/LSWR_loss_function_PINN

A kind of loss function based on Least Squares Weighted Residual method for computational solid mechanics

Primary LanguagePython

LSWR_loss_function_PINN

This repository provides numerical examples of the Least Squares Weighted Residual loss function for physics-informed neural network-based computational solid mechanics framework.

Numerical examples include:

  • 2D pure bending beam problem
  • 2D in-plain stretching preforated plate problem
  • 3D streching cube with a sphere hollow problem

Information regarding the numerical examples, please refer to our paper.

Enviornmental settings

  • TensorFlow 2.8.0
  • Keras 2.8.0
  • SciPy 1.8.0

Paper link

https://doi.org/10.1007/s00466-022-02252-0

Cite as

[1] J. Bai, T. Rabczuk, A. Gupta, L. Alzubaidi, Y. Gu, A Physics-Informed Neural Network technique based on a modified loss function for computational 2D and 3D solid mechanics, Computational Mechanics, (2022).