A combined finite element–finite volume framework for phase-field fracture

Juan Michael Sargado,Eirik Keilegavlen,Inga Berre,Jan Martin Nordbotten

doi:10.1016/j.cma.2020.113474

Juan Michael Sargado, Eirik Keilegavlen + Show 2 more

Open Access

https://doi.org/10.1016/j.cma.2020.113474

Copy DOI

Abstract

Numerical simulations of brittle fracture using phase-field approaches often employ a discrete approximation framework that applies the same order of interpolation for the displacement and phase-field variables. In particular, the use of linear finite elements to discretize both stress equilibrium and phase-field equations is widespread in the literature. However, the use of P1 Lagrange shape functions to model the phase-field is not optimal, as the latter contains cusps for fully developed cracks. These should in turn occur at locations corresponding to Gauss points of the associated FE model for the mechanics. Such a feature is challenging to reproduce accurately with low order elements, and element sizes must consequently be made very small relative to the phase-field regularization parameter in order to achieve convergence of results with respect to the mesh. In this paper, we combine a standard linear FE discretization of stress equilibrium with a cell-centered finite volume approximation of the phase-field evolution equation based on the two-point flux approximation constructed over the same simplex mesh. Compared to a pure FE formulation utilizing linear elements, the proposed framework results in looser restrictions on mesh refinement with respect to the phase-field length scale. This ability to employ coarser meshes relative to the traditional implementation allows for significant reductions on computational cost, as demonstrated in several numerical examples.

Highlights

Brittle fracture is an important failure mechanism against which engineering structures must be properly designed to ensure their safety
While the initial intent of [18] was for the regularized energy functional to approximate a body with discrete cracks through the concept of Γ -convergence [21], lately it has come to be understood that variational phase-field models are a subset of gradient-enhanced damage models, and that the phase-field regularization parameter should be related to the intrinsic length scale of the material [22,23,24,25]
We propose instead to combine a finite element approximation of the linear momentum equation using P1 shape functions with a cell-centered finite volume scheme for the phase-field subsystem based on the classical two-point flux approximation (TPFA)

Summary

Introduction

Brittle fracture is an important failure mechanism against which engineering structures must be properly designed to ensure their safety. Griffith [1] was the first to formulate a theory of brittle fracture based on thermodynamic arguments, utilizing earlier results by Inglis [2] concerning stresses around elliptical holes. This theory was later amended to include the effect of plastic zones around crack tips by Irwin, who introduced the notion of stress intensity factors [3]. Together, these form the basis of classical linear elastic fracture mechanics (LEFM).

Methods

Results

Conclusion