Closed-loop intelligent optimization of blasting-induced permeability enhancement in low-permeability sandstone-type uranium deposits: Coupling boundary conditions, fracture features, and flow rate

This study proposes a closed-loop framework that integrates physical simulation, data-driven analysis, and intelligent optimization to improve the permeability and fluid-injection efficiency of low-permeability sandstone uranium deposits. First, the evolution of the fracture network under different boundary conditions (uncoupling coefficient, short delay time, and pressure rise time) was simulated using the finite discrete element method (FDEM). Subsequently, key fracture characteristics were extracted from binarized and skeletonized images using ImageJ, and principal parameters were selected based on Pearson correlation analysis and principal component analysis (PCA). Second, a conditional generative adversarial network (cGAN) was constructed to generate high-fidelity fracture images based on blasting boundary conditions, thereby expanding the dataset. Finally, a three-stage machine learning model was developed. In Stage 1, fracture features were inverted from the boundary conditions. In Stage 2, the model predicts the pumping-hole flow rate based on fracture characteristics, original boundary conditions, and injection pressure, achieving high prediction accuracy (R2 = 0.98). In Stage 3, a differential–evolution algorithm is employed to jointly optimize boundary conditions and injection pressure, increasing the extraction-hole flow rate at 5 m from 21.22 × 10−6 m3·s−1 to 31.97 × 10−6 m3·s−1; the predicted peak flow at the field center shows an error of less than 6% relative to measurements. The results demonstrate that this closed-loop framework not only efficiently characterizes the relationship between fracture network features and seepage response during the blasting-induced permeability enhancement process but also provides reliable decision support for designing blasting and fluid injection schemes in low-permeability sandstone uranium deposits.