DLRBniCSx

PyTorch-RBniCSx-FEniCSx based open source library for deep learning based reduced order modelling.

Dependencies

• DOLFINx (v0.6.0)
• RBniCSx

  pip install git+https://github.com/RBniCS/RBniCSx.git
  

• ufl4rom

  pip install git+https://github.com/RBniCS/ufl4rom.git
  

• MDFEniCSx
• DLRBniCSx

  pip install git+https://github.com/Wells-Group/dlrbnicsx.git
  

• pytorch

  pip install torch
  

Other

• plotly
• Matplolib

Overview

The finite element calculations are performed using dolfinx. We use RBniCSx for Proper Orthogonal Decomposition (POD) and construction of reduced basis dataset. Once the dataset has been constructed, typical workflow in DLRBniCSx is as follow:

• Create training dataset and validation dataset using CustomDataset
• Use datasets DataLoader as train_loader and valid_loader for easy access to samples
• Initialise neural network model using HiddenLayersNet
• Use train_loader for training of the neural network using train_nn function and valid_loader for validation of the neural network using validate_nn function
• Perform error analysis using error_analysis function
• Compute reduced basis solution at a given online parameter using online_nn function

Disclaimer

In downloading this SOFTWARE you are deemed to have read and agreed to the following terms: This SOFT- WARE has been designed with an exclusive focus on civil applications. It is not to be used for any illegal, deceptive, misleading or unethical purpose or in any military applications. This includes ANY APPLICATION WHERE THE USE OF THE SOFTWARE MAY RESULT IN DEATH, PERSONAL INJURY OR SEVERE PHYSICAL OR ENVIRONMENTAL DAMAGE. Any redistribution of the software must retain this disclaimer. BY INSTALLING, COPYING, OR OTHERWISE USING THE SOFTWARE, YOU AGREE TO THE TERMS ABOVE. IF YOU DO NOT AGREE TO THESE TERMS, DO NOT INSTALL OR USE THE SOFTWARE.