Abstract: For in-situ leaching of fragmented uranium ore by blasting, the ore heap is constructed by drilling and blasting and is a loose fragmented medium. Its particle size distribution obeys the Rosin-Rammler distribution law, and its characteristic particle size, particle size distribution index and porous ratio vary with the employed blasting techniques and technical parameters. In order to study the regularities for saturated water seepage in the medium, a self-designed apparatus is used to perform tests on seven samples compounded by using the Rosin-Rammler distribution law. And the adaptive neuro-fuzzy inference system (ANFIS) is used to establish the ANFIS models for predicting the seepage ratio and flow state index on the basis of the test results. The research results show that the saturated water seepage in the medium is Non-Darcy flow, that the relationship between the seepage velocity and the hydraulic gradient can be described in an exponential function, and the ANFIS models can give highly accurate predictions. It will provide a new approach to the estimation of the seepage ratio and flow state index.