This project implements a Physics-Informed Neural Network (PINN) using TensorFlow to simulate and analyze the thermal performance enhancement of surfactant-enhanced ultra-thin vapor chambers (UTVC). The PINN integrates governing physical equations, including the continuity equation, Navier-Stokes equations, energy equation, and surfactant transport equation, to accurately model heat dissipation mechanisms within UTVCs under various surfactant conditions.
- 3D PINN Architecture: Designed to handle three-dimensional simulations for comprehensive analysis.
- Integration of Governing Equations: Enforces physical laws directly within the neural network's loss function.
- Detailed Boundary Conditions: Models evaporation, condensation, adiabatic, and convective boundaries accurately.
- Normalization and Denormalization: Ensures numerical stability and efficient training.
- Visualization Tools: This tool generates plots for temperature distribution, velocity fields, pressure distribution, and surfactant concentration at various time points.
- Optimized Training: Implements strategies to handle high-dimensional data efficiently, including gradient clipping and learning rate scheduling.
Ultra-thin vapor Chambers (UTVC) are critical components in thermal management systems, particularly in electronics cooling. Enhancing their thermal performance can significantly improve device reliability and efficiency. This project leverages PINNs to model the complex heat transfer and fluid dynamics within UTVCs, incorporating the effects of surfactants to optimize performance.
- Python 3.8 or higher
- TensorFlow 2.8.0 or higher
- NumPy
- Matplotlib
Ensure you have Python and pip installed on your system. It's recommended to use a virtual environment to manage dependencies.
-
Clone the Repository
git clone https://github.com/yourusername/pinn-utvc.git cd pinn-utvc
Efficient thermal management is critical in modern electronic devices, especially those housing high-performance components like CPUs and GPUs. Ultra-Thin Vapor Chambers (UTVCs) are widely used for dissipating heat effectively. However, the thermal performance of UTVCs can be further enhanced by modifying the working fluid properties. This project leverages Physics-Informed Neural Networks (PINNs) to simulate and analyze the thermal performance of Surfactant-Enhanced Ultra-Thin Vapor Chambers. By integrating various dilute surfactant-water mixtures, the model aims to optimize heat dissipation, ensuring optimal performance and longevity of electronic devices.
- 3D Physics-Informed Neural Network (PINN): Extends traditional PINNs to three dimensions, capturing the thickness of the UTVC for accurate thermal modeling.
- Surfactant Integration: Evaluates ten different dilute surfactant-water mixtures to assess their impact on surface tension reduction and heat transfer efficiency.
- Comprehensive Loss Function: Incorporates governing equations (Continuity, Navier-Stokes, Energy, and Surfactant Transport) into the loss function to ensure physically consistent predictions.
- Normalization and Scaling: Applies normalization techniques to inputs and outputs to stabilize training and improve convergence.
- Training Optimizations: Utilizes learning rate scheduling, gradient clipping, and a streamlined network architecture to reduce training time and enhance performance.
- Visualization: Generates clear visual comparisons of maximum and average temperatures across different surfactant mixtures, facilitating easy identification of optimal mixtures.
- Efficient Training: Optimized to run on GPUs, significantly reducing training time from hours to minutes where possible.
Results Upon training, the PINN predicts the thermal and fluid dynamics within the UTVC. The visualization plots provide insights into temperature distribution, velocity fields, pressure gradients, and surfactant concentrations at various time points, facilitating the analysis of thermal performance enhancements.
- Clone the Repository
git clone https://github.com/yourusername/utvc-pinn.git cd utvc-pinn
Usage Configure Surfactant Properties
The surfactant_data dictionary in pinn_model.py contains the properties of each surfactant mixture. Modify this section if you wish to add or update new mixtures.
Run the Simulation
Execute the main script to train the PINN and evaluate thermal performance:
bash Copy code python pinn_model.py View Results
After training, the script will generate bar charts comparing each surfactant mixture's maximum and average temperatures. These visualizations help identify the most effective surfactants for enhancing UTVC performance.
Results The simulation evaluates the effectiveness of each surfactant mixture in reducing maximum temperatures near the evaporation area and maintaining reasonable average temperatures across the UTVC. The results are visualized through bar charts, clearly comparing thermal performance enhancements.
Project Structure Copy code utvc-pinn/ │ ├── assets/ │ ├── utvc_simulation.png │ └── results_comparison.png │ ├── pinn_model.py ├── requirements.txt ├── README.md └── LICENSE assets/: Contains images for the README and other documentation. pinn_model.py: Main script implementing the 3D PINN model. requirements.txt: Lists all Python dependencies. README.md: Project documentation (this file). LICENSE: License information. Contributing Contributions are welcome! Your input is valuable, whether it's improving the model, adding new surfactant mixtures, or enhancing documentation.
Fork the Repository
bash Copy code git clone https://github.com/yourusername/utvc-pinn.git cd utvc-pinn Create a Feature Branch
bash Copy code git checkout -b feature/YourFeature Commit Your Changes
bash Copy code git commit -m "Add your message here" Push to the Branch
bash Copy code git push origin feature/YourFeature Open a Pull Request
License This project is licensed under the MIT License.
Acknowledgements Physics-Informed Neural Networks by Raissi, Perdikaris, and Karniadakis. TensorFlow for providing the deep learning framework. NumPy and Matplotlib for data handling and visualization.
Contact For any questions or suggestions, please open an issue or contact r02522318@gmail.com