Grant: Discovery Grants Program - Individual

The propagation of short cracks significantly influences the overall lifespan in high cycle fatigue (HCF) and very high cycle fatigue (VHCF) of structural components. Despite extensive studies, no effective damage modeling approach has been developed to comprehensively understand the complex behavior of short cracks. Current methods face challenges due to their limitations in capturing non-local material damage near growing short cracks. To bridge this knowledge gap, our research project aims to enhance the understanding of short crack behavior in HCF and VHCF regimes through advanced experimental and novel modeling studies. The primary objective is to lay the foundation for the development of physics-based peridynamics and machine learning models using multi-scale mechanistic data from propagating cracks. Multi-scale mechanistic field data near a growing crack are critical to quantitative damage assessments of short crack propagation behavior. The intricate interplay between crack driving force variables, mechanical deformation fields, and microstructure results in the irregular behavior of short crack growth. We will collect high-resolution crack propagation data using advanced measurement techniques to investigate multi-length scales damage mechanisms in the short crack growth regime. A recently proposed peridynamics theory will be employed to develop new peridynamics-based damage model(s). Peridynamics considers fatigue damage based on a non-local damage concept beyond a material point at the crack tip. Peridynamics, with its capacity to model material behavior beyond classical continuum mechanics, shows promise in capturing the complex behavior of short crack propagation. Machine learning, on the other hand, can enhance the computational efficiency and predictive capabilities of the modeling framework by learning from experimental and synthetic datasets. Peridynamic models will be correlated with experimental data to generate synthetic datasets for training machine learning models like Graph Convolutional Networks (GCN) and Convolutional Neural Networks (CNN). This integration will allow physics-based constraints, leading to accurate predictions of short crack growth by considering underlying physics and mechanistic field interactions. As peridynamics and machine learning advance, this research may provide groundbreaking insights into structural integrity analysis, benefiting various industries such as aerospace, automotive, transportation, and defense industries. It promotes model-based design practices, leading to optimized engineering component design. The results will be shared with the scientific community through peer-reviewed journals and national/international conferences. This research also fosters innovation and excellence among diverse student teams at Concordia University, preparing the next generation of engineers and researchers highly skilled in cutting-edge modeling techniques for structural component design.