Zicheng Yang

Covered karst collapse is a key geotechnical hazard in infrastructure construction in karst regions of China. In particular, strata consisting of an overlying clay layer and an underlying sand layer are prone to abrupt collapse induced by sand leakage under construction disturbances, which poses serious risks to pile foundation safety. To clarify the disaster-forming mechanism and develop a quantitative analysis method, this study investigates the mechanical behaviour of the entire collapse process by combining theoretical analysis with numerical simulation. A continuous mechanical analysis framework is established that follows the sequence from sand layer leakage to cavity expansion and then clay layer instability. Within this framework, a calculation model for the angle of repose of the sand layer is proposed that considers seepage and confined pressure effects. Simultaneously accounting for the influence of the casing, stability models for overall and localised collapses are developed using limit equilibrium theory. A comprehensive safety factor criterion K_c based on the critical span (or radius) is then proposed, leading to a linked evaluation method that couples the potential span of the sand layer with the ultimate span of the clay layer. The results show that an increase in Δh/h significantly reduces the angle of repose of the sand layer; the mechanical mechanism is confirmed whereby an increase in the roof span leads to shear stress exceeding the soil’s shear strength, thus triggering instability; the proposed safety factor K_c can effectively predict both overall and localised collapse, and case verification demonstrates that the predicted spans match well with actual collapse dimensions. The results provide a theoretical and technical basis for risk prediction, as well as for the prevention and control of pile foundation construction in karst areas.

The spatial characteristics of the dry beach in tailings ponds are critical indicators for the safety assessment of tailings dams. This study presents a method for dry beach extraction that combines deep learning-based semantic segmentation with 3D reconstruction, overcoming the limitations of 2D methods in spatial analysis. The workflow includes four steps: (1) High-resolution 3D point clouds are reconstructed from UAV images, and the projection matrix of each image is derived to link 2D pixels with 3D points. (2) AlexNet and GoogLeNet are employed to extract image features and automatically select images containing the dry beach boundary. (3) A DeepLabv3+ network is trained on manually labeled samples to perform semantic segmentation of the dry beach, with a lightweight incremental training strategy for enhanced adaptability. (4) Boundary pixels are detected and back-projected into 3D space to generate consistent point cloud boundaries. The method was validated on two-phase UAV datasets from a tailings pond in Yunnan Province, China. In phase I, the model achieved high segmentation performance, with a mean Accuracy and IoU of approximately 0.95 and a BF of 0.8267. When applied to phase II without retraining, the model maintained stable performance on dam boundaries, while slight performance degradation was observed on hillside and water boundaries. The 3D back-projection converted 2D boundary pixels into 3D coordinates, enabling the extraction of dry beach point clouds and supporting reliable dry beach length monitoring and deposition morphology analysis.