We built a mobile vibration analysis tool — plug USB sensor, get bearing health diagnosis in seconds, no training needed
I work on a mobile app for bearing fault detection using USB accelerometers (USB serial). The goal was simple : plug the sensor, press start, get a GREEN/ORANGE/RED diagnosis — no laptop, no software install, no calibration.
How it works :
- Connect USB accelerometer to an Android device
- App captures vibration signal
- Backend API does FFT + ACF envelope analysis
- Result : bearing health state + defect breakdown (inner race, outer race, ball, cage)
No training required — the absolute mode works from the first reading. Also tracks degradation over time with relative mode (baseline comparison).
Tested on public datasets :
- 98.3% accuracy on Ottawa bearing dataset (zero training)
- 97% with training on same dataset
We're looking for maintenance technicians or reliability engineers who work with rotating equipment (pumps, motors, fans, compressors) to beta test on real industrial signals.
If you have a USB accelerometer laying around (or want to try with recorded data), DM me and I'll set you up with an APK and API key.
Happy to answer technical questions in the comments.