2D Unsteady Heat Diffusion (PINNs + FDM)

📝 Overview

This project explores the 2D unsteady heat diffusion equation in aluminum using:

• Physics-Informed Neural Networks (PINNs) (two different versions)
• Finite Difference Method (FDM) for comparison

We implement deep learning models that respect the heat equation as a constraint, then compare their accuracy and performance to a classical numerical solver.

🧑‍💻 What’s Included

• square.ipynb — First version of the PINN model
• Final_void.ipynb — Improved version of the PINN model
• data/ — Input files (e.g. .xlsx)
• figures/ — Plots and screenshots of results
• requirements.txt — Python dependencies

⚙️ Installation & Usage

1. Clone this repository:

   git clone https://github.com/Adi-0202/PINN-Heat-Diffusion.git
   cd PINN-Heat-Diffusion

2. Install dependencies:

   pip install -r requirements.txt

3. Run the scripts:

   Final_void.ipynb
   square.ipynb

4. Results: Visualizations and screenshots of results can be found in the figures/ directory. Examples: PINNs solutions vs. FDM baseline Error and convergence plots
5. Acknowledgement: Thanks to Professor Dr.Kritesh Gupta for guidance and support. 6.Keywords: Physics-Informed Neural Networks, Heat Diffusion, Finite Difference Method, Deep Learning, Numerical Methods