The grout injection forms a compacted backfill in the separation chamber, and most of the grouting water seeps into the overburden pores. The initial saturation of overlying rock has been found to affect the seepage distribution of effluent in overlying rock. In the actual grouting process, the flow direction of grouting water in overlying rock and the initial saturation of overlying rock may affect whether the bled water will leak to the working face. In order to investigate the influence of initial saturation on the seepage law of isolated overburden grout injection, the present paper studies the saturation change process of bled water seepage at three different initial saturations (S_r = 29%, 51%, 73%). The water leakage of the working face was monitored with the self-developed 3D visual simulation experimental system, without considering the overlying rock of the injection layer. The results show that the bled water flows mainly in the horizontal direction, and the horizontal seepage velocity is about 10 times the vertical seepage velocity. The higher the initial saturation of the overburden, the larger the seepage range of the overburden, and the closer the seepage boundary is to the working face, the easier the bled water will leak into the working face. Therefore, the grouting design and process control can be carried out better. When the thickness of the selected isolation layer is insufficient, the bled water may appear on the working face, so this problem can be improved by increasing the thickness of the isolation layer.

Traditional physical similarity simulation methods struggle to replicate the cumulative unloading–expansion effect in overburden, particularly due to inherent limitations in representing the bulking–compaction behavior of fractured rock masses in the caving zone. This significantly hinders a deeper understanding of overburden movement mechanisms. To address this technical challenge, this study innovatively proposes an experimental method designed to simulate the bulking–compaction process of rock masses in the caving zone. The method employs a composite of EPE and PP sheets. Through systematic uniaxial compression tests and orthogonal experimental design optimization, an optimal material mix ratio with superior performance was identified. Its stress–strain behavior was systematically analyzed, and its feasibility was comprehensively verified from the perspective of the synergistic evolution of displacement and stress fields. The results demonstrate that the stress–strain response of the new similar simulation material (SSM) aligns highly with the Salamon model. Furthermore, its load-bearing capacity exhibits a non-linear strengthening characteristic with increasing EPE thickness. Physical simulation validation tests, based on the engineering context of the Shilawusu Coal Mine, showed that all the relative error parameters were strictly controlled within 12%. The overall accuracy was significantly superior to existing simulation methods, achieving a substantial reduction in prediction errors for key parameters.