0
$\begingroup$

I have 3 identical gears that are meshing together, on each shaft that is holding these gears I have an accelerometer recording vibration data. Once I move to the frequency domain via MATLAB's fft I am unable to see any specific frequency that would describe the meshing frequency of my gears, it looks all the same.

Are there any methods I should be looking at to tackle such a challenge of identifying the gear meshing frequency and probably identify which gear is inducing what frequencies?

enter image description here

Also would it make any sense to run the motor while disconnecting the gears and record the vibration induced by the motor, and analyze that to filter out the motor's influence?

| improve this question | |
$\endgroup$
1
$\begingroup$

Yes it would make sense to analyse the sound of the vibrations without the gear interactions if you can, and then delete that sound profile from the overall sound using a noise remover that works with a sample of the noise only prior to the noise + recorded signal. You should say roughly what frequencies you expect from the gears based on the RPM and number of teeth, because it affects analysis methods. On this and many other programs there is "noise reduction using noise sample profiling" http://wiki.audacityteam.org/wiki/Noise_Reduction it is tremendously efficient.

There are no obvious correlations or patterns in the graph, it is true. Read about engine vibration analysis. I found a paper on engine sound graphs when i was researchign SPWVD, and indeed it is 100ds of times more precise than FFT and pseudo smoothed wigner ville distribution makes waterfall graphs which show the clearest patterns of sound that are physically measurable.

To seriously analyse your vibrations, you should not use coarse FFT, instead use pseudo smoothed wigner ville distribution which is available as an addon. this program can do it too, see the graphs: http://www.christoph-lauer.de/sonogram

| improve this answer | |
$\endgroup$
  • $\begingroup$ As the years have gone by, the link to Audacity's noise reduction wiki page no longer exists. As a replacement, maybe the corresponding manual entry is of use... $\endgroup$ – applesoup Feb 5 '19 at 13:21
0
$\begingroup$

If the gears are identical they'll produce the vibration at identical frequency. The amplitude should sum up in this frequencies. From industry machine the only way to know what gear is producing the most energy is to move the accelerometer close to each one of the gearboxes and check differences. Also, the gear mesh frequency is supposed to be easily detectable only when there is a problem with the gear, for example, a crack in one of the teeth.

Take a look on this resource

| improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.