Analysis of bearing mechanism of large-diameter under-reamed piles based on model tests.

Given the insufficient understanding of the bearing mechanism and failure modes of large-diameter under-reamed piles in complex strata, this study conducted scaled laboratory model tests based on similarity theory. A visualized "semi-model pile static loading-reaction frame" system was established to systematically investigate the influence of under-reaming angle (0° ~ 25°) and pile embedment depth (60 ~ 80 cm) on the bearing characteristics and failure mechanisms of the pile foundation. The results show that: 1) The under-reaming angle is the dominant factor controlling bearing performance. A reasonable increase in this angle can significantly enhance the ultimate bearing capacity, with a 204.99% improvement observed at 20° compared to the uniform-section pile. However, the enhancement effect weakens with increasing embedment depth. Comprehensive analysis suggests that 15° ~ 20° is the optimal under-reaming angle range; an excessive angle induces stress concentration at the "shoulder" of the pile, leading to interfacial detachment between pile and soil, thus limiting further improvement in bearing capacity. 2) The under-reamed structure effectively optimizes the load transfer path along the pile shaft. The end bearing resistance ratio increases first and then decreases with the under-reaming angle, reaching a maximum of 65.09% at 20°, indicating a transition of the load transfer mechanism from shaft-resistance-dominated to end-resistance-dominated behavior. 3) The failure morphology of the pile toe rock evolves from "penetrative shear failure" in the uniform pile (failure zone ≈ 1.25D) to "fan-shaped compaction" at the optimal under-reaming angle (≈ 4D), and further enlarging the angle results in unstable "bulging and collapse" failure. This study systematically reveals the full-process mechanism from load bearing to failure in large-diameter under-reamed piles, providing a theoretical basis for optimizing design parameters and predicting failure behavior. The findings offer valuable references for engineering design and improvement of design codes.