// the find
ShivaprasadVToggi/Industrial-Machine-Health-Monitoring-System
Real-time predictive maintenance system for CNC machines and Induction Motors using STM32F446RE, MEMS Accelerometers, and IoT Edge Computing.
A student/academic project that wires an STM32F446RE to an ADXL345 accelerometer and DHT22, then ships 5-dimensional telemetry to ThingSpeak every 16 seconds. The goal is generating a labeled time-series dataset for predictive maintenance ML experiments. It's a hardware bring-up and data collection exercise, not a production-ready system.
The interrupt masking for atomic sensor reads is the right call on a Cortex-M4 and is explicitly handled rather than ignored. Using the hardware FPU for float formatting is a sensible choice given the M4's FPU availability. The data schema is clean and well-documented — 5 named fields with units makes the CSV immediately usable for training. Separating edge preprocessing (normalization on the MCU) from cloud ingestion is architecturally sound for bandwidth-constrained IoT.
16-second sampling intervals are far too slow for meaningful vibration fault detection — real PdM systems sample accelerometers at 1–25 kHz and do FFT in the frequency domain; the 'future scope' FFT is actually the core requirement, not an enhancement. There is no actual ML model here — it's a data collector presented as a predictive maintenance system. No RTOS or scheduling discipline is visible; running LCD, I2C, UART, and sensor reads in a bare-metal loop will produce jitter that corrupts the time-series. Single contributor, 1 star, no issues, no release — this hasn't been stress-tested beyond the lab bench.