About The Project
MotorGuard is an end-to-end predictive maintenance platform that combines embedded sensing, real-time data streaming, cloud analytics, machine learning fault classification, and a modern web dashboard.
The system continuously monitors electric motor behavior using an ESP32 + sensors (ADXL345, INA219, DS18B20, Hall effect), sends telemetry to a FastAPI backend, and predicts likely faults using a trained Random Forest model. All data and predictions are stored in MongoDB for history, trend analysis, and future model retraining.
🏗️ System Architecture: Sensors → ESP32 → Wi-Fi JSON → FastAPI Backend → ML Inference + Analytics → MongoDB → Next.js Dashboard (live + historical)
⚡ Fault Classes Monitored: • NORMAL — Healthy motor operation • OVERHEATING — Temperature threshold exceeded • OVERLOAD — Current/power spike detected • BEARING_FAULT — Vibration anomaly signature • STALL — RPM dropout or locked rotor condition
Key Features & Architecture
Real-Time Motor Telemetry
JSON over Wi-Fi from ESP32 with ADXL345 (vibration), INA219 (voltage/current), DS18B20 (temperature), Hall sensor (RPM)
ML-First Fault Prediction
Random Forest classifier with confidence-aware reporting and rule-based fallback
Remaining Useful Life (RUL) Approximation
Predictive degradation modeling for temperature and vibration trends
MongoDB-Backed Analytics
Historical storage for trend analysis, health scoring, and model retraining pipelines
Comprehensive Backend API
Endpoints for data ingestion, predictions, model management, retraining, and AI-powered diagnostics (Groq RAG)
Live Dashboard UI
Real-time monitoring, trend visualization, fault distribution, and diagnostic deep-dives