LEFM

I would like to draw your attention to our recently proposed predictive method based on a semi-empirical model (LEFM) and Neural Network, exploiting the Physics-informed Machine Learning concept. We show how the accuracy of state-of-the-art fatigue predictive models, based on defects present in materials, can be significantly boosted by accounting for additional morphological features via Physics-Informed Machine Learning.

Journal: International Journal of Fracture

Abstract: A simple nonlocal field theory of peridynamic type is applied to model brittle fracture. The kinetic relation for the crack tip velocity given by Linear Elastic Fracture Mechanics (LEFM) is recovered directly from the nonlocal dynamics, this is seen both theoretically and in simulations. An explicit formula for the change of internal energy inside a neighborhood enclosing the crack tip is found for the nonlocal model and applied to LEFM.

Recent advance in adaptive remeshing techniques now gives the possibility to efficiently simulate complex 3D crack growth using conform meshing of the discontinuity. An exemple of such kind of mixed mode simulation with adpative remeshing can be seen on YouTube:

http://www.youtube.com/watch?v=uQrado7el8I