OpenPMX — Open-Source Predictive Maintenance Platform

An open-source predictive maintenance platform for small and mid-sized manufacturers. Built to democratize industrial AI for the 300,000+ SMB manufacturers in the US who can't afford enterprise solutions like Siemens MindSphere or PTC ThingWorx.

🚀 Live Demo

Try it now — no installation required: 👉 https://openpmx-frontend.onrender.com

API Documentation: https://openpmx-backend.onrender.com/docs

What it does

• 🔍 Anomaly Detection — Detects machine anomalies before they become failures
• ⏱️ Remaining Useful Life — Predicts how many days before equipment failure
• 🏥 Health Scoring — Real-time health score (0-100) for each bearing
• 🚨 Instant Alerts — Critical alerts when machines need immediate attention
• 📊 Visual Dashboard — Factory managers see machine health at a glance
• 🔒 On-premise Ready — Runs fully on your own infrastructure

Tech Stack

• Backend: Python · FastAPI · Scikit-learn · NumPy · Pandas
• Frontend: React 19 · Recharts · Vite
• ML: Statistical anomaly detection · RMS analysis · Health scoring
• Deployment: Docker · Render · GitHub Actions
• Data: NASA IMS Bearing Dataset (University of Cincinnati)

Results on NASA Bearing Dataset

Bearing	Final Health	Status
Bearing 1	80/100	Healthy
Bearing 2	86/100	Healthy
Bearing 3	0/100	Failed ✓ Correctly identified
Bearing 4	39/100	Warning

Platform detected 908% vibration increase in Bearing 3 and correctly triggered a critical alert.

Quick Start

Option 1 — Use the live demo (no installation)

👉 https://openpmx-frontend.onrender.com

Option 2 — Run locally with Docker (one command)

git clone https://github.com/SahDhirendra/openpmx
cd openpmx
docker-compose up

Open http://localhost:5173 in your browser. That's it — no Python, no Node.js, no setup required.

Option 3 — Run without Docker (development mode)

# Terminal 1 — Backend
git clone https://github.com/SahDhirendra/openpmx
cd openpmx
pip install -r requirements.txt
uvicorn app.main:app --reload

# Terminal 2 — Frontend
cd dashboard
npm install --legacy-peer-deps
npm run dev

Open http://localhost:5173

Development Phases

• Phase 1 — Data exploration & NASA bearing dataset analysis
• Phase 2 — Anomaly detection & RUL predictor
• Phase 3 — FastAPI backend with ML engine
• Phase 4 — React dashboard with real-time alerts
• Phase 5 — Cloud deployment on Render
• Phase 6 — CSV upload for custom machine data
• Phase 7 — Federated learning across factories

Author

Dhirendra K. Sah
Controls & Automation Engineer | MS Mechatronics, NDSU
LinkedIn · GitHub

Contributing

Contributions welcome! This project is built for the US manufacturing community. See PHASES.md for roadmap.

License

MIT License — free to use, modify, and distribute.