Spectrum tolerance

[JeanMicheling]

Hi folks,

I'm looking at a spectrum signal of a 3600 rpm spindle. one of the bearing defect frequencies is 2332 hz and 2915 hz according to NSK spec. I got a high peak (about 450mg) at 2337 hz and 2918 hz (about 520mg). Is there a tolerance on those readings? Can I condiser those frequencies as bearing defect frequencies?

Thanks

 

[Tmoose]

A typical new machine tool standard is 0.5 g peak overall for ball bearings and maybe 0.75 g for roller bearings. The definition of frequency range (forever, 600 kcm, ??) to use for "overall" has sparked more than one debate, and your instrument may limit the options too.

Any way, with two discreet frequencies each containing about 0.5 g, whether peak or rms, your overall almost certainly exceeds 1 g. The fact the vibration frequencies are such a good match for fault frequencies is remarkable, and indeed suspicious. Fault frequencies 38 and 48 X rotation are numerically pretty high. Are these double row roller bearings?

Is there any vibration history for this spindle? If the vibration is new, it is suspicous also. A good crash can create inner and outer race defects instantly. I like to check my analyzer is configured correctly for the accelerometer sensitivity, and the mounting method is suitable for these high frequencies too. How does the spindle sound, by maked ear, and also listening with a rod or screwdriver?

 

[sms]

Yes those numbers are certainly and remarkably close and can be considered bearing defect frequencies. Adjusting both the Fmax and the number of lines of resolution on your analyzer will change the ability to resolve a frequency.

-The future's so bright I gotta wear shades!

 

[cincibcats]

sms, is correct I would make sure that your doing an FFT with enough spectral lines. If you have 4000Hz of bandwidth, then you are most likely sampling at 8000Hz and thus need a 8000 point FFT to get a 1Hz frequency resolution.

 

[Hatch]

Over many years, I have never actually hit a bearing defect frequency dead on (i.e. spectrum peak differs from defect frequency). Even if my FFT resolution is OK, there always seems to be a small 1 to 5 Hertz differnce.

In fact if you calculate the bearing defect frequencies from the standard equations (using the manufacturers data on contact angle, # rollers/balls etc), it also differs by a few Hertz from the manufacturers quoted frequencies.

In short, you are close enough in my books. What are the chances of something else generating that particular frequency anyway.