Causes of Rapid and Extensive Fracture Propagation After Fracture Coalescence

Abstract

Abstract The paper identifies specific spatial arrangement of the existing crack pathways and the crack network at the time of crack coalescence that lead to a rapid and extensive crack growth immediately after crack coalescence. To that end, we engineer relevant and informative features from simulated crack network generated using the HOSS simulator and then train machine learning model to relate the most significant crack-network features with the onset of rapid and extensive crack propagation immediately after crack coalescence. These features are indicators of the underlying mechanism of crack propagation and coalescence. The features serve as the inputs of a SVM classifier with RBF kernel to find the decision boundary between the two types of samples, those that undergo rapid and extensive crack propagation after coalescence and those that don't. Through permutation feature importance evaluation, we determine the 7 most important crack-network features associated with sudden, large crack growth. Notably, these features are associated with the status of energy buildup, energy distribution, and energy release inside the material. The primary aim of this study is to explain the sudden and extensive fracture-propagation post coalescence in terms of the KDE-based and subgraph-based features. The underlying assumption is that the stress and energy state inside the material cannot be directly observed; hence, the phenomenon of coalescence and large fracture propagation post-coalescence need to be answered in terms of the mechanisms observed based on the spatiotemporal evolution of the crack network under the uniaxial compression.