I always supplement balancing software calculations with a vector diagram. It gives me some insight of the process. In particular, it provides the lag angle which can be obtained after trial weight (TW) run and is equal to the angle between TW position and T vibration vector.

Usually for speeds below the critical the angle is equal to 10-20 deg with the T vector lagging behind in reference to the direction of rotation. Recently I was surprised to find out that TW position was actually lagging behind by 20 deg. !!!!!

The balancing job went well and vibration came down to very low level. But the leading T vector is obviously weird.

Did anyone have similar experience?

Thanks,
David

Were you using mils, ips, or g's?

Mils.

I am not familiar with the "T" in reference to the original and correction weight positions. I am familiar with "TW" being trial weight position. Sorry for being so simple minded but could you show a drawing with these positions marked?

Can you describe specifically how you determined the response angle?

Did you happen to get any coastdown data?

Based on the limited info provided, I will take a stab at what is happening.

My guess -- You are balancing a fan using bearing cap readings. The bearing cap vibration is somewhat (or a lot) out of phase with the shaft vibration. Because of this, the textbook relationship of high spot to heavy spot doesn't necessarily apply. Next time, try taking shaft stick readings to compare with bearing cap data to see which one better fits the "normal" heavy spot (TW) to high spont (T) relationship.

I have seen similar strange behavior when balancing turbine generators using relative vibration. In general, I believe in using absolute shaft vibration for balancing turbines when possible. I never use it for fans although it might be a good idea for some cases. I have never observed the behavior on low speed equipment, but I could definitely see it happening.

Michael Titone

And lag should be dependant upon type of transducer and what you are doing with it. Plus is in the opposite direction of rotation.

Attached is a file with some clarification. There is nothing new there, although not everyone measures "heavy spot - vibration response" angle at the trial run stage. I find it very useful though.

I agree with Michael that one can not expect a text book classic relationship between heavy spot and vibration response since vibration measurement most of the time is indirect, namely, on the bearing. Because of that transmitted vibration may be different, at least phase wise.

This message has been edited. Last edited by: David_G,

Lag_angle.doc (36 Kb, 68 downloads)

You can do a successful balance without paying close attention to location of stationary keyphasor sensor in relation to vib sensor and without paying attention to keyphasor rotating target in relation to weight positions. (In this case there is an unknown constant phase angle difference introduced between phases measured on rotor and phase angles measured with vibration, but again it doesn’t affect the balance.)

However to develop a response lag angle you would need to consider those factors and also possibly lag angles associated with the instrument (or else check coastdown). Did you consider those things?

In theory, the phase angle of the High Spot (O) and the Heavy Spot should be equal (or at least pretty close) when you are measuring displacement and are well below the critical.

One assumption that you are making is that the heavy spot corresponds to the angular location of your trial weight. The actual location of the heavy spot is the vectorial sum of the trial weight and residual unbalance of your component. This may or may not account for your leading angle.

Steve

Is the drawing you have posted a "true" representation of the balance job in question?

Are you saying you have been "used" to seeing your "T" on the "right" side of the TW as opposed to this drawing's "T" being on the "left" side? Or am I still not understanding what is being said(asked)?

Interesting there are a lot of different perspectives. I think I understand where Ralph and Mike are coming from. I'm not sure I understood Steve's comments.

Dave - To expand on my earlier comments, let's say you had a given balance job with a given set of fixed keyphasor and sensor locations.

Now, let's say I perform 20 experiments in balancing that rotor. Each time I start with the same machine condition. Pick a random location on the rotor and call it "0" degrees for purposes of determining trial weight/correction weight angles. Perform balance including vector diagram. End of 1st experiement. Now remove previous correction weight to get back to the initial condition and repeat the experiment but this time pick a different 0 degree reference on the rotor. Repeat 20 times with 20 different 0 degree references on the rotor. I have 20 balance diagrams. Each gives the same final actual correction weight and position (although it appears at a different angle on the diagram). But the 20 vector diagrams look quite a bit different including the angle between T and TW.

