IOT - Internet of Things - Monitoring Motor Arrays
Video - Early Test - Old Geophone
DC Bias is fiddly depending on transducer - what worked for Geophone did not for these GIKFUN-HDX-2's (no data sheet)
//const int sdvig = 32767; //DC bias of the ADC, approxim +2.5V.
//const int sdvig = 15000; //DC bias of the ADC, approxim +1V???? THE BLUE SENSORS DC BIAS IS VERY DIFFERENT FROM GEOPHONE. (( WORKS with 2.5v dc bias in, 1v out)
const int sdvig = 00000; //DC bias of the ADC, approxim +1V???? THE BLUE SENSORS DC BIAS IS VERY DIFFERENT FROM GEOPHONE. (( WORKS)
- Have an array of motors running - sensors picking up time domain data for each one
- Do a FFT Fourier Transform on it - then time slice into 'windows'
- Tag each window over time for Spectrum
- Over time, use prior knowledge of 'known good' and if available 'known faulty' to classify into good, change, warn and fail states - moving window
- Unpack data a little more (e.g. frequency shift/delta, noise floor, skirt size, signal to noise ratio, signal purity, harmonics) for windows, e.g. N, N-1, N-10 etc..
- Use basic open source Machine Learning tools (e.g. Random Forest) for classify into good/change/warn/fail based on prior knowledge
- Rinse and repeat
There are more sophisticated Deep Learing and ML tools that probably handle the time domain more elegantly, so that's worth reading up on...
About this blog
Description is...<br/>Data Analytics & Visualization Blog - Generating insights from Data since 2013
Created: July 25, 2014Englishfrançais