First critical speed question

[geesamand]

I work with vertical shafted machines, usually supported at the top with a pair of tapered roller bearings. For machines driven by an output gear located between the bearings, our traditional critical speed calculation has been fine. (We neglect housing and bearing stiffness effects). Our past experience has shown this calculation to be within a few percent of any ring tests on actual machines in air. These machines output at 20-200rpm.

Recently we produced a couple of direct drive machines, both driven by a 900rpm motor attached above using a flexible coupling. One machine used a pair of carefully clearanced spherical roller bearings, the other used tapered roller bearings. Both of them ring tested 20-30% lower critical speed than we predicted.

What might be overlooked that is causing such lowered critical speeds?

 

[BobM3]

Sphericals won't have the moment stiffness that tapered roller bearings have. I assume you included the mass and spring rates of the coupling into your model.

 

[geesamand]

I have not included the motor rotor or the coupling. Since it's a flexible coupling and the motor has its own bearings, how non-conservative is it to neglect items above the flex coupling?

But now that I think more, it would be sensible to rerun the calculation with those parts to bracket the result.

Dave

 

[Tmoose]

Does "in air" testing mean without product, or in a rig that simulates a free-free mounting?

I obviously don't have a clear picture of your system, but it's hard for me to imagine a structure so stiff it can be neglected in critical speed calculations. Add a motor and the problem could evolve into the famous reed frequency of vertical pumps, for which one of the most important responses is often dominated by structure, and the bearing stiffness is trivial.

 

[geesamand]

Thanks Tmoose and BobM3.

I have had a feeling that bearing stiffness is insignificant in this problem. No doubt I feel the structure is a major factor and maybe the motor as well.

Yes, 'in air' means no product. This is an agitator that runs with the lower end of the shaft in water. Prior to final balancing, the assembled machine was ring tested.

I will look into this Reed frequency phenomenon, which I infer includes the motor into the analysis. Like I said before, our traditional designs are not direct-drive, so our most detailed shaft frequency model uses a 2-bearing support. We neglected any effect of the motor being attached to the upper end, and due to the flexible coupling it did not concern us. The coupling in this case is a LoveJoy L225 jaw coupling with the Hytrel insert.

So the logical next step is to study this Reed frequency analysis. If you know of a practical, approachable treatment of the subject I would really appreciate the recommendation.

Thanks for your help.

 

[Tmoose]

I picture your set-up looking like this.

http://www.rainbird.com/images/products/turf/pumpstations/VerticalTurbine_bg.jpg

If so, the pump and etc structure is all important, and the motor and its mass is pretty much a lump. Like a ball of mud on the end of a fishin pole, in the words of Charlie Jackson or another vibration legend. To such an extent that Some times tuning to shift resonance is accomplished with thin shims in the joint of the motor face mount.

If your application already exists there is a good chance you have un-intentionally nearly hit a reed frequency resonance, and things like creative de-tuning, extra fine trim balancing of the assembly, or a tuned absorber (for constant speed apps) need to be in your field service tool kits.

 

[geesamand]

Yes Tmoose, that pic is representative of the layout of the structure. Thanks again for your help.

I'm still looking for a good approachable reference that will explain how to proceed with structure and motor effects.

In fact, I have yet to determine the difference between a reed frequency resonance and a traditional resonant frequency.

Dave

 

[Tmoose]

When first bought my own FEA program I could only afford several hundred node capability. That was enough to hit some vertical pump criticals with decent accuracy. Manual calculations of the all-important casing stiffness were 'way too simplified, especially if the stiffness is derived from piece-wise assembly of "I"s of the various sections - example is a transitional diameter or housing/volute. If the I of a thin disk is modeled as a thin slice of a thick walled tube the I is real big, but in reality it is a mighty flexible diaphragm. It should at least be modeled as a flat plate on the end . One test is to see if the manual solution has significantly (several hundred rpn) different frequencies in different directions. Real pumps often do. On top of that, a bed plate of poor proportions or installed with good intentions but poor workmanship (leveling nuts left in place, etc) is far from a rigid support and thus bring the reed frequency WAY down all by itself.

 

[geesamand]

Well our equipment is usually mounted above a tank, so the support structure is far from solid. I believe the only reason the flexbility of the structure isn't normally considered in our designs is because we typically build long slender shafts running at lower rpm. A 2-bearing-support calculation usually got really close to test values. For this analysis, we even did a 3D FEA analysis, so I'm very confident the rotating components (not including motor) are very accurately considered.

This design, being direct drive (higher speed) and having a fairly short shaft, means the first critical speed is more affected by support stiffness than our other designs. The challenge on this product line will be managing the support structure requirements - I suspect the ultimate answer will simply be to design further below first critical speed so that the flexibility of the support structure won't catch us as often.

Dave