Here are links to several interesting articles where the authors have done some experiments to introduce faults in the lab and measure the results:
http://www.spectraquest.com/tech/index.html
(You have to log in to access the articles)

I would order them from most interesting to least interesting as follows:

"Observations Concerning Misalignment Vibration Signatures" - Reports results of intentionally introducing various misalignment on a machine. Tried with 3 different couplings. Results were generally very little correlation between misalignment and vibration.

"Vibration Analysis of Belt Drives in Rotating Machinery" - shows results of adjusting belt tension and introducing misalignment in belt driven machine.

"Analyzing Gearbox Degradation Using Time-Frequency Signature Analysis" - TWF and spectrum were collected and defect was created they increased the size of the defect through several tests. Acceleration TWF peak and gearmesh sidebands increased as they made the defect larger. (I know...the effects of wear smoothing the edges might change the pattern, so peak acceleration doesn't tell the whole story... but I still like it for bearings...don't look at many gearboxes myself).

"Dynamic Behaviors of Rotor with Fluid Film Journal and Rolling Element Bearings" - shows the change in the shape of the orbits as the machine coasts down....in addition to getting larger tends to get elliptical near resonance for these machines.

Two articles show some testing to determine belt natural frequency and effect of tension etc.

The most disappointing article: "A Study of Rolling Element Bearing Defect Analysis" - A fairly elaborate test to measure bearing defect pattern at 4 different speeds and 3 different loads and different rotor weights and defect severity. They mention their equipment can measure up to 102,000 lines. One single graph was presented with 25,600 lines. Based on fault frequencies falling very close to harmonics of running speed they warn "It can be seen from our table that extremely high resolution is needed to detect bearing faults and the leakages or spectral smearing is eminent in the analysis using conventional FFT Analyzers." I am expecting some really good results. Unfortunately the ONLY data they for the results of their experiment varying speed, load, severity etc is the magnitude of the fundamental BPFO!!! It is very small magnitude and wanders up and down randomly with load and speed (who knows how the shape of the envelope of harmonics is changing). Their conclusion: "The results of the study seem to indicate that BPFO amplitude is not a strong function of speed, rotor weight, and fault severity within the range of this investigation.....No other parameters such as RMS level, crest factor, or kurtosis were used to compare the results..."

Gosh, would it have been too much trouble with that superwhambedyne analyser to record an acceleration TWF peak, an enveloped overall... or something just a little more interesting than BPFO fundamental ?!???! ;-)

Oh well, nothing is perfect. In fairness they did say they are going to do further review of their data in their next study. All in all some pretty interesting articles based on controlled testing.

This message has been edited. Last edited by: electricpete,

Hello Pete, this series of articles has been spread for some time. I have sent them direct comments and in no case have the come back, nor corrected the texts published.

The demo machine has a nice bunch of faults but as all demo/training situations, it is essential to have clean nicely repeatable faults to demonstrate. The people using a machine like this are unexperienced and if the demo shows ambigous results, it is something to work on.

A problem with rotor kits in general is that they rarely simulate actual bearing static and dynamic loads, and the structural properties (stiffness, mass, and damping) are far from real life with full size machines. I think rotor kits, including the "super deluxe" one, have value for demonstration and training, but the real limitations should be known and taught as well. I question whether the design was ever intended or optimized for bearing fault detection R&D.

Walt

I agree that even a complex rotor kit is of limited value in exploring "typical" machine behavior, and I would tend to take with a grain of salt any "conclusions" that someone might draw from a study using a rotor kit.

What I'd like to eventually do is set up a decent sized motor in the shop that could be coupled to a variety of machines that had already failed (fans, motors, gearboxes) and spin them to see what the defects look like. Then you could possibly "enhance" the existing defects to simulate deterioration. It'd be better, but still not "the real thing."

For that, you need to take data on real machines out in the field, that you had control over. That doesn't happen a lot for most of us.