Because of impact from the ancient river erosion , Luning Mine No. 2 coal seam thickness change was relatively large, from 0.35 ～ 6.8 m. In order to improve the prediction accuracy of thickness of