Welcome to PINN4GPR’s documentation!

This project enables you to:

1. Generate Ground Penetrating Radar (GPR) datasets of realistic railway track configurations with gprMax.

2. Train a CNN-based surrogate model for gprMax on the generated data, which at inference time is two orders of magnitude faster than FDTD simulations.

3. Use the surrogate model for faster large-scale dataset generation.

4. Explore the use of physics-informed neural networks (PINNs) for the approximation of GPR wavefield data in complex railway track geometries.

Getting started

• Installation
  • Prerequisites
  • Quick install
  • 3D ballast simuation
• Tutorial
  • Dataset creation
  • CNN surrogate model training
  • PINN models training

API reference

• dataset_creation package
  • create_dataset module
  • geom2bscan module
  • inputfile module
  • ballast_simulation module
  • convert_to_np module
  • configuration module
  • statistics module
• pinns package
• tests package
• visualization package
  • initial_comparison module
  • field module

Indices and tables

• Index

• Module Index