Integrated simulation, machine learning, and experimental approach to characterizing fracture instability in indentation pillar-splitting of materials
Measuring fracture toughness of materials at small scales remains challenging due to limited experimental testing configurations. A recently developed indentation pillar-splitting method has shown promise of improved flexibility in fracture toughness measurements at the microscale, partly due to the occurrence of an unusual fracture instability, i.e., a transition from stable to unstable crack propagation. In spite of growing interest in this method, the underlying mechanism of this phenomenon is yet to be elucidated. Furthermore, we provide a comprehensive description of fracture instability in indentation pillar-splitting by combining in situ experiments with high-fidelity simulations based on cohesive zone and J-integral methods. In addition, a machine-learning-based solution for predicting the critical indentation load of fracture instability is established through Gaussian processes regression for broad use of this method by the community.