Novel approach to predict the spatial distributions of hydraulic conductivity of rock mass using convolutional neural networks

Characterizing the spatial distributions of hydraulic conductivity of rock mass is important in geoscience and engineering disciplines. In this paper, the architecture of CNN is proposed to predict the spatial distributions of hydraulic conductivity based on limited geologic factors. The performance of CNN model is evaluated using the new data of hydraulic conductivity. A comparative study with the empirical method is performed to validate the reliability of CNN model. The effect of weathering and unloading on the spatial distributions of hydraulic conductivity is studied using the CNN model. The result shows that the hydraulic conductivity predicted by CNN model is within the error range of 5% compared to the Lugeon borehole tests. The predictive accuracy of the CNN method is higher than the estimations of the empirical relations. The spatial distributions of hydraulic conductivity versus depth can be divided into three stages. At first stage, the hydraulic conductivity is slightly reduced with the increasing of depth. Increasing to the depth range of 300-600 m (second stage), the hydraulic conductivity is slightly reduced as a function of lower weathering degree. At last stage, the hydraulic conductivity is not changed by the weathering, and converge to a constant with the depth increasing.