect on the system. You have a mass/spring/mass system and the motor torque is a part of the spring. The drive doesn't need to change anything to contribute to the oscillation.

Keith

 

[Laminar]

Why should that surprise you?

Because I've always been able to make control worse. To divide the Kp and Ki by 100 or multiply by 100 and see no effect, better OR worse is what gets me.

 

[Laminar]

Anyway, I please you to publish the graphs of the system response and "mark" the PID controller settings changes "points"

And one more. If you "play" with main drive PID settings, I please you to show with main drive curves (armature amps on the main motor)

Apologies, I re-added the arm amps in green.

Within the highlighted box, Kp was reduced from 10.88% to 6.0%.

 

[Laminar]

Today's a weird day - things are different. Here's an instance where they were running at the danger speed - it looks like the system wanted to start oscillating, but then decided not to.

And something I've never seen (pulled from Historian, so scaling and smoothing are a little different). The oscillations are coming and going at a steady speed.

I have made no changes to the controls or program vs. the previous charts.

 

[drbitboy]

The historian image is probably the result of sample aliasing, i.e. the oscillations were (mis-)behaving as they usually do at ~2Hz, and historian was sampling at some near multiple of 2Hz, so it stored a different point, in phase, of the sine curve for each sample.

e.g. if I sample a sine curve at phase intervals of δ, e.g. phase=0,π,2π(≡0),3π(≡1π),etc, I would get a flat line at zero. If I sample at intervals of [(1+δ) × π], where δ is a small number, then I would get something like that Historian plot.

 

[Laminar]

Yep, you're right - I caught an instance live and compared it side-by-side with Historian and Historian is lying to me.

 

[MaxK]

A little curves analysis.
1. Please pay attention to red and blue curves #46 graph "Oscillation" and graph " Stabilization" BEFORE 7:54:20. It is clearly seen that the red curve is ahead of the blue one.
2. Now pay attention to red and blue curves #46 graph "Stabilization" after 7:54:20 and graph attached below (#52 scaled curves). Such a noticeable shift of the curves is not observed.

So Now notice the purple curve.
In the first case, the curve oscillates with the oscillation frequency of the blue and red curves.
In the second case, no such oscillation is observed.

Assumptions:
1. The system has mechanical (?) self-oscillations (relatively small amplitude)
1.1 It is practically impossible to suppress these self-oscillations (it is possible that the tension drive is the cause of the self-oscillations, but until we see the interpreted current values, it is impossible to assert anything).
2. The PID controller of the main drive brings the system into resonance
2.1 I believe that the settings can bring the PID controller out of the resonant frequency zone.
I would suggest "playing around" with the PID controller settings +/-25% and evaluate the result (my guess: increase Kp, decrease Ki).

Request:
Please explain what the parameter "I.Feedback" means, how it is measured (I would like to understand the physical relationship between “measured web tension”, “tension control motor RPM via encoder feedback” and “I.Feedback”).
Is the PID controller of the main drive controlling the frequency?

 

[joseph_e2]

It may not be helpful or relevant, but...

I recently posted about an issue we're having with a VFD driving a flywheel machine, along with trending data. After some back and forth with Rockwell and our local support, we learned not to trust that the trending was exactly accurate as to what happened in what order. We normally saw the drive output frequency drop to 0Hz, then the Ready/fault relay drop out, then the bus would swell. Occasionally, though, we saw the ready/fault relay drop out one sample before the frequency dropped out. They explained that away saying that the trending isn't always perfectly reliable when you're trying to record cause vs effect.

So...if you're getting consistent results that the output frequency starts "going crazy" before the RPM feedback does, then that might be a clue...or a fluke caused by limitations in the trending. Especially when you're trying to isolate something that happens very very fast.

 

[Laminar]

Request:
Please explain what the parameter "I.Feedback" means, how it is measured (I would like to understand the physical relationship between “measured web tension”, “tension control motor RPM via encoder feedback” and “I.Feedback”).
Is the PID controller of the main drive controlling the frequency?

I.Feedback is the motor speed available in the PLC. In the case of this drive, it is also encoder feedback.

Thank you for all of your analysis and work - I will adjust the main drive tuning parameters and report back!

 

[MaxK]

Another thought to test the resonant theory: set a dead band in the PID controller of the main drive

 

[drbitboy]

Another thought to test the resonant theory: set a dead band in the PID controller of the main drive

^ ahh of course.

The solution was obvious, but @MaxK is the only one with his thinking cap one!

 

[Peter Nachtwey]

^ ahh of course.

The solution was obvious, but @MaxK is the only one with his thinking cap one!

I thought the resistor/damping in the RL circuit was the solution.


I don't see were dead bands solve anything except for those systems that only have on-off control.