Permeability evolution of uranium reservoirs under multi-field coupling (gas-temperature-force-seepage) and artificial intelligence multi-parameter target optimization study

In insitu leaching uranium mining, insufficient permeability of the rock formation is a primary bottleneck limiting uranium resource extraction. This study focuses on a sandstone-type uranium mine in Xinjiang, where a coupled gas-temperature-force-seepage model was developed using the finite discrete element method. The rock's damage mechanism and permeability evolution during blasting for permeability enhancement were systematically analyzed. First, a three-factor, five-level orthogonal numerical experiment was designed to quantify the effects of the uncoupling coefficient, short delay time, and pressure rise time on fracture network expansion. The optimal blasting parameter combinations for reservoir modification were identified, and the regulatory effects of geostress and blasting sequence on fracture formation and connectivity were elucidated. Second, the effect of injection pressure on the reservoir's leaching range and flow distribution was analyzed, providing theoretical support for the optimization of injection parameters. Finally, an optimization framework combining machine learning and genetic algorithms was introduced to further enhance the flow rate. The framework accurately predicts the flow rate and optimizes blasting parameter combinations using the eXtreme Gradient Boosting model. The results show that a maximum flow rate of 3.2491 × 10−4 m3 s−1 can be achieved under various parameter combinations, demonstrating the robustness and broad applicability of the optimization framework. This study provides additional insights into blasting for permeability enhancement in sandstone-type uranium reservoirs.