The Research And Application Of Selforganized Competitive Neural Network In Interpretation Of Well Logging Data In Sandstone Type Uranium Deposits

With the development of science and technology, logging interpretation methods and techniques are also increasing.Interpretation of logging data processing method tends to integration, man-machine interaction and automatic interpretation. The comprehensive interpretation of logging data, the method is more efficient, fast, accurate and has become the direction of the research of lithology identification. Lithology identification is a key part in the interpretation of logging curves. However, the traditional artificial recognition of lithology interpretation methods havea long cycle and low precision.Resulting in the subjective factors have great influence on the interpretation of the results. Therefore, this paper uses self-organizing competitive neural network recognition method to explain comprehensive lithology identification of well logging data.This paper studies the regional geological conditions of a basin uranium deposits.On the basis of the characteristics of sandstone type uranium deposit andthe rules of radioactive equilibrium, the relationship of formation lithology and the logging curve amplitude is summarized. Combined with previous research results, and analyzes the effects of two standard samples of index selection, the indicators of the samples are determined. According to the practical well logging data, with SP, LL3, CAL, DEN logging data as learning samples index, the standard sample of 6 kinds of common drilling data in lithology are established, that is, shale, siltstone, fine sandstone, sandstone,coarse sandstone and coal seam of the standard sample.By using the application of self organizing competitive neural network in the sandstone type uranium deposit in the drilling data, formation and its lithology aredivided. Research shows that self-organizing competitive neural network method in lithology recognition has faster convergence speed and higher recognition accuracy.Compared with the expert logging lithology section,automatic identification of the lithology are in accordance with the geological rules.According to the quantitative interpretation of nuclear logging data using hierarchical principle and deconvolution interpretation method, prepared for parameters related to the function code, wells abnormal mineralization layer is quantitativelyinterpreted.Calculation method in combination with production practice of the content, content of the given method and the three-point deconvolution method,five-pointdeconvolution method of calculating results are discussed. The results show that compared with content given interpretation method used in the production, the result of five-point deconvolution method are more accuracy thancontent of the given method.