Can I read A Machine Learning Material Model for Structural Adhesives in Finite Element Analysis on EtoBox?

A Machine Learning Material Model for Structural Adhesives in Finite Element Analysis by Sommer, Daniel (author);Haufe, André (author);Middendorf, Peter (author) is a Engineering article available to read on EtoBox.

What is A Machine Learning Material Model for Structural Adhesives in Finite Element Analysis about?

The computational analysis of adhesives within finite element simulations is important for many engineering applications, from automotive structures to civil engineering. The underlying material descriptions are closed, analytically formulated models representing the stress-response of the material for given strains. In this publication, we investigate and present a methodology for adopting artificial neural networks for this purpose. The aim is to show feasibility and performance of such models to be able to uncouple the material description from fixed formulations to reap the benefits of neural networks for materials modelling. A methodology is presented for creating training data for a structural adhesive, whereby the data is obtained from specimen and from generic strain paths of a baseline material model. From this, feed-forward neural networks are trained and evaluated using machine learning tools as well as from an engineering perspective by pre-defined test strain paths. It is shown that, given the quality and the quantity of training data is adequate, the model is able to generalize with respect to the response. Especially, the coverage of complete strain space is shown to

Who reads A Machine Learning Material Model for Structural Adhesives in Finite Element Analysis?

Author

Sommer, Daniel (author);Haufe, André (author);Middendorf, Peter (author)

Publisher

Elsevier BV

Published

2022

Language

EN

Field

Engineering (Physical Sciences)