Intel Digital Twin for Predictive Maintenance

Overview

This workspace contains a collection of projects and prototypes for building a Smart Digital Twin platform (Technovate) that predicts industrial equipment failures before they occur. The system simulates machines, streams sensor data, applies AI for predictions, and provides a dashboard for monitoring. Features are split across multiple folders, requiring integration for a complete solution.

The platform focuses on predictive maintenance for machines like robotic arms and aircraft engines, using physics-based simulations and machine learning.

What is Done

• Simulation Engines: ROS2 and Gazebo setup for robotic digital twins (in digital_twin_robot/robot_digital_twin/).
• AI Models: Deep learning models for aircraft engine diagnostics and prognostics (TensorFlow/Keras in digital-twin-for-aircraft-engine-maintenance/).
• ML Pipelines: Predictive maintenance scripts using scikit-learn and XGBoost (in predictive-maintenance/ and Predictive-Maintenance-for-Industrial-Equipment/).
• Frontend Prototype: React-based dashboard with Tailwind CSS for monitoring and controls (in digital_twin_robot/frontend/).
• Data Handling: CSV/JSON export, feature engineering, and exploratory data analysis (notebooks across folders).
• Hybrid Modeling: Python package for digital twins with physics-based and ML components (in Digital-Twin-in-python/).
• Visualization: Charts and plots using Plotly, Matplotlib, and Recharts.

What Needs to be Done

• Integration: Combine simulation (ROS/Gazebo), AI models, and frontend into a single platform.
• Backend API: Implement FastAPI endpoints to connect simulation data to AI predictions and serve the frontend.
• Real-Time Streaming: Enable live sensor data flow from simulation to dashboard.
• Model Deployment: Deploy AI models for inference (e.g., using TensorFlow Serving).
• Database: Add persistent storage for sensor logs and predictions.
• Testing: Unit and integration tests for reliability (basic structure in Digital-Twin-in-python/tests/).
• Documentation: Update guides for setup and usage.

What is Not Done Yet

Full Digital Twin Simulation: Complete integration of Gazebo/ROS with AI for real-time predictions.

Advanced Explainable AI: SHAP or similar for model interpretability.

Fault Injection: Tools for simulating failures in the digital twin.

Historical Playback: Time-scrubbing for past data analysis.

Multi-Machine Support: Extend beyond single machine types (e.g., support for CNC machines).

Intel Technologies Used

"Intel" here refers to intelligent technologies (AI/ML and simulation tools) for building smart systems. The following are the key tools and technologies employed:

AI/ML Frameworks

• TensorFlow/Keras: Deep learning for diagnostics, prognostics, and sequential modeling (e.g., RUL prediction).
• Scikit-learn: Traditional ML for classification, regression, anomaly detection (e.g., RandomForest, Isolation Forest).
• XGBoost: Gradient boosting for failure prediction and RUL estimation.
• PCA: Dimensionality reduction for feature engineering.

Simulation and Modeling

• ROS2: Robotic simulation and sensor data handling.
• Gazebo: Physics-based simulation engine for digital twins.
• OpenModelica: System dynamics modeling (state-space equations).
• MATLAB/Simulink: Multibody dynamics and control systems.
• URDF: 3D robot model definitions.

Frontend and Visualization

• React.js: Component-based UI for dashboards.
• Tailwind CSS: Utility-first styling.
• Recharts/Plotly.js/Chart.js: Interactive charts for sensor data and predictions.
• Streamlit: Rapid prototyping for ML apps.

Backend and Data

• FastAPI: Planned for REST APIs (Python-based).
• Pandas/NumPy: Data manipulation and analysis.
• Hydra-core/Pydantic: Configuration and validation.
• MLflow: Experiment tracking for ML models.

Other Tools

• Jupyter Notebooks: For EDA, model training, and evaluation.
• Docker: Containerization for deployment.
• Poetry/Pip: Dependency management.

How to Run/Combine

1. Setup Environment: Install Python 3.8+, Node.js. Use pip install -r requirements.txt from relevant folders.
2. Run Simulations: Navigate to digital_twin_robot/robot_digital_twin/ and launch ROS/Gazebo.
3. Train Models: Execute notebooks in digital-twin-for-aircraft-engine-maintenance/ for AI models.
4. Start Frontend: cd digital_twin_robot/frontend && npm install && npm run dev.
5. Integrate: Build a FastAPI backend to bridge simulation and AI, then connect to React frontend.
6. Test: Run scripts in predictive-maintenance/ or the Streamlit app.

For full integration, combine components into a monolithic app or microservices architecture.

Contributing

Features are modular; contribute to specific folders and propose merges for integration.

License

Varies by folder; check individual READMEs.