Predicting Hydraulic Conductivity of Rock Masses using Machine Learning and Hydro-Jacking Tests – A Case Study of Mohmand Dam

Hydraulic conductivity of fractured rock masses is a controlling parameter in dam engineering, governing seepage and grouting performance. In practice, hydraulic conductivity is commonly evaluated using in-situ packer (Lugeon) or commonly known hydro-jacking tests. However, these tests are costly, time consuming and cumbersome, requiring skilled technical staff. Therefore, empirical models are often used to estimate the hydraulic conductivity, which generally rely on a limited number of input variables, and therefore inadequately represent the nonlinear permeability behavior of rock masses. To address these limitations, this study proposes a machine-learning based modeling framework for predicting hydraulic conductivity of fractured rock masses using published data comprising packer test results, hydro-jacking reopening pressures, and geological parameters including depth, rock quality designation (RQD), and fracture characteristics. Hydro-jacking tests are performed at the Mohmand dam site, and the model performance is evaluated against the test data. The results indicate that the machine-learning based model is reliable and can accurately capture hydraulic conductivity in fractured rock masses. The proposed approach offers a reliable alternative to traditional empirical methods and has practical implications for seepage assessment, grouting design, and dam foundation permeability evaluation in complex geological settings.