Predictive Maintenance & KPI Dashboard

🚀 Project Overview

This project aims to build a predictive maintenance system using machine learning and data analytics. The system will:

• Store and process turbine engine sensor data.
• Predict engine failure using RUL calculations.
• Provide real-time KPIs through Power BI & Tableau dashboards.

📌 Project Structure

• SQL Database: Data storage, preprocessing, and querying.
• EDA & Feature Engineering: Sensor analysis and ML feature extraction.
• Predictive Models: RUL prediction using XGBoost & Random Forest.
• KPI Dashboard: Power BI/Tableau visualization for insights.

🔹 Data Sources & Methodologies

Component	Methodology	Data Sources	Tools Used
Data Ingestion	Convert NASA C-MAPSS .txt data to structured SQL tables.	NASA Turbofan Sensor Data (C-MAPSS)	Python (Pandas), SQL (PostgreSQL)
Data Cleaning & Structuring	Handle missing values, standardize types, index data for fast queries.	Preprocessed Sensor Readings	SQL, Python (Pandas)
Failure Threshold Estimation	Calculate statistical anomaly thresholds (95th percentile) & health index functions.	Sensor readings (temperature, pressure, speed)	SQL (PERCENTILE_CONT), Python (Z-score, IQR)
Remaining Useful Life (RUL) Prediction	Compute RUL dynamically based on last recorded cycles per engine.	Time-series cycles per engine	Python (Pandas), SQL
Predictive Maintenance Model	Machine Learning (XGBoost, Random Forest) on sensor degradation trends.	Historical failure data	Python (Scikit-learn)
KPI Dashboard & Reporting	Aggregate trends, visualize real-time failure predictions.	SQL-queried data	Power BI, Tableau

🔹 Key Performance Indicators (KPIs)

Metric	Definition	Purpose
Production Efficiency (%)	(Actual Output / Expected Output) × 100	Tracks overall engine performance.
Failure Rate (%)	(Failed Engines / Total Engines) × 100	Measures reliability of maintenance cycles.
Mean Time to Failure (MTTF)	Avg. cycles before engine failure	Estimates expected operational life.
Remaining Useful Life (RUL)	Cycles left before failure	Enables predictive maintenance.
Downtime Cost ($/hr)	Unplanned downtime cost per failure event	Justifies ROI of predictive maintenance.

🛠 Tech Stack & Tools

• SQL (PostgreSQL, MySQL) → Data storage & processing
• Python (Pandas, NumPy, Scikit-learn) → Data analysis & ML modeling
• Power BI / Tableau → Dashboard visualization
• Docker & API → Deployment & accessibility

🚢 Running the API & Frontend Stack

The repository ships with a FastAPI service (deployment/API_endpoint.py) and a Streamlit UI (frontend/streamlit_app.py). Both components are packaged together through the Dockerfile in deployment/.

Prerequisites

• Docker 24+
• (Optional) An external SQL database that exposes the engine_predictions, kpi_trends, and kpi_aggregates tables/views. If not supplied the stack falls back to the bundled sample dataset located at deployment/sample_data.json.

Environment Variables

Variable	Purpose	Default
PREDICTIVE_API_KEY	Shared secret expected in the X-API-Key header for /predict, /metrics, /reload.	dev-secret
PREDICTIVE_API_BASE_URL	Base URL the Streamlit client uses to reach the API.	http://localhost:8000
DATABASE_URL	Optional SQLAlchemy connection string for production data.	(unset)

Build & Run with Docker

docker build -t predictive-maintenance -f deployment/Dockerfile .
docker run \
  -p 8000:8000 \
  -p 8501:8501 \
  -e PREDICTIVE_API_KEY="your-strong-secret" \
  -e PREDICTIVE_API_BASE_URL="http://localhost:8000" \
  predictive-maintenance

• The FastAPI service is exposed at http://localhost:8000 (health check at /health).
• The Streamlit dashboard is available at http://localhost:8501.

To point the service at a managed database, supply a DATABASE_URL (e.g. postgresql+psycopg://user:pass@host:5432/db). The expected schema is documented in deployment/API_endpoint.py.

Authenticating Requests

All non-health endpoints require an API key supplied through the X-API-Key header. Use the same secret for local development and the container runtime:

curl \
  -H "X-API-Key: your-strong-secret" \
  http://localhost:8000/predict

If a key is not provided or mismatched, the service returns 401 Unauthorized.

Local Development

To run the stack without Docker make sure the dependencies in deployment/requirements.txt are installed, then execute in separate terminals:

uvicorn deployment.API_endpoint:app --reload --port 8000
streamlit run frontend/streamlit_app.py --server.port 8501

Update the PREDICTIVE_API_BASE_URL environment variable if the API is hosted elsewhere.

📝 Documentation & Portfolio Links

• GitHub Repository (Code & Implementation)
• Power BI/Tableau Dashboard (Live KPI Dashboard)
• Business Impact Report (Financial Justification)

👨‍💻 Author

Anthony Rodriguez
🔗 LinkedIn Profile
📧 Email

📜 License

This project is licensed under the MIT License.