Investigating creep damage initiation at the mesoscale using high-resolution electron microscopy, crystal plasticity modelling, and a classification algorithm

Accurate modelling of plastic and creep deformation, along with the associated damage mechanisms in 316H stainless steel under high-temperature and complex loading conditions, is essential for ensuring the long-term structural integrity of power plant components. Robust physics-based models contribute to more accurate life assessment procedures, thereby improving safety and extending component service life under creep conditions. However, current approaches often lack accurate microstructure-sensitive models that can correlate experimentally observed local creep damage with key microstructural features such as grain orientation and morphology in creep damage prediction. To address this knowledge gap, a combined modelling and experimental approach is employed to investigate creep damage initiation in 316H stainless steel at 550 °C. A crystal plasticity finite element (CPFE) model is developed to simulate the primary and secondary stages of creep deformation. To accurately predict local deformation under realistic boundary conditions, a new modelling strategy is introduced, embedding crystal plasticity domains within larger-scale geometries. Furthermore, a novel methodology is introduced to define damage initiation criterion by employing a classification algorithm to correlate experimentally observed creep damage with internal variables from the CPFE model. This data-driven approach enables the development of a predictive equation for identifying damaged grain boundaries. This equation represents a significant advancement over phenomenological approaches, such as the stress-modified ductility exhaustion (SMDE) model. The proposed model predicts approximately 67% of observed creep cavities at grain boundaries in the analysed regions, demonstrating the strong potential of a data-driven modelling framework for microstructure-sensitive damage prediction.