Resonant frequency eliminator 2

[Masbibe]

Hi All,

I am interesting does someone have some nice and simple literature about resonant freguency eliminator.

Usualy in drives manuals there are nothing special about this. There are no further information about tuning of this filter.

Does anyone have some positive experience about using this filtar.


Milovan Milosevic

 

[electricpete]

That would be blocking the speed range where the machine is resonant, so that the machine accelerates or decelerates quickly through this speed range. If you didn't do this and spent a long period of time operating in resonant speed range, the vibration could build to high vibration levels and remain there as long as you're in that speed range... generally not a good thing for the machine.


=====================================
(2B)+(2B)' ?

 

[Masbibe]

electricpete,

Thank you very much for your quick response.

But it seems to me that job of RFE is not so simple just go quickly through resonant area.

ABB drives has this RFE possibility and it is written that it is 2nd order filtar and there are next few parameters regarding RFE:
- frequency of pole
- damping of pole
- frequency of zero
- damping of zero

I am not so familiar with automation theory especially with zeros and poles of system. So I would be interesting to read some explanation simple as it possible what this filter do (from physical aspect) and what are method to tune this parameters.







Milovan Milosevic

 

[electricpete]

But it seems to me that job of RFE is not so simple just go quickly through resonant area.

You are probably right. I have never heard the term "resonant frequency eliminator" before, so I took a guess that it was resonant speed blocking function. But I can't reconcile that with the parameters you provided.... so I'm thinking I guessed wrong.



=====================================
(2B)+(2B)' ?

 

[electricpete]

My next guess would be that the filter you are referring to would be the feedback filter H(s) in a closed loop system, used to coordinate with the open loop transfer function G(s) to maintain stability as shown here:

http://en.wikipedia.org/wiki/Nyquist_stability_criterion

That's a long way from anything practical. I can tell you how to build a G(s) for given H(s) that provides stability from control theory. I have no idea how this concept is applied in a vfd.

=====================================
(2B)+(2B)' ?

 

[Masbibe]

electricpete,

Thanks again for your response.
I am on some commissioning and tomorrow should be my last day here.

On one DC drive I have some strange oscilation always with same period regardless of how much is the speed.
Period of this oscillation is about 300ms, and this period is not matching with motor rotation period or load rotation period so my guess is that this is some resonant frequency of the load.

For those who are more familiar with metal industry motor is driving ladle on Continuous Casting Machine.

So for me it is obvius that this oscilation are not coming from current controller. So only thing I should change is speed controler parameters. But because this oscillation period is too high I could not use clasical low pass filter to eliminate this oscillations.
So my idea is maybe to use this RFE (resonant frequency eliminator) filter. But I never used this filter before and I dont know what this parameters exactly means:
- frequency of pole
- damping of pole
- frequency of zero
- damping of zero

Also I read before that for using RFE could cause dangerous oscillation if you are using it and you are not familiar with automation theory.

So as it is written in the manual this fiter is 2nd order band stop filter and it eliminate area around some particular frequency, and for me it will be very helpfull if somebody can said to me transfer function of this filter (but in the form of using this parametars that I mentioned before) that I could test response of this filter in Matlab and to see how it realy work because I would not like to test it on real machine.

Also it will be very helpfull if somebody has experience with using this or some familiar resonant frequency eliminator.



Milovan Milosevic

 

[cswilson]

Milovan:

What you are asking about is usually known in feedback control circles as a "band-reject" (as you call it, "band-stop") or "notch" filter. In the most common implementation, the input to this filter is the error between desired speed (or position) and measured speed (or position). I am presuming you have some measurement device for speed or position here.

The typical form for this filter is that of a "bi-quadratic" -- 2nd order numerator and denominator terms. The typical (analog) transfer function is:

s^2 + 2*zetaz*wz*s + wz^2
N(s) = K*------------------------------
s^2 + 2*zetap*wp*s + wp^2

where:

wz is the frequency of the zeros (rad/sec)
zetaz is the damping ratio of the zeros
wp is the frequency of the poles (rad/sec)
zetap is the damping ratio of the poles

While some people advocate setting the frequency of the zeros (there is a pair of zeros at the same frequency in this filter) to the same frequency as the mechanical resonance poles, I believe this is a mistake -- pole/zero cancellation is unwise in my view.

I generally recommend putting the filter zeros at about 90% of the mechanical resonance frequency, lightly damped (zeta of about 0.2). The phase lead from thes zeros is what aids in the stability of the system.

I usually like to put the poles at about 150% of the resonant frequency, heavily damped (zeta of about 0.8). The purpose of the poles is to keep the high-frequency gain of the filter from getting too large.

The "K" gain term should be set so that the "DC gain" of the filter (when s=0) is 1.0. Often, it can be "absorbed" into another gain term.

The warning you received is correct. If you don't know what you are doing with these filters, you can create a dangerous, violent response. Make sure you handle your units (rad/sec vs Hz) properly, and make sure the conversion to digital form is handled properly.

With such a low resonant frequency,a high digital sampling rate can actually be a disadvantage.

Curt Wilson
Delta Tau Data Systems

 

[Masbibe]

Thank you very much cswilson, this will be very helpfull for starting using RFE filter in the future.



Milovan Milosevic

 

[electricpete]

Is this filter applied in the forward path (like G(s) or portion of G(s) in the link) or in the negative feedback path (like H(s))?

Out of curiosity, what kind of parameter is sensed for feedback control (speed? current?)

=====================================
(2B)+(2B)' ?

 

[Masbibe]

electricpete,

This filter is applied on speed error signal.

Milovan Milosevic

 

[controlsdude]

If at all possible I would run this thru MathLab to see if the parameters your putting in will give you stability. My experiences with these calculations was that what I thought would work usually made it unstable.

Does this manufacturer offer any type of math model to crunch your numbers before hand to see what your doing is stable?

 

[Masbibe]

I can only look in Matlab what is behavior of band stop filter on some input signal but what will be response of your real system you can not know.

I do some simulation in Matlab and result for elimination some particular frequency are amazing, but you must know exactly which is frequency you want to eliminate if you make mistake here results are not any more so superior.

I didnt had time to try this on real machine so it will wait until some next time I deal with some oscilation problem.

Also in my case oscilations I want to eliminate are with quite small frequency around 3 Hz so right now I am afraid to turn on this filter and to go out from this site, because maybe this filter because of this small frequency will have strong impact during acc or dec ramp or during some load changes.

It will be nice to hear also somebody else experience using this or some similar band reject filter.





Milovan Milosevic