/ANN-FEAD-2021

Scripts for the ANN publication submited to FEAD 2021

Primary LanguageJupyter Notebook

ANN VUHARD repository

This repository contains datafiles related to the publication:

Efficient Implementation of Non-linear Flow Law Using Neural Network into the Abaqus Explicit FEM code

submitted to : Finite Elements in Analysis and Design journal.

Content of the directories

Main directory

  • ANN-Learning.ipynb : Jupyter notebook main training program. This one build and train the ANN.

  • PythonToFortran-3layers.ipynb : Jupyter notebook converter to write the VUHARD from the ANN parameters. This one generates the FORTRAN subroutines for Abaqus Explicit.

ANN-JohnsonCook

This directory contains the following files:

  • Datatest.npz : numpy datafile containing the testing data without derivatives
  • DatatestWithDerivatives.npz : numpy datafile containing the testing data with derivatives
  • JC-Experiments : excel file containing training data for the ANN

that can be used to create new ANN.

VUHARD

Contents the following FORTRAN files which are pre-trained ANN ready to use:

  • VUHARD-ANN-3-15-7-1-sigmoid.f : ANN 3-15-7-1 as proposed in the paper
  • VUHARD-ANN-3-7-4-1-sigmoid.f : ANN 3-15-7-1 as proposed in the paper
  • VUHARD.f : Analytical model of the VUHARD

BarNecking

Built-in model

Built-in model is defined by the BarNecking.inp file for Abaqus. Running the model can be done using the following command:

abaqus job=BI input=BarNecking double=both cpus=2 mp_mode=threads

Analytical VUHARD model

Analytical VUHARD model is defined by the BarNecking_VH.inp file for Abaqus. Running the model can be done using the following command:

abaqus job=VH input=BarNecking_VH user=../VUHARD/VUHARD.f double=both cpus=2 mp_mode=threads

ANN VUHARD models

ANN VUHARD model is defined by the BarNecking_VH.inp file for Abaqus. Running the model can be done using the following commands:

abaqus job=ANN-1 input=BarNecking_VH user=../VUHARD/VUHARD-ANN-3-7-4-1-sigmoid.f double=both cpus=2 mp_mode=threads
abaqus job=ANN-2 input=BarNecking_VH user=../VUHARD/VUHARD-ANN-3-15-7-1-sigmoid.f double=both cpus=2 mp_mode=threads

Taylor

Built-in model

Built-in model is defined by the Taylor.inp file for Abaqus. Running the model can be done using the following command:

abaqus job=BI input=Taylor double=both cpus=2 mp_mode=threads

Analytical VUHARD model

Analytical VUHARD model is defined by the Taylor_VH.inp file for Abaqus. Running the model can be done using the following command:

abaqus job=VH input=Taylor_VH user=../VUHARD/VUHARD.f double=both cpus=2 mp_mode=threads

ANN VUHARD models

ANN VUHARD model is defined by the Taylor_VH.inp file for Abaqus. Running the model can be done using the following commands:

abaqus job=ANN-1 input=Taylor_VH user=../VUHARD/VUHARD-ANN-3-7-4-1-sigmoid.f double=both cpus=2 mp_mode=threads
abaqus job=ANN-2 input=Taylor_VH user=../VUHARD/VUHARD-ANN-3-15-7-1-sigmoid.f double=both cpus=2 mp_mode=threads

Olivier Pantalé
Full Professor of Mechanics
email : olivier.pantale@enit.fr

Laboratoire Génie de Production
Ecole Nationale d'Ingénieurs de Tarbes
Université de Toulouse
47 Avenue d'Azereix - BP 1629
65016 TARBES - CEDEX - FRANCE