Shaft Whirling

[rickfischer51]

I am modeling a pump shaft with Ansys. My Campbell diagram shows a separation between the forward and backward whirl at zero rpm. I did some reading and found that this can happen when the bearing stiffnesses are not axisymmetric, but mine bearings are axisymmetric. What else can cause this?

 

[MikeyP]

Maybe I am missing something, but a forward whirl is a whirl orbit in the direction of rotation and a backward whirl is an orbit in the opposite direction to rotation. By this definition then there cannot be whirl at zero speed.

--
Dr Michael F Platten

 

[electricpete]

Mike - Think of it as a limit. Gyroscopic effect speed-dependent stiffening... differt for forward and backwardwhirl speeds. As speed approaches zero, the effect goes to zero and the forward/backward frequencies converge to the same frequency at w = 0.
Example: Figure 14 here

http://www.sandv.com/downloads/0505swan.pdf

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

 

[rickfischer51]

Electricpete, that is the reference I used to brush up on rotordynamics. I should, I think, get plots that look like fig. 12, but they look instead like fig. 19. Swanson says at the bottom of page 15 that asymetric bearing properties can cause the separation seen in fig. 19. I built my model with symmetric bearing properties, and I did not model the housing. I fixed the od of the bearing, (actually the second node of a COMBI214 in Ansys) in space. My issue here is that I am ignoring the backwards whirls because I believe they dont actually exist if the bearings are symmetric. But I am getting separation at zero rpm, which indicates asymmetric bearings. So, either I have an error in my model and dont actually have symmetric bearings, or something else causes the separation. I've checked my model, and it looks right, but I have another pair of eyes looking at it. Can anything else cause this?

 

[electricpete]

My issue here is that I am ignoring the backwards whirls because I believe they dont actually exist if the bearings are symmetric.

One thing to comment - I don't think it is fair to neglect backwards modes on the basis that there is no asymmetry. Distinct backward modes (at different frequencies than associated forward modes) can exist due to gyroscopic effects if there is a disk that tilts in the mode of interest. Whether or not they will likely be excited is different question.

But I am getting separation at zero rpm, which indicates asymmetric bearings. So, either I have an error in my model and dont actually have symmetric bearings, or something else causes the separation. I've checked my model, and it looks right, but I have another pair of eyes looking at it. Can anything else cause this?

I'm not as familiar with effects of effects of "asymmetric bearings". It does make sense to me that fluid film bearings in particular would have a different behavior in forward whirl than to backward whirl (because the whirl direction reverses but shaft rotation direction does not... so one is whirling with the oil velocity and the other opposing the oil velocity).

To my way of thinking (*) the more important aspect in creating a difference between forward and backward is the asymmetric matrix describing the bearing (rather than just asymmetry in the sense of differing self-stiffness in x and y directions). I think maybe the cross coupling you mentioned will show up as a skew symmetric term in the matrix? (Kxy = -Kyx or Cxy = -Cyx) ? That might be the source of the behavior. Maybe try getting rid of that cross coupling term and see if this feature goes away?

* Disclaimer - I'm in over my head. Others can speak more intelligently to these things than me.

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

 

[rickfischer51]

Thanks for the reply. I'm also in a bit over my head. I'm working off the sample problems in the ansys documentation, making simple model to start, etc.

First, I found the problem. The CAD file I used to calculate mass properties used the Z axis as the axis of rotation. My FEA model used the X axis. So, my moment of inertia values were in the wrong fields on the R card, i.e. it was a asymmetric mass problem, not an asymmetric bearing problem. Now my plots look just like fig 15.

Swanson discusses forward/backward modes in the middle of the first column on page 15. When I specify an imbalance and do the harmonic analysis, and plot bearing deflections, I only get peaks at the forward whirls. I think this has to do with the whetehr or not there is an imaginary component.

I think the biggest contibution to bearing asymmetry, and perhaps the most common, is if the bearing mounting is not symmetric. If the housing where the bearing is mounted is cylyndrical, then no problem. But often this is not the case. If for instance, there were gussets cast into the housing, or other geometry nearby that would locally affect stiffness, then the bearing response is asymetric and the whirl otbits are ellipses instead of circles.

In my model, I set Kyy = Kzz, and Cyy = Czz, and let the cross terms equal zero, so this gives me symmetry, assuming I dont make a stupid mistake and hose the mass properties.