Linel2D-Net

Introduction

The Linel2D-Net emulates a solution of an arbitrary 2D linear elastic, quasi-geometric boundary value problem defined on a 128x128 image domain using a U-Net model trained on a large set of "exact" FDM solutions.

Key Features

The Linel2D-Net tool is called from the command line as follows:

Forward model, Linux:
Linel2DNet_fwd "input" "output" "models/fwdMODEL_PR025.h5"

Forward model, Windows:
Linel2DNet_fwd.exe "input" "output" "models\fwdMODEL_PR025.h5"

Inverse model, Linux:
Linel2DNet_inv "input" "output" "models/invMODEL_PR025.h5"

Inverse model, Windows:
Linel2DNet_inv.exe "input" "output" "models\invMODEL_PR025.h5"

whereas *.h5 are models pre-trained using FDM forward and inverse solutions of Lame-Navier PDE with the Poisson's ratio = 0.25. The "input" folders in the case of forward models contain three subfolders with examples of geometrical boundary conditions (BC) and XY components of the displacements (CompX, CompY) defined on non-zero pixels from 128x128 BC images. Forward solutions of 2D linear elastic BVPs (i.e. XY components of the displacement field for the entire image domains) are written out into predX and predY subfolders of the "output" folder. In the case of inverse models, "input" folders contain the components of the displacement field for the entire images, and "output" folders contain three subfolders for predicted boundary conditions (BC) as well as displacement components defined on BC regions. The structure of input and output folders is fixed and should not be changed. Users can compute solution of their own 2D linear elastc BVP by additing new boundary condition (BC) and displacmeent images to BC nad CompX/CompY subfolders of the intput folder. Thereby BC images have to be TIFF images with values 0 (free pixels) or 1 (Dirichlet pixels with prescribed displacements), and CompX and CompY are floating point TIFF images indicating the values of XY displacement components on non-zero (Dirichlet) regions of BC images, see examples in the "input" folder.

Downloads

The Linel2D-Net tool for Linux and Windows OS with example data can be downloaded using the links below:

Forward models:
Download the Linel2D-Net forward tool for Windows x64 + examaple data(3GB)
Download the Linel2D-Net forward tool for Linux + examaple data (3GBB)

Inverse models:
Download the Linel2D-Net inverse tool for Windows x64 + examaple data(3GB)
Download the Linel2D-Net inverse tool for Linux + examaple data (3GBB)

By downloading, installing and/or using this software the user agrees with the following license conditions.

References

[1] Antony et al. (2023) Linel2D-Net: a PINN approach to solving 2D linear elastic boundary value problems on image domains.

Leibniz Institute of Plant Genetics and Crop Plant Research (IPK)
OT Gatersleben
Corrensstrasse 3
06466 Stadt Seeland,
Telefon: +49 (0)39482 50
Telefax: +49 (0)39482 5139
E-Mail: info@ipk-gatersleben.de
Internet: www.ipk-gatersleben.de