Puzzling machine vibration

[geesamand]

I'm trying to troubleshoot a recurring issue when we build a new machine.

It's a 3" horizontal shaft with a 60# impeller that is mounted at the end of the shaft about 40" away from the inner bearing. The inner bearing is a deep groove ball bearing (6022) permitted to float axially in it's housing bore and the outer bearing is a 6321 that is fixed axially in the housing bore. Both have tight fits on the inner race. The critical speed of the shaft assembly is about 850rpm and the operating speed is about 425rpm. It's driven by an 1800rpm motor and 4.18:1 toothed belt. The shaft is fully machined and the impeller is balanced to G2.5.

The build these machines and spin test them in air. The problem is that a fraction of them will show much higher vibration (8-9 mils displacement) vs. the normal 2 mils. The frequency analysis shows the vibration is predominantly at 850cpm (2x shaft). 1x and 3x shaft usually shows at a much lower level. I've reviewed this problem many times and unfortunately I still can't pin down why some of these machines run poorly and the others don't.

I thought I understood this situation but I'm going to back up and ask some very fundamental questions:
1) Does the fact that operating speed = 1/2 critical speed tend to be an issue?
2) Is there any particular guideline for the alignment of the bearings to the shaft? (I have the bearing mfr's recommendations but that's in relation to bearing life, not acceptable vibration)
3) We often disassemble these bad actors and reinspect key components only to find nothing particularly wrong with them.

I'm ready to call a consultant and watch them waste our money just so I can step away from this thing for a while. I'd appreciate any suggestions that might help me avoid that equally unpleasant situation.

 

[MikeHalloran]

Put a VSD on the motor at final test, and sweep the drive frequency above and below the nominal. You probably don't have to do it forever, but do a significant sample.

I'm guessing you'll find a strong resonance not far away from the nominal, and it will exist over a relatively small range of frequencies, and that for some units their response band includes the nominal.

I.e., the critical speed is very nearly an integral multiple of the normal operating speed, so small production variations are getting the unit into the area where the response is disproportionate.

Change almost anything that moves the critical frequency by say ten pct, and the problem may disappear. The intent of the sweep test is to find out in which direction to make the change.


Mike Halloran
Pembroke Pines, FL, USA

 

[desertfox]

Hi geesamand

There are lots of little differences that might cause some shafts to vibrate more than others,you mention that the impeller was balanced but are the shafts balanced? I have posted some links that might be of interest:-

http://www.update-intl.com/VibrationBook2h.htm

sig=Kcn9CY1wzQB0vcf6DYur_gFTPTM&hl=en&ei=_
tFDTcuQFcqahQfIhfCWAg&sa=X&oi=book_result&ct=
result&resnum=3&ved=0CCQQ6AEwAjgU#v=onepage&q&f=false

http://books.google.co.uk/books?id=...unning shafts at below critical speed&f=false

http://www.roymech.co.uk/Useful_Tables/Drive/Shaft_Critical_Speed.html

desertfox

 

[geesamand]

desertfox,

The shaft is not balanced separately but it is fully machined with the highest runout being .005". (If that were the issue, wouldn't we see a strong 1x vibration signature)

Mike,

So in effect you're saying that operating at 50% of first critical is a problem? Do you know why this is never documented in rotordynamics references?
Moving critical speed by 10% will be very difficult as this is a production design. Several years ago we revised the shaft length with little improvement. On a few of these bad actors we have moved operating speed on a few occasions with little improvement.

 

[desertfox]

Hi geesamand

If the shaft isn't balanced, when you add the extra mass of the impeller wouldn't that magnify the vibration, if you look on the first link I posted it talks about shafts having keyways in two positions (180 degrees apart) and running the shaft at half its critical speed and setting up a vibration due to varying shaft stiffness when the keyways in the shaft are at top and bottom dead centre.

desertfox

 

[MikeHalloran]

I should have highlighted the phrase "integral multiple". You will find it all over vibration texts.

If you go to the trouble of recording the entire response of a bunch of production sweep tests, not just the peak frequency, the product itself will tell you how far you have to move the critical frequency.

