4 Run Balance Method

[dynaman]

Hi guys,

Have been performing the 4 run method for a while now. Works well enough with a vibration meter. The balance weight location works fantastic. However I find that the weight required according to the method is often incorrect (typically lower than required). What am I doing wrong?

Is there a way to improve on this?

thanks

Dynaman

 

[GregLocock]

Impossible to say given the information supplied. How big is the error? what software are you using?

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[dynaman]

Hi Greg, the error is about about 60% lower than required. I'm not using software, just doing the vector math by hand.

cheers

M.

 

[Strong]

Try doing the 4-run method graphically with the intersection circles. Perhaps there is a math error rather than a physical error in weight size or placement.

Walt

 

[GregLocock]

Since you are doing it mathematically try adding noise (ie random phase, say 10% amplitude) to your measurements and seeing if that tends to pull the results in or out with your particular procedure.

Use one of the on line calculators to check your working, it is much easier to see what is going on graphically.

I agree 60% is a huge error.



Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[dynaman]

I tried the method using intersecting circles. I can get the circles to intersect at a point really well. Its just that the weight is off.

 

[Tmoose]

Is this a rotor with the possibility of some bowing in response to the weight? like a center hung fan on relatively long shaft or some lengthy rotor operating at 50% of first bending critical or above?

 

[dynaman]

It is an overhung rotor, shaft length and diameter is considerable but don't believe it is bowing. I check the vibration results using frequency analysis and I'm measuring the imbalance signal (no hidden effects that I can see). I have tried the process on many different fans and gave the same results.

 

[Tmoose]

what instrument(s) do you have? Do you have a photocell, etc for sychronous measurements?

I'd do some "bump" test with the sensor radially on the shaft out by the fan wheel, impacting out by the wheel, and some impacting other locations.
The running 1st bending resonance is possibly ( I first wrote likely, then changed my mind) to be somewhat higher than a static test if the fan wheel back plate is stiff enuff to provides some gyroscopic stiffening.
Is the fan isolation mounted?
Do you take vibration measurements elsewhere on the fan, structure and foundation?

 

[electricpete]

I'm surprised the angle comes out right and not the magnitude. The angle is the trickier part of that calculation.

Is the correction weight attached at same radius as the trial weights were?
Is it possible you have a weighing error for the trial weight or correction weight... that would give correct angle but wrong magnitude.

Sorry to suggest such simple things... just mentioned in the unlikely event you haven't looked at these already.


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

 

[dynaman]

Hi all. Thanks for your feedback. I went back to the lab (my garage)and tested Greg's hypothesis. I superimposed white noise over the imbalance signal using my oscilloscope software and ran the 4 point method again (graphically). The phase solution was the same however the correction weight was way worse now, in factor double the error. Hmmmm.... not sure why exactly but it had a dramatic effect. I then re-ran the solution again, this time applied filtering. The solution gave the same phase angle but now the correction weight was pretty close to the actual solution. So it seems that noise can really throw things out.

For those who asked, I'm using frequency analysis (FFT) to measure the peak amplitude at RPM. That way I can concentrate on the imbalance signal only. There were some other noise elements at higher frequencies but their magnitude was quite low relative to the imbalance. I also thought about coupling effects at RPM because its an overhung rotor........maybe.

So I think the 4 point method is fine its just how I measured the data.

 

[GregLocock]

I thought up some rules of thumb for the 4 plane method - waddayathink?

1) use a triggered method instead unless you love puzzles

If we call the 4 runs U, T1 T2 and T3, where T1 is the trial weight at position 0 degrees, T2 is at 120 degrees, T3 at 240 degrees:

2) one or two of the runs T1-3 needs to be less than U. if not, use a smaller trial mass

3) one or two of the runs T1-3 needs to be greater than U, if not then it is physically impossible (with 120 spacing) so you must have a noisy measurement. Use a bigger trial mass.

4) if any of the T runs is zero then that run is the solution



Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[dynaman]

What triggered method are you referring to Greg?

 

[GregLocock]

If you use a trigger and then measure your vibration's phase relative to that (easy to do if you a tracking filter and scope or two channel FFT) then you only need 1 trial mass run, not 3 .



Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[dynaman]

How do you use a 2 channel FFT to calculate phase shift?

 

[electricpete]

Two channel analyser can tell you relative phase of different points on the machine at a given time. I don't think it would be much use for balancing since the things we want to compare occur at different times. You need something some kind of pickup that is going to look at something on the rotating shaft and give you a once per revolution pulse can be used as reference for phase measurement.

2) one or two of the runs T1-3 needs to be less than U. if not, use a smaller trial mass

3) one or two of the runs T1-3 needs to be greater than U, if not then it is physically impossible (with 120 spacing) so you must have a noisy measurement. Use a bigger trial mass.

What on earth are you talking about?

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

 

[electricpete]

Hi all. Thanks for your feedback. I went back to the lab (my garage)and tested Greg's hypothesis. I superimposed white noise over the imbalance signal using my oscilloscope software and ran the 4 point method again (graphically). The phase solution was the same however the correction weight was way worse now, in factor double the error. Hmmmm.... not sure why exactly but it had a dramatic effect. I then re-ran the solution again, this time applied filtering. The solution gave the same phase angle but now the correction weight was pretty close to the actual solution. So it seems that noise can really throw things out.

I'm not sure what he purpose was of adding noise. It was apparently to test a hypothesis which as I understood it was that random measurement error would affect the magnitude of calculated correction but not angle of the calculated correction. I see no reason for that to be the case (anyone want to explain). The result that was reported was that filtering the signal gave good results on magnitude and phase. What does filtering do: get rid of other frequency components. So now we know the likely cause is that you had other frequency components present which were increasing your overall magnitude. Filtering gave the running speed amplitude. Why the original error (From extra frequencies) affected only the magnitude and not the angle of calculated correction is still somewhat of a mystery to me.

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

 

[electricpete]

Last comment - I want to clarify I was not trying to downplay Greg's contribution (sorry if it came accross that way). His general idea was that he suspected measurement error of some type and that's what it turned out to be (error due to extra frequencies present). My focus is trying to learn the lesson of why it acted the way it did (why did it affect only magnitude and not phase of calculated correction)

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

 

[dynaman]

Like I said, I couldn't explain it but it worked. The measurement for RPM peak amplitude was way more stable without noise present + filtering. So with noise the peak would shift erratically so you had to question where exactly was the amplitude peak. Sometimes it was high and sometimes lower so I got erratic results... but for some reason the phase angle didn't change (maybe +/-2 degree at most).

 

[GregLocock]

ep - you are wrong. A 2 channel FFT analyser makes a great balancing analyser.


Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?