Now if I go back and carefully consider the factors above (stationary keyphasor location compared to stationary vib prob location and rotating keyphasor target compared to rotor trial weight and correction weight zero reference), I can develop only one effective zero reference, only one vector diagram, and now the angle between trial weight unbalance vector and trial weight vib response vector is my true response lag angle.

Sorry to be longwinded. I apologize if I'm telling you something you already know.

This message has been edited. Last edited by: electricpete,

quote:

Originally posted by electricpete:
But the 20 vector diagrams look quite a bit different including the angle between T and TW.

Pete, all 20 diagrams are going to look the same, just angularly shifted. The angle between T and TW is also going to be same. How do I know this? From the theory and practical tests.

For simplicity, I have zero shift between the tach and the accelerometer (see the previous attachment), so the angle between T and TW obtained from the diagram represents the actual angular difference between newly created heavy spot (TW) and its vibration effect (T).

Steve, when I say "heavy spot" I mean heavy spot created by TW as oppose to overall heavy spot in a rotor since I only analyze the response T caused by weight TW. The residual (original) unbalance of the rotor is not a factor for a linear system.

IMHO, the lag angle measured on the bearing is a function of M,K,C combination of a rotor plus M2,K2,C2 combination of the structure (bearing). Eccentually we are talking of a second degree of freedom system in this case.

M2,K2,C2 of the structure may distort classic angular relationship between TW and T when T lagging behind TW and is small for speeds below first critical.

quote:

Originally posted by David_G:
Pete, all 20 diagrams are going to look the same, just angularly shifted. The angle between T and TW is also going to be same.

I disagree. Let us further clarify that in spite of the change in 0 angle reference system for labeling weight position between each of the experiments, that we keep the trial weight in the same actual phyical position for these 20 experiments (it's not necessary to prove my point but it helps my explanation).

Now, since we don't change the stationary vib probe and stationary keyphasor, the physical experitment is the same all 20 times and we have the identical VIBRATION vectors O, O+T and calculated T. There is no change or rotation of these 3 vectors among the 20 experiements. The vibration sensor and vib phase measurement don't know or care about your labeling of 0 rotor phase reference for weights. (they only care about actual vib pattern, physical position of vib sensor, physical position of stationary keyphasor and physical position of rotating key-phasor target)

But remembering that the physical TW location on the rotor was the same among all 20 experiements, the plotted TW vector must shifts among all 20 experiements as we change our 0 weight refrence.

Same T, differerent TW, means different angle between T and TW in all 20 experiements.

That doesn't prevent us from doing a successful balance. The random angle included becomes part of our influence coefficient. We will caculate the same physical CW each time.

So if we look at all 20 graphs, the VIB vectors are the same in all 20. But the WEIGHT vectors TW and CW are shifted a random angle with respect to the vib vectors among runs. The TW and CW will maintain the same angle between each other.

So if you are randomly establishing 0 reference position on the rotor for weight labeling and not accounting for relative position of stationary keyphasor compared to stationary vib probe and the relative position of 0 reference compared to rotating keyphasor target, as well as any vib instrument lag time (I'm not sure about that last one), then the angle between your vib vector T and weight vector TW will vary randomly depending on your random choice of 0 reference for weight labeling on the rotor.

This message has been edited. Last edited by: electricpete,

Pete - I think the confusion with your idea is that you have seperate angular references for the vibration phase angle (fixed keyphasor) and the position of the trial weight. It doesn't make sense to plot the trial weight position on a vector diagram when using a different 0 degree reference. When you calculate the infleunce vectors, you will have to account for the differences in reference positions.

I agree with Dave that all 20 diagrams will look the same, just rotated.

Steve

Steve - I think you are saying there is only one "logical" choice of 0 degree rotor reference for weight position, correct?

If so, then you're saying that we change nothing among the 20 experiements (not even the zero reference for weight angle) and we expect no change in the outcome if we didn't change anything during the experiement. No one could argue with that.