Speaking of two keyways, our college had a critical speed demonstrator comprising a disc on the middle of a long double-D shaft. I.e., the shaft had two flats at 180 degrees along most of its length. It would reliably excite itself very nicely.

Which says to me that if your product has that kind of rotational symmetry, you might be able to solve your problem by disturbing it. E.g., if you have two keyways or flats, provide three instead, and see what happens. See if you can find other ways to make features of the product so they are guaranteed _not_ rotationally aligned.



Mike Halloran
Pembroke Pines, FL, USA

 

[geesamand]

OK, I think you guys are bringing me into new territory and parts of this topic I need to better understand. I really appreciate this.

I'm familiar integral multiple thing before, but our company has considerable experience in slower turning vertical shafts that NEVER excite significantly when we design the unit to run at/below 70% of first critical speed. When I read "integral multiples" I take integers 1,2,3,...n and multiply them by critical speeds. So it seems you're confirming it can mean fractional multiples too. Which fractions matter and when? I will do some more reading but if you can point me on these questions that would help immensely.

As for the rotational symmetry, it is rotationally symmetric except for the keyways. The impeller is 3 bladed and checked for consistent geometry. The shaft has 4 keyways along it's length: (1) a 7/8" keyway at "0" degrees on the outboard end for the pulley, (2,3) a short 3/8" keyway under each bearing at 180 degrees, and (4) a 3/4" keyway at the impeller at 0 degrees at the inboard end. The keyways and mounted components are the same length as the keys so once assembled there is not excess imbalance.

A rough sketch showing locations:
(1)x(2)xxxxxxx(3)xxxxxxxxxxxxxxxxx(4)

Which keyways are the problem?

While I would need to get new shafts made to do this (cutting extra keyways will warp the shaft I'm sure), a new shaft can be made with the keyways clocked in any direction. I might be able to get a shaft rekeyed and straightened and leave the old keyways in place.

Changing critical speed is tough because past analytical studies show that the impeller weight and location dominate the critical speed. The impeller shape and location is not negotiable. Turning down the shaft near the impeller has minor effect only. For other reasons I have sketched up a concept using angular contact bearings that put the inboard bearing effective center closer to the impeller but that's a major undertaking and my ass would be on the line in a big way if it didn't work out really well and on schedule. (As in, modifying the existing design is politically/commercially better than throwing something new into the situation).

Thanks again,

David

 

[geesamand]

Darn, this forum won't let me edit what I just wrote. It seems this first link:

http://www.update-intl.com/VibrationBook2h.htm

specifically talks about my situation. It does make sense that uneven stiffness create 2x input to the system where imbalance and misalignment only input at 1x.

If uneven stiffness is the issue, it follows that the two small keyways under the bearing are the reason. On paper this sound good. Experience has shown this was a problem before we added keyways under the bearings. (We had no keyways for quite a while). Plus the keyways under the bearings are shallow and short.

 

[geesamand]

Mike,

Regarding the VFD idea we don't have one in that area of our facility but we can play with coast-down tests. I will investigate that. I might also be able to reconfigure with a lower belt ratio and perform a coast-down test so that the machine spins at say 500rpm and coasts down through 425 to see where vibration response peaks.

David

 

[electricpete]

I have seen 2x on a belt driven machine, discussed extensively here:

http://maintenanceforums.com/eve/forums/a/tpc/f/3751089011/m/4421034563?r=4421034563#4421034563

There are of course a lot of possible causes of 2x in general.

In our particular case, resonance was not a factor but I suspected the keyway causing bending stiffness varying at twice per revolution as you discussed. The keyway is ridiculously long to allow adjustment of pulley position (there are photo's that you can see if you register at that site).

We did an experiment to adjust the pulley tension. The 2X went up with higher tension, down with lower tension. In my mind that is consistent with the bending stiffness theory since the radial loading that acts on the varying stiffness is created by the belt.

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

 

[electricpete]

I guess part of what I'm suggesting is that radial load is a necessary component for the 2x-varying stiffness to result in vibration. In theory the weight of the machine can provide that loading in horizontal machine but that would be rare. I believe it is more common and more expected when you have a belt.

