Abstract
The movement of overburden strata under valley geomorphological conditions presents significant challenges for ground pressure control and the protection of surface ecosystems and aquifers. In this study, geological and boundary mechanical conditions were obtained through field investigation and monitoring, and were subsequently applied in both physical similarity simulation and numerical modeling. The research focuses on two mining sequences-steep gully to deep gully (SG→DG) and deep gully to steep gully (DG→SG)-to analyze the evolution of stress and displacement in overburden strata at different mining stages. A novel approach integrating deep learning techniques was employed to identify and analyze stress-displacement image patterns. Results indicate that when vertical stress is considered, the proportion of high-risk stress zones under the DG→SG method reaches 5.57%, compared to only 3.28% under the SG→DG method. This implies that the probability of stress-related hazards in the SG→DG sequence is approximately 58% of that in the DG→SG sequence. These findings suggest that the SG→DG method offers significant advantages in mitigating ground pressure risks. It is therefore recommended as the preferred mining approach, especially when combined with ground grouting and underground hydraulic fracturing techniques. Furthermore, the study reveals that mining direction is a critical human-controllable factor affecting overburden behavior. By investigating stress redistribution under different mining sequences, this research provides new insights into stress transfer mechanisms in overburden strata, offering theoretical guidance for the safe and efficient extraction of coal resources in complex valley terrain environments.