Abstract: By analyzing the limitations of the traditional logging data interpretation methods,we proposed an artificial-intellig ence-neural-net work-based method for lithology identification and porosity and permeability prediction according to the mechanisms and charact eristics of neural net works.Atraining patter matching from properlogging data vectors is selected first,and then alearning sample union from several training pattern matchings is constituted,making the networks remember this characteristics and format the prediction model by learning this sam pleunion;finally,the required parameters calculated.Lithology identification,and porosity and permeability prediction of the main set of uraniumore body occurring in Shihongtan uranium deposit with this method are consistent with real documentation.Practical application of this approach shows that it is feasible for sandst one uranium deposit.Compared with traditional methods,the approach does not require establishing concrete interpretation model and computational formula. As aresult,a better adaptability and higher accuracy of prediction are obtained.The approach is valuable practice.