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

 

[MikeyP]

from "Rotordynamics" by J S Rao

Chapter 9 - Shafts with dissimilar moments of area

"In the previous chapters, we considered the shaft to be circular in cross-section, which has the same moments of area about any diameter. For rotors on horizontal shafts, besides the main critical speed, disturbing whirl amplitudes have been observed at half the critical speed. On vertical shafts such is not the case indicating that gravity is one of the causes for this."

The chapter then goes on to discuss how dissimilar moments of area have an even greater effect as you have already pointed out.

Whether or not theory says there will be a self excited force at 2nd order or not. I would still stay away from half critical speed because you never know what external 2nd order forces may be applied from e.g. the motor or the load or something else you might not have thought of.

M

--
Dr Michael F Platten

 

[geesamand]

I've been pondering the effect of the keyways on this machine (again, these are small / short keyways compared to the shaft) and it got me to thinking the real issue with stiffness might not be the shaft itself but the shaft support.

This machine is mounted in such a way that the vertical direction is more rigid than the horizontal. This mounting is not negotiable and we have stiffened the horizontal as much as is practical. We measure a substantial difference in 1st critical speed depending whether the probe is horizontal (785cpm) or vertical (900cpm).

Therefore isn't it possible that the inherent imbalance will excite at 2x shaft speed, as it encounters a change in mounting stiffness twice per revolution?

If that is true, then it looks like I have little choice but to change first critical speed. We have a coast-down test coming shortly.

 

[electricpete]

Therefore isn't it possible that the inherent imbalance will excite at 2x shaft speed, as it encounters a change in mounting stiffness twice per revolution?

No, not in my opinion. Stationary asymmetric stiffness is different than rotating asymmetric stiffness. The rotating assymetric stiffness acted on by a constant direction (stationary) force causes 2x response. Stationary asymmetric stiffness operated by a constant-direction force causes constant (DC) force. Stationary asymmetric stiffness operated on by rotating force will cause 1x. Of course there are things like looseness that can result in harmonics of 1x, but that’s a different story.


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

 

[electricpete]

I think you could probably come up with a scenario where stationary asymmetric stationary stiffness (support asymmetry)) increases the 2x response in presence of asymmetric rotating stiffness (keyway) ... because the asymmetric stationary stiffness may encourage shaft bending. But take away the rotating asymmetry and it’s an irrelevant scenario to our discussion.

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

 

[JJPellin]

You should be very careful about changing the mounting stiffness. I just got back from a rotordynamics class in Houston. They had a demonstration model that was highly unstable when mounted with the same stiffness in both directions, but completely stable when mounted with drastically different stiffness in one plane versus the other. If you are experiencing a rotor instability, you could easily make it worse by trying to increase the horizontal stiffness. A round orbit tends to be less stable than an elliptical orbit.

You are assuming that the vibration is associated with two times fan speed. I find it more interesting that the vibration is at about 47% or motor run speed. This is the sort of sub-synchronous relationship that starts to look like a rotordynamic instability. Since you are belt driven, there is potential for other exciting forces. An eccentric pulley on the motor will produce an excitation at motor run speed which could excite a rotor instability at about half of that (since the belt ratio is almost 4 to 1).

Johnny Pellin

 

[geesamand]

I hope it doesn't sound like I'm assuming that the vibration is 2x shaft speed. We've measured it at 850cpm and that happens to be 2x shaft speed and happens to be in the range of first critical speed of the shaft. Our coast-down test tomorrow should resolve which of these is not circumstance.

David

 

[Tmoose]

Something to consider with belt driven equipment is belt fat spots (V and poly V) and pulley/sheave eccentricity creating once per rev "tugs" of extra belt tension that can excite all kinds of things.

Triggered strobe lights used to be standard equipment with vibration analyzers, and made locating guilty components "stand still." One time we selected a few sets of "good" v-belts (one just for today and a few years worth of spares for the future) for one of those fancy bug killing HVAC units by measuring the vibration on the hospital director's desk.

Do the inner races butt against shaft shoulders? I'd also measure the axial runout of the inner race faces as installed on the shaft, and the axial runout of the outer race ( should be NONE ). Traditional/classic vibration analysis tables often link 2X vibration with "bent" shafts, which is effectively what a cocked inner race is.