The Relationships Between Rock Elements And The Igneous Rocks, The Lithologic Discrimination And Mineral Identification Of Sedimentary Rocks:A Study Based On The Method Of Artificial Neural Network

Lithology identification is the prerequisite and foundation work of well logging and e-logging interpretation. As the advances in oil and gas exploration, the unconventional oil and gas reservoirs accumulated in the complex lithology by become the key target of this field. Two major methods, the Fluorescence Spectrum Analysis for oil logging and elemental capture spectroscopy for e-logging have been developed for the complex lithology. The common nature of both methods are based on the examination of the major elements of strata; however, it is difficult to acquire the quantitative data of the lithology and the contents of the minerals.Some foreign oilfield technology service companies, represented by Schlumberger Company, had done a lot of research work on the lithologic discrimination using elements of strata. However, all known documents reveal that their works were concentrated on the sedimentary rocks, instead of igneous rocks. Concerning the complicated geological background of China, it is necessary and critical to carry on such research particularly on the relationships between rock elements and their mineral contents.On the basis of the 7834 data of the element composition of igneous rocks from 27 provinces, and 9066 data of X-ray from 13 different types of continental petroiliferous sedimentary basins, we investigated the complex matching relationships between rock elements and their minerals using Artificial Neural Network, especially multi-layer perception and BP neural networks in the methods of nonlinear mathematical classification. Using multi-layer perception neural networks, we constructed a model to discriminate the lithology of both sedimentary and igenous rocks based on the element composition. A quantitative predication model was also established to evaluate the mineral composition of sedimentary rocks based on the elemental composition. We trained the models with a large amount of test and training, and developed some applications using Studio.NET 2003. We achieved three major results as follows.According to the element compositions of Si, Ca, Al, Mg, K and other major elements of the rocks, our models has an accuracy of 94.3% on identification of igneous and sedimentary rocks.The quantitative predication model can discriminate the contents of many mineral including quartz, feldspar, clay, calcite, dolomite, etc. The mean absolute error for quartz is ca. 1.72,4.5 for feldspar, and 1.72-4.5 for other minerals.We analyzed the logging data with our models and applications, it is well consistent with experimental results and the results of some foreign oilfield technology service companies.