https://github.com/sinhaparth5/heat-transfer-cooling-fin-model
A Physics-Informed Neural Network (PINN) implementation for solving 1D transient heat transfer in aluminum cooling fins.
https://github.com/sinhaparth5/heat-transfer-cooling-fin-model
physics pinns
Last synced: 10 months ago
JSON representation
A Physics-Informed Neural Network (PINN) implementation for solving 1D transient heat transfer in aluminum cooling fins.
- Host: GitHub
- URL: https://github.com/sinhaparth5/heat-transfer-cooling-fin-model
- Owner: sinhaparth5
- License: apache-2.0
- Created: 2025-08-11T18:53:08.000Z (11 months ago)
- Default Branch: master
- Last Pushed: 2025-08-11T20:11:27.000Z (11 months ago)
- Last Synced: 2025-08-11T21:12:59.364Z (11 months ago)
- Topics: physics, pinns
- Language: Python
- Homepage:
- Size: 82 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Physics-Informed Neural Network for Cooling Fin Heat Transfer
[](https://www.python.org/)
[](https://tensorflow.org/)
A Physics-Informed Neural Network (PINN) implementation for solving 1D transient heat transfer in aluminum cooling fins.
## Problem
Solves the 1D heat equation with convective losses:
```
ρcp ∂T/∂t = k ∂²T/∂x² - hP(T - T∞)
```
**Conditions:**
- Initial: T(x,0) = 373K (100°C)
- Base: T(0,t) = 373K (fixed)
- Tip: Convective cooling boundary
## Quick Start
```bash
pip install tensorflow numpy matplotlib seaborn scipy
python cooling_fin_pinn.py
```
## Results
- ✅ 5+ orders of magnitude loss reduction
- ✅ Physical temperature evolution (373K → 356K at tip)
- ✅ Smooth, continuous solution
- ✅ All boundary conditions satisfied
## Key Features
- **Meshless**: No grid discretization needed
- **Physics-embedded**: Heat equation built into loss function
- **Automatic differentiation**: TensorFlow handles gradients
- **Multiple visualizations**: Temperature evolution, heatmaps, 3D plots
## Network Architecture
```python
Input: [x, t] → Hidden: 4×64 neurons → Output: T(x,t)
Loss = PDE_residual + IC_loss + BC_loss
```
## Files Generated
- `cooling_fin_temperature_evolution.png` - Spatial temperature profiles
- `cooling_fin_heatmap.png` - 2D temperature field
- `cooling_fin_training_convergence.png` - Loss evolution
- `cooling_fin_temporal_evolution.png` - Time series at fixed positions
- `cooling_fin_3d_surface.png` - 3D visualization
## Applications
- Electronics cooling (CPU heat sinks)
- Automotive thermal systems
- Heat exchanger design
- Thermal optimization
## License
MIT