Engineering systems
with intention.

STM32 vibration
fault classifier

A reproducible motor-condition prototype that turns streaming vibration samples into DSP features and readable C inference for resource-constrained microcontrollers.

• 1 kHz sampling with deterministic 256-sample windows
• RMS, kurtosis, crest factor, and Goertzel frequency features
• Python-trained decision tree exported directly to embedded C

Problem

Detect motor health changes from noisy vibration signals before they become visible failures.

Build

Generated synthetic regimes, trained an interpretable model, then verified the same logic in C.