It is the assumption of the experiment that we change the 0 reference for weight labeling among each experiment. Randomly pick a different blade on the fan each time as our 0 reference. (don't take this into account in any software... just do graphical solution). We can still balance successfuly this way but we just can't determine lag angle.

This message has been edited. Last edited by: electricpete,

Here is an example of how I think two of these experiements would look. Again, there no change between experiments other than shifting my 0 reference for purposes of labeling weight positions (actual trial weight position remains the same). We do a simple single-plane graphical solution using phase angle from our meter and weight positions measured on rotor in direction opposite rotation from our arbitrary reference. (that arbitrary reference changes between experiments)

Balance #1
O = 1 < 150
Ut = 1<0 (same parameter as Dave's TW)
O+T = 2<100
Calculate
T = 1.56 < 70.6
Ucw = 0.64 < -100.6 (same parameter as Dave' CW)
Response Lag angle = Angle(T) - Angle(Ut) = 70.6 - 0 = 70.6

Balance #2 - Same actual trial weight position but different 0-reference for weight angles
O = 1 < 150
Ut = 1<90 (same as above except shifted 90 degrees in labeling but physically at the same position)
O+T=2<100
Calculate:
T=1.56<70.56
Ucw = 0.64 < -10.6

Response Lag angle = Angle(T) - Angle(Ut) = 70.6 - 90 = -19.4

You can see that Ucw is the same magnitude but changed by the same angle as Ut, which is the 90 degree change in our reference position. Ucw ends up in the same acgtual physical position.

BalanceWithChangingReference.ppt (377 Kb, 18 downloads)

quote:

Originally posted by electricpete:
Same T, differerent TW, means different angle between T and TW in all 20 experiements.

The above is where the logic is flawed. As you chose different reference mark (0 deg) on the rotor, the TW position in degrees will also be different and so will be T. The angle between T and TW will stay the same though as it is a physical parameter of this rotor.

Again, sometimes the angle becomes weird because of another mechanical system getting in between the rotor and accelerometer, thus skewing the angle. It does not prevent us from successful balancing though.

I am considering two different references on the rotor One is the keyphasor target which forms a reference for the vibration phase. The other is the arbitrary reference used for identifying weight positions.

I agree with Steve's' analysis, if you lock these two references together, and then perform the experiments by moving them together to different points of the rotor, the only effect is to rotate the entire vector diagram. (By the way Steve - my previous comment about "changing nothing during the 20 experiments" was a little off-base - sorry about that).

While it may be good practice and a logical choice to lock these two references together, the point of my experiment was to demonstrate that the relationship between phase reference and weight reference is important to determining response angle, but is not important to performing a successful balance. So in my experiment, you put the vibration phase reference tape on and don't move it again (or use the shaft keyway). But use a different weight reference position for each experiment.

The results of this experiment would be as I have identified above.

quote:

electricpete: "Same T, differerent TW, means different angle between T and TW in all 20 experiements."
David_G: "The above is where the logic is flawed. As you chose different reference mark (0 deg) on the rotor, the TW position in degrees will also be different and so will be T"

You're right that there was an error. I reversed the T and TW. It should have been: "Same TW, differerent T, means different angle between T and TW in all 20 experiements."

The TW is the same because we didn't change the vibration phase reference or physical location of trial weight so vibration vectors don't change. The T is different because we changed its reference.

The conclusion is the same: the angle between T and TW changes during each experiment. The point of the whole excercize is to show that the relationship between phase angle reference and weight angle reference cannot be ignored in this question.

This message has been edited. Last edited by: electricpete,

One factor that must be taken into account is mode shape. Keeping it simple and using only a shaft mode and not including foundation/support mode shape, one could look at the simple example of an 'S' shaped mode. If one places a weight at one end and measures at the other end at low speed the phase is 180 degrees, etc.

Take a rocking mode of a support system. One never or rarely measures the vibration at the point of application of the balance weight. If you did measure exactly where you put the force in you would expect the simple 0 to 180 degree relationship, but we don't do that in practice.