A variational asymmetric phase-field model of quasi-brittle fracture: Energetic solutions and their computation

TL;DR

This paper develops a variational asymmetric phase-field model for quasi-brittle fracture, incorporating energetic solutions and a novel numerical scheme that improves crack propagation predictions by including two-sided energetic inequalities.

Contribution

It introduces a new variational formulation with energetic solutions for fracture modeling, enabling more accurate and consistent numerical simulations of crack evolution.

Findings

The model captures different behaviors at traction and compression.

Including energetic inequalities improves the description of crack initiation.

The approach is validated on 2D and 3D benchmark problems.

Abstract

We derive the variational formulation of a gradient damage model by applying the energetic formulation of rate-independent processes and obtain a regularized formulation of fracture. The model exhibits different behavior at traction and compression and has a state-dependent dissipation potential which induces a path-independent work. We will show how such formulation provides the natural framework for setting up a consistent numerical scheme with the underlying variational structure and for the derivation of additional necessary conditions of global optimality in the form of a two-sided energetic inequality. These conditions will form our criteria for making a better choice of the starting guess in the application of the alternating minimization scheme to describe crack propagation as quasistatic evolution of global minimizers of the underlying incremental functional. We will apply the procedure for two- and three-dimensional benchmark problems and we will compare the results with the solution of the weak form of the Euler-Lagrange equations. We will observe that by including the two-sided energetic inequality in our solution method, we describe, for some of the benchmark problems, an equilibrium path when the damage starts to manifest, which is different from the one obtained by solving simply the stationarity conditions of the underlying functional.