4DCT Observation of crack evolution in coal under continuous uniaxial compression

To address the limitations of current real-time CT technology in capturing dynamic crack evolution throughout the entire uniaxial compression process of rock and the lack of efficient crack extraction methods, 4DCT approach is developed to accomplish the continuous observation of crack evolution inside coal samples under continuous uniaxial compression. This study proposes a temporal crack dynamic tracking model that integrates the single-image crack segmentation network with a multi-image associative memory method. Based on the Atkinson-Shiffrin memory architecture, the model establishes a three-level storage system comprising sensory memory, working memory, and long-term memory. Combined with segment voting denoising and merged propagation mechanisms, it achieves precise correlation and dynamic tracking of cracks in long-sequence and high-noise CT images. Utilizing a self-developed 4DCT system, continuous uniaxial compression and real-time CT scanning of coal samples are performed to obtain dynamic evolution sequences of internal cracks in coal during the complete stress-strain process. A microscopic damage constitutive model is established based on porosity evolution. This model successfully quantifies the mechanical property degradation caused by crack development and accomplishes accurate correlation between microscopic crack propagation and macroscopic stress-strain responses. The chain regulation mechanism of crack evolution and the heterogeneous progressive failure is revealed.