configuration is two 8" cage fans on a through shaft motor(one fan on each side). Shaft dia is roughly 1" and 20 inches shaft length out each side! To add to it, the only place to mount a transducer is on the motor can with only 10 inches or so between transducers. It's really a balancing nightmare. My software keeps coming up with absurd amounts of weight to add(too much).
When adding trial weights, I get a good phase shift, but I can add quite a bit of trial weight without affecting the amplitude as much as I think it should.
The cages are balanced before mounting, but they are trying to hit final tolerances after assembly

any suggestions would be super helpful

If the fans were not balanced at speed before mounting, You might try removing one fan, balancing the other as low as you can, then removing it and installing the first one and balancing it as low as possible. You should try to get well below your tolerance on each fan so that when they're both installed the total unbalance won't drive the level above your limit. If they're not keyed to the shaft make sure you mark their orientation on the shaft.

Are you sure this is imbalance issue? Sounds like you are operating near a resonance/bent shaft per your description of what is happening during balancing procedure. Is there multiple turning speed harmonics present on the motor? If the fans are already balanced the likely faults are looseness in motor/mounting structure, air flow restriction, resonance. You can balance when operating near resonance, pain in the butt and almost guartanteed the fan will not stay in balance.
When you add trial weight, amplitude does not change much? Try moving your tape 90 degrees on the shaft. Large amounts of final weights is another indicator that you are operating at or near a resonance. I agree with Joe, check/balance one wheel at a time.

quote:

configuration is two 8" cage fans on a through shaft motor(one fan on each side). Shaft dia is roughly 1" and 20 inches shaft length out each side! To add to it, the only place to mount a transducer is on the motor can with only 10 inches or so between transducers.

8 inch fans? Hmm. Small huh? How wide are they, 18 inches?

40 inches apart? Each shaft 20 inches long.

Transducers 10 inches apart? Must be a very small motor, huh? Might could use jus one transducer with them that close together and sort of come in the back door with a poor man's 2 plane balance job.

What amplitudes are present?

How is the motor mounted, in its actual running position or on a test stand? Is it bolted down or what? Is it sitting out in the open and not in its housing and the fan being able to "suck" and blow air in every diferction as opposed to be controlled air flow?

Not being able to change the amplitude with a large amount of weight doesn't sound like it is running in a resonance state, IMO. The phase shift you say you get, how much is it? May be going past the correction point.

What is the speed of the motor?

Too many questions? Might even have more after these are answered.

I had to balance one like this last year. It was on an air curtain for a roll up door.

What I found was broken welds underneath the motor base. Fixed the broken welds, and THEN balanced.

May not be your problem, but it sounds very similar.

Thanks guys.
After more analysis, I don't think unbalance is their primary issue. I think most of what is causing the problems is turbulence in the air handling system that creates a low frequency pulse, I would really have liked to see the rotor run with the shaft only as well as one cage on each side as suggested, but the entire unit is assembled and the customer has to "get them out the door". Honestly, with this configuration, I think the balance is as good as its going to get, as they are coming in between .1 and .04 g acceleration 0-peak.

The fans are approx 10 inches in length, with the shaft mounting in the middle of the cage.

Andy, anytime I balance anything that has an width to it, if I have access to both end planes, I "pair" my weights, with one in each plane at the same angular position. This assumes that it's mostly a static imbalance and the initial phase is fairly close on both ends. If the rotor is fairly symetric, I will ratio the amount of the trial weights to match the amplitudes at each end. In the software, I just treat the trial weights as a single weight. I do one calculation using both readings and the single (combined) weight. Then I do a calculation for each end, using only the weight on that end. It helps to see what each end "wants". If there is enough seperation between the bearings, each bearing will respond almost independently of the other. Of course this is for centerhung rotors....

....you said "overhung"... my bad. Regardless of the setup, sometimes the "solution" doesn't make sense. You have to go "intuitive" at that point. If you got very little phase change, but the amplitude went down, then just add a little more weight. If the phase changes very little, you're on the right spot... just keep adding weight, a little at a time. Likewise, no phase change, but amplitude increases, you either have too much weight, or you are adding weight at the 'heavy' spot and you are 180 out. You probably know all that, but the main thing is to keep doing what makes the vibration better, regardless of what your instrument says.

This may well be a two plane balance. I would want software that could handle a different number of balance planes than measurement planes.

An interesting approach would be first to see if the vibration is mostly static or couple across the motor, the 10 inches - Use your 10 inches if that is all you have . This can be done by taking amplitude and phase measurements in a plane, horizontal or vertical (I would use both H and V to calculate the balance.).

Next it would be interesting to apply a static or couple type shot across the machine, one weight on each fan according to whether the vibration is mostly static or couple.

If the static or couple balance doesn't fix it, next would be a couple or static, different from the first attempt. If this or just a single static or couple works, you will have simplified the balance by doing essentially one or two single plane balances.

If after this, the results are not sufficient (trims might be a good idea if the data supports that approach), one will have enough data from the separate 'static' and 'couple' shots to calculate a true 2-plane balance. In fact prior to calculating the second single plane balance for the couple or static weight, one can also compute a true 2-plane balance to weigh the relative benefits of using a true 2-plane balance instead of a second single plane calculation with the couple or static.