Choose Language Hide Translation Bar
Level I

Vibration Analysis: Fault Frequencies Correlations



I am currently performing a vibration analysis using JMP for the first time. As data input I have FFT (Fast Fourier Transform) data from an extruder gearbox for 5 different time stamps (baseline, pre-event, post-event, 6 month increments after event X2). Each FFT data set comes with 16,000 frequencies (Hz) vs magnitudes (g). In vibration analysis the primary objective is to identify "fault frequencies" for bearings, gears, etc. These fault frequencies and respective harmonics have a specific frequency that when they become an issue, they will show up in an FFT with a significant peak. 


My question for this community: is there to have JMP "look" for these known frequencies in this vast data pool of FFTs and extract their respective magnitud in order to observe their progression in time (and possibly their harmonics)? The only inneficient way I have managed to this is by using a script that compares the frequency in the FFT to these local constants (fault frequencies) and if they "match" within a tolerance, it prints the magnitud. However, it is not very efficient and considering I have 12 bearings and 9 gears, this process will take a very long time. I just know that JMP can do better than that!!


Thank you for any help or guidance you can provide!


Re: Vibration Analysis: Fault Frequencies Correlations

Hi, I'm not completely clear what the challenge is here. It sounds like you have specific known fault frequencies from the 16000 that you would like to track. But there is some difficulty. From your description it sounds like this might be because the "known" frequencies are somewhat variable - you need to match within a tolerance. You have a solution using scripting - why is this inefficient? Do you have too many fault frequencies?


Re: Vibration Analysis: Fault Frequencies Correlations

Do you have JMP Pro? You might try analyzing the spectra with the Functional Data Explorer. You could then save the functional principal components. You could fit a model with FPCs versus time or event versus FPCs.

Learn it once, use it forever!
Article Labels

    There are no labels assigned to this post.