Skip to content

Physics Informed Neural Networks (PINNs) is a machine learning technique that incorporates physical laws and constraints into the neural network training process for solving partial differential equations (PDEs) in various fields of science and engineering, including solid mechanics.

License

Notifications You must be signed in to change notification settings

arre-ankit/Physics-Informed-Neural-Networks-PINNs-

Repository files navigation

Physics Informed Neural Networks(PINNs)

Physics Informed Neural Networks (PINNs) is a machine learning technique that incorporates physical laws and constraints into the neural network training process for solving partial differential equations (PDEs) in various fields of science and engineering, including solid mechanics.

In this repository PINN model are integrated in the Timoshenko equation to predict the deflection of the beam at any distance and time and compared the results with those obtained through the traditional finite element method.

Timoshenko Eqution : https://en.wikipedia.org/wiki/Timoshenko%E2%80%93Ehrenfest_beam_theory

Research Papers on PINNs :

  1. https://faculty.sites.iastate.edu/hliu/files/inline-files/PINN_RPK_2019_1.pdf
  2. https://arxiv.org/pdf/2303.01055

Setup:

$ git clone https://github.com/lululxvi/deepxde.git
$ cd deepxde/docker
$ docker build -f Dockerfile . -t deepxde
docker run -v $(pwd):/root/shared -w "/root/shared" -p 8888:8888 deepxde

Add the file then run the Jupyter Notebook

About

Physics Informed Neural Networks (PINNs) is a machine learning technique that incorporates physical laws and constraints into the neural network training process for solving partial differential equations (PDEs) in various fields of science and engineering, including solid mechanics.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published