Abstract
In this work, we describe our contribution to the Purdue-SANDIA-LLNL Damage Mechanics Challenge. The phase field fracture model is adopted to blindly estimate the failure characteristics of the challenge test, an unconventional three-point bending experiment on an additively manufactured rock resembling a type of gypsum. The model is formulated in a variationally consistent fashion, incorporating a volumetric–deviatoric strain energy decomposition, and the numerical implementation adopts a monolithic unconditionally stable solution scheme. Our focus is on providing an efficient and simple yet rigorous approach capable of delivering accurate predictions based solely on physical parameters. Model inputs are Young’s modulus E, Poisson’s ratio ν, toughness Gc and strength σc (as determined by the choice of phase field length scale ℓ). We show that a single mode I three-point bending test is sufficient to calibrate the model, and that the calibrated model can then reliably predict the force versus displacement responses, crack paths and surface crack morphologies of more intricate three-point bending experiments that are inherently mixed-mode. Importantly, our peak load, crack trajectory and crack surface morphology predictions for the challenge test, submitted before the experimental data was released, show a remarkable agreement with experiments. The characteristics of the challenge, and how changes in these can impact the predictive abilities of phase field fracture models, are also discussed.
•The phase field fracture model is used to deliver blind failure predictions.•The model takes as input only Young’s modulus, Poisson’s ratio, toughness and strength.•A simple, mode I three-point bending test suffices to calibrate model parameters.•Predictions show a remarkable agreement with all other tests, which are more intricate and mixed-mode.•The agreement is excellent across all data: crack paths, peak load and crack morphology.