Response properties of geometries of coal penetrating fracture on seepage behavior

The fracture surfaces of coal-rock masses formed under mining-induced stress generally exhibit complex geometries, and the fracture geometry is one of the primary factors affecting the seepage characteristics of coal-rock penetrating fracture. This paper investigates the seepage characteristics of 5 groups of coal penetrating fracture (CPF) with different joint roughness coefficients (JRCs). Based on 3D morphology scanner tests and hydraulic coupling tests, a characterization method of effective geometric parameters in fracture surfaces under various confining pressures was improved, and a relationship between effective geometric parameters and the confining pressure is established. The results indicate that the nonlinear flow behavior in a CPF primarily includes three types: non-Newtonian fluid seepage under high confining pressure and low JRC, non-Darcy seepage under low confining pressure and high JRC, and the whole process of seepage characteristics between these two conditions. Among them, non-Newtonian fluid seepage is caused by significant fracture expansion, while non-Darcy seepage can be attributed to turbulence effects. During the seepage process, the geometric parameters with different JRC fracture samples all exhibit exponential changes with the increase of confining pressure. In addition, under high confining pressure, the effective contact ratio, effective fracture aperture, and void deviation ratio with high JRC fracture samples under high confining pressure increase by 93.5%, 67.4%, and 24.9%, respectively, compared with those of low JRC fracture samples. According to the variation of geometric parameters in a CPF with external stress, a seepage model considering geometric parameters in a CPF is proposed. By introducing the root mean square error (RMSE) and coefficient of determination (R²) to evaluate the error and goodness of fit between model curves and experimental data, it is found that the theoretical curves of model in this paper have the best matching with the experimental data. The average values of RMSE and R² for model in this paper are 0.002 and 0.70, respectively, which are better than models in the existing literature.