Learning to fail: Predicting fracture evolution in brittle material models using recurrent graph convolutional neural networks
This addresses a key challenge in materials science by enabling faster predictions of material failure, though it is incremental as it builds on existing simulation data and deep learning methods.
The paper tackles predicting fracture propagation and failure in brittle materials under stress using a machine learning approach, achieving results within 3% for fracture damage and length and 15% for time to failure compared to simulations, with predictions generated in seconds instead of hours.
We propose a machine learning approach to address a key challenge in materials science: predicting how fractures propagate in brittle materials under stress, and how these materials ultimately fail. Our methods use deep learning and train on simulation data from high-fidelity models, emulating the results of these models while avoiding the overwhelming computational demands associated with running a statistically significant sample of simulations. We employ a graph convolutional network that recognizes features of the fracturing material and a recurrent neural network that models the evolution of these features, along with a novel form of data augmentation that compensates for the modest size of our training data. We simultaneously generate predictions for qualitatively distinct material properties. Results on fracture damage and length are within 3% of their simulated values, and results on time to material failure, which is notoriously difficult to predict even with high-fidelity models, are within approximately 15% of simulated values. Once trained, our neural networks generate predictions within seconds, rather than the hours needed to run a single simulation.