Work

A Neural Network Constitutive Model and Automatic Stiffness Evaluation for Multiscale Finite Elements

Year: 2025

Type: article

Abstract: A neural network model for a constitutive law in nonlinear structures is proposed. The neural model is constructed based on a data set of responses of representative volume elements, calculated by finite elements. An open scientific software machine learning platform Tensorflow and an application programming interface, intended for a deep learning Keras library, provided by Python are used for the development of the artificial neural network. The tangential stiffness matrix within a multi-scale model is calculated via the method of automatic differentiation of Tensorflow. The results are compared with given data set. The loss function, including the Sobolev metrics is computed. The results can be integrated into a multiscale finite element analysis and provide results with less effort. The technique is also tested on hyperelastic materials.

Source: Applied Sciences

DOI: 10.3390/app15073697

Authors: Aliki D. Mouratidou, Georgios E. Stavroulakis

Institution: Technical University of Crete

Language: English

Topic: Model Reduction and Neural Networks

Subfield: Statistical and Nonlinear Physics

Field: Physics and Astronomy

Domain: Physical Sciences

Open Access status: gold

Estimated APC paid: $2,896

Funder: European Commission