oa Применение искусственных нейронных сетей к скважинным данным для картирования литофаций плиоцена, плейстоцена и голоцена

Successful artificial neural network analyses of the lithology data – prediction of either sandstones or shale were obtained. These were performed in the youngest (Pliocene, Pleistocene and Holocene) sediments in Sava Depression, Croatian part of the Pannonian Basin, and the input data was based on well log curves (spontaneous potential, shallow and deep resistivity curves). These analyses were successfully preformed in 20 wells of which in two, the training and validation of the networks was made. Such obtained lithology data was used for the construction of lithofacies maps which clearly show the distribution of the overall thickness of sandstone bodies and their number. Palaeotransport directions and areas with thicker sandstone interval for possible gas accumulations can also be observed in the constructed maps. The main influx of sediment was from the NE trough a channel which coincides with recent structural low defined by the absence of Pliocene outcrops while the thickest sandstone intervals can be delimited to the SW part of the mapping area.