I have always assumed if balancing was done using prox probes, we would use the filtered runout-compensated 1x. Is it a good assumption, or would there ever be any situation when it would make sense to balance using uncompensated?

If the compensation exist, is good and to be trusted, that is what I would do, using the compensated. But I am not a eddy probe user to that extent as you know. If the balancing "calibration" data for some reason is uncompensated I would scratch my head and probably use it if data could not be recaptured so that may be a difficult case. OLI

Age old discussion no doubt for 30 yrs I'm aware of.

Run-out say of 2 mils and mass offset causing unbalance in addition to or solely. Should that run-out be subtracted? It is a contributor to 1X and that's what we're trying to reduce. I have leaned to uncompensated for balancing for reason above. After all, it is a contributor to the system.

If you can convince yourself you have quality 1X slow roll data, then you should in most cases balance with compensated data.

Think about this hypothetical scenario. You have a shaft that has a 1X runout of 2.0 mils pp at 90 deg lag. You think you do an absolutely bang up balancing job using uncompensated data and get the uncompensated level down to essentially "zero" 1X uncompensated. Guess what, the shaft is now "dancing" at 2 mils and at a lag angle of 270 deg. In effect the 1X "true" dynamic motion is vectorially cancelling the indicated runout vector of 2 mils pp at 90 deg lag. Again, just hypothetical, but if you have good 1X slow roll data, balance based on compensated data.

John from PA

Formally, technically the run out that is compensated are errors in the measurement from surface, electrical and other interference that should not be included in data for balancing as balancing is aiming at moving CG to Center of rotation for the rotor. If some other technology like capacitive or for that matter shaft riders with other limitation been used and given data more similar to the compensated that would strengthen that principle. It may then be the case that other derived indications from the eddy probe signal like some overall reading or so that is not compensated will give a false indication that the result is getting worse after balancing but that is, however maybe hard to explain for others, a result from limitations of that measurement and presentation and can´t be avoided unless the errors requiring compensation is reduced. So the basic rule apply, shit in give the same as output even if it some cases can be compensated for. Only my point of view. OLI

Glitch, run-out, glitch + run-out: as Ralph points out, slow roll data and it being valid is or must be considered and evaluated. Again, different views and/or approaches.

For a big machine and if your runout is low, (less than 0.5 mil pp) then it does not matter much.

If you are trying to really be precise, you have to subtract the runout.

If it is a very small high speed machine where even 0.5 mil pp is a significant percentage of the signal, the runout must be subtracted.

If the runout is high, and you are trying to fool the monitors, balance with uncompensated. But be warned you may actually be unbalancing the machine. I do not recommend trying to fool the monitors under any circumstances, I would recommend fixing the runout.

I'd subtract the runout first. The balance method, whether with computer software or by graph paper tries to reduce the vibration vector to Zero, within some acceptable residual tolerance. If data is uncompensated, then the "apparent or measured" vibration would/could be reduced to Zero, but the actual unbalance would not be Zero.
Not only could I be wrong, but I am willing to learn to be right!

Walt

Ideally, one has small runout, and it doesn't make much difference. In such cases I have often balanced the uncompensated, because it leads to less discussion and argument from the customer. ... And really there is no practical difference.

For larger runout (significant for the situation) the answer is generally subtract the runout. Runout can be glitch and/or bow. There are numerous papers discussing the theoretical aspects.

One can search for 'balacing bowed rotors.' If it is just glitch then one is balancing 'noise.' At the monitor or in the control room the vibration will look lower if one balancesl the 'noise;' however, the real vibration will be bigger. GET THE SHAFT NOISE UNDER CONTROL!