Through the explanation of complex lithology character of a reservoir's bore log, INET is used to get the permeability. INET needs training data, which is composed of a certain number of inputs and the relevant target signals, then with the quick produced results, a comparison can be made among them. As the result shows, the data generated by the INET is much closer to the core data. INET can get the optimal network operating system in a short time and provide an inherent asperity permeability profile. By applying the delay depth network model, two samples were taken for each depth; the trained network can find the relation of present depth point with log parameter of its upper and lower layer and present depth core analysis parameter automatically. INET is a good way for the simulating and predicting of reservoir rock permeability and is also of valuable significance as the study of theoretical approach.