High-Inertia Load VFD - Dynamic Brake Overheat!

[marv_tech]

Hey All,


I've got a Vacon x4 VFD running in open-vector mode. It is spinning a heavy pin-mill wheel. It gets frequency reference from a PID running in a SLC5.


The PID is not tuned well (but unfortunately I cannot change it). As a result, the drive is constantly hunting, DC bus voltages hit their limit due to decel, which kicks in the dynamic brake, which overheats and eventually faults.


Turning off the dynamic brake leads to (rapid) DC overvoltage faults. Making my decel time longer maybe helps but I still get nuisance faults.



TL;DR: Is there any way for a VFD to allow a high-inertia load to decelerate naturally rather than trying to "absorb" that load?

 

[drbitboy]

LN₂ cooling?

 

[GaryS]

I would change the vfd to a line regenerative type
that way you would have 100% braking power 100% of the time
you could even push it 150% for some time
if you really have a High-inertia load them make sure the vfd is rated for heavy duty
I think the Yaskawa matrix drive would be a good choice

 

[marv_tech]

LN₂ cooling?

Cost prohibitive for my application, the only nitro I get is to purge the cabinet

I would change the vfd to a line regenerative type
that way you would have 100% braking power 100% of the time
you could even push it 150% for some time
if you really have a High-inertia load them make sure the vfd is rated for heavy duty
I think the Yaskawa matrix drive would be a good choice

That is an interesting idea, and costing it out might be enough to convince the powers-that-be to allow me to tweak the PID instead. I was considering a big dynamic brake resistor (which should at least buy me runtime) as a last ditch effort.

 

[sbaum]

I'm not familiar with the Vacon but many VFDs will have a DC bus regulation mode that automatically extends the decel time when the bus voltage rises. Typically need to remove the dynamic brake in that mode.

However in your case it sounds like a regenerative drive would be best.

 

[MaxK]

costing it out might be enough to convince the powers-that-be to allow me to tweak the PID instead.

I suggest starting by learning the process:
Perhaps the PID controller is poorly tuned and causes unnecessary decelerations - in this case, it makes sense to tune the controller or soften the deceleration imposed by the PID controller.
If the need for active decelerations is dictated by the process, then you need to look for a way to dissipate energy.

Do you have the opportunity to feed us curves (PV, CO, current, voltage etc)?

 

[rdrast]

Bonitron Line Regen Controller


These things are amazing, and relatively inexpensive.

 

[GaryS]

As I said
here's what you need
https://www.yaskawa.com/products/dr...general-purpose/u1000-industrial-matrix-drive

the U1000 Matrix drive has no DC buss it's direct line to motor full regen
you could then tune the PID loop

just hook it up and go

 

[Peter Nachtwey]

This is a classic case of expecting more out a system that what iit is capable of. It probably worked just fine at a lower frequency but some thought they would try to increase production.

It is all about the energy! The pin wheel has K=(1/2)*I*ω^2 energy. You can see the energy is proportional to the frequency squared. All this energy is dissipated when the pin well is stopped. Changing the PID will not change how much energy that must be dissipated. Only reducing the frequency will reduce the amount of energy to be dissipated. Reducing the deceleration time requires that all that energy be dissipated in a short time and that any regenerative currents will be greater thus have a larger chance of saturating.

Tuning a PID that is in saturation is useless. Then in saturation the PID is basically not controlling anything. One of the things I like to do in my simulations is to make the system saturate so I can see how well the PID recovers.

REDUCE THE FRRQUENCY KENNETH!

 

[MaxK]

It is all about the energy! The pin wheel has K=(1/2)*I*ω^2 energy.

Right! However, in this case we are talking about energy dissipation per time (power). I.e.

(K0-K1) / Δt = ½ I * (ω0^2 – ω1^2) / Δt

I.e. it's needed to reduce the change in frequency per time (ω0^2 – ω1^2) / Δt. Thus, if at the current settings the PID controller produces sharp CO, and even more so oscillatory changes, while the process requirements allow for "smooth" control, then retuning the PID controller can reduce the specific power dissipation.

REDUCE THE FRRQUENCY KENNETH!

I'm afraid that the process requires exactly the energy (frequency) that is being used now and OP is unable to change this

 

[ndzied1]

REDUCE THE FRRQUENCY KENNETH!

Nice one

 

[lfe]

Yours is not going to have more inertia than a train at 300 km/h.

You need a VFD with regenerative braking.

 

[marv_tech]

I suggest starting by learning the process:
Perhaps the PID controller is poorly tuned and causes unnecessary decelerations - in this case, it makes sense to tune the controller or soften the deceleration imposed by the PID controller.
If the need for active decelerations is dictated by the process, then you need to look for a way to dissipate energy.

Do you have the opportunity to feed us curves (PV, CO, current, voltage etc)?

The PID controller is definitely poorly tuned and causing the unecesssary decels. It would make perfect sense to tune the controller, I've just been instructed not to unfortunately.


The curve I have been working with has been PID output vs feedback loop input. Unfortunately I don't have the facilities to plot the DC bus voltage and the load is low enough that I'm not concerned with amperage. I know for sure that the problem is the dynamic brake - I can see it kicking on frequently and the VFD fault history spells it out loud and clear.



I've attached a 5 minute trend - blue line is the actual pinwheel speed, green line is the PID output command to the drive. Both are 14 bit integers (0-16k counts). You can see the PID swinging all the way to the limits trying to stay in control, which makes the drive constantly ramp up and decel.



Definitely some interesting ideas here. I was just hoping I was missing some VFD terminology and that coast-to-setpoint was a common control scheme that I was just missing.

 

[MaxK]

I've attached a 5 minute trend - blue line is the actual pinwheel speed, green line is the PID output command to the drive.

hmm... this is so bad (it's even worse than I expected)
Perhaps there is some reason for this?
Perhaps there are system operating modes in which such PID adjustment is justified? Acceleration/deceleration of the mill, interaction with other units...?
Is this the maximum chart time resolution?

 

[marv_tech]

hmm... this is so bad (it's even worse than I expected)
Perhaps there is some reason for this?
Perhaps there are system operating modes in which such PID adjustment is justified? Acceleration/deceleration of the mill, interaction with other units...?
Is this the maximum chart time resolution?

That is what I am wracking my brain for as well. Unfortunately the only history I can get is "the drive has always hunted but didn't always throw faults." On the plus side, this is basically a stand-alone system and is only operated when the mill is running in the steady state and the speed setpoint is not really changed during the run, so I can rule out some of those concerns.



I suspect the original system drive benefited from PID in some way that the newer drive doesn't (mid 90s vintage VFD/ASDs had some weird quirks) or maybe responded much slower. I actually get tighter speed control running without PID entirely but that necessitates keying the setpoint directly into the drive, so not a long term solution.



I actually have a clue - someone modified the top speed without documenting it about 4 years ago (as someone above mentioned in a comment.) With the way Vacon calculates ramp rates (using top speed and accel/decel times), this actually impacts the responsiveness of the drive. I am going to try with the original settings to see if the PID has better control but I don't have high hopes.


I can definitely get a longer capture, I'll upload a 30 minute window when I get a chance. Thanks all for the input and ideas.