https://github.com/mariamabidi/pinn-based-flow-prediction

This repository contains code and experiments for predicting 3D aerodynamic flow around car geometries using Physics-Informed Neural Networks (PINNs) and for analyzing flow features via autoencoder-based clustering.
https://github.com/mariamabidi/pinn-based-flow-prediction

computer-vision machine-learning neural-network numpy pytorch pyvista scikit-learn

Last synced: 12 months ago
JSON representation

This repository contains code and experiments for predicting 3D aerodynamic flow around car geometries using Physics-Informed Neural Networks (PINNs) and for analyzing flow features via autoencoder-based clustering.

• Host: GitHub
• URL: https://github.com/mariamabidi/pinn-based-flow-prediction
• Owner: mariamabidi
• Created: 2025-03-28T19:00:49.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2025-07-02T18:08:34.000Z (about 1 year ago)
• Last Synced: 2025-07-29T08:50:56.593Z (12 months ago)
• Topics: computer-vision, machine-learning, neural-network, numpy, pytorch, pyvista, scikit-learn
• Language: Python
• Homepage:
• Size: 43 KB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# PINN-Based Aerodynamic Flow Prediction and Clustering 🚗

This repository contains code and experiments for predicting 3D aerodynamic flow around car geometries using Physics-Informed Neural Networks (PINNs) and for analyzing flow features via autoencoder-based clustering.

## Features

- ✅ **Physics-Informed Neural Networks (PINNs)**  

  Predict 3D velocity fields around car shapes without full CFD simulations.

- ✅ **Streamline Visualization**  

  Visualize predicted flow fields and identify regions of interest.

- ✅ **Autoencoder for Feature Compression**  

  Reduce high-dimensional CFD data to meaningful latent representations.

- ✅ **KMeans Clustering in Latent Space**  

  Detect and classify distinct aerodynamic zones like wakes, stagnation points, and freestream regions.

- ✅ **3D Visualization of Clusters**  

  Overlay cluster labels on mesh geometry for intuitive interpretation.

## Use Cases

- Aerodynamic analysis and design exploration  

- Data-driven identification of critical flow regions  

- Reducing reliance on computationally expensive CFD runs

## Technologies

- PyTorch  

- PyVista  

- scikit-learn  

- NumPy

## Authors

- Mariam Abidi — PINN & Autoencoder & clustering implementation 

- Suhas Vittal — PINN implementation, streamline visualizations 

- Nishith Hingoo — Dataset sourcing, preprocessing pipelines

---

> This project demonstrates the feasibility of physics-guided machine learning for aerodynamic analysis and provides a framework for faster, simulation-free flow predictions and feature detection.