Abstract: To address the limitations of traditional long-distance running injury assessment methods in quantifying tissue cumulative load and mechanical fatigue effects, this study integrates personalized musculoskeletal system modeling, fatigue failure mechanics analysis, and deep physics-informed neural networks to establish a biomechanical research framework spanning from the elucidation of mechanical mechanisms to the prediction of relative risk. Focusing on tissue damage evolution under continuous impact loading, this research aims to systematically elucidate the synergistic regulatory mechanisms of foot biomechanical characteristics and running distance on tissue fatigue damage, thereby providing a new theoretical foundation for the mechanical risk assessment of long-distance running injuries. The study first constructs an anatomically specific lower limb musculoskeletal model based on individual medical imaging data. It incorporates a nonlinear viscoelastic constitutive relationship to characterize dense connective tissues such as tendons, enabling accurate depiction of their time-dependent mechanical behaviors, including stress relaxation and creep, under cyclic loading. Subsequently, a tissue-level fatigue failure mechanics model is established based on cumulative damage theory. This model quantitatively reveals a distinct three-stage nonlinear evolution pattern in damage accumulation for the plantar fascia, Achilles tendon, tibia, and patellofemoral joint during long-distance running. A critical “fatigue-sensitive window” is identified within the 11-16 km running distance range. Quantitative analysis further demonstrates that the Foot Posture Index (FPI) changes significantly with running distance and exhibits a highly nonlinear positive correlation with the fatigue failure probability of each tissue. Furthermore, to enable dynamic assessment of injury risk, this study innovatively embeds the fatigue failure mechanics mechanism into a deep physics-informed neural network. By incorporating the tissue cumulative damage evolution equation as a physical constraint into the loss function, an intelligent prediction model is developed, using FPI and running distance as core input parameters. Validation based on experimental data from 45 runners completing a half-marathon confirms that the proposed model significantly outperforms purely data-driven methods in both prediction accuracy and generalization capability. The results indicate that fatigue damage in lower limb tissues during long-distance running exhibits nonlinear, multi-stage biomechanical evolution characteristics, with different tissues showing spatiotemporal heterogeneity in their response to cumulative loading. The novel framework that integrates mechanical principles with deep learning effectively achieves dynamic mapping and mechanical trend prediction from kinematic parameters to internal tissue damage states. This study not only provides a deeper mechanical explanation for understanding the mechanisms behind running-related injuries but also offers important methodological support for developing personalized injury prevention and rehabilitation strategies based on biomechanical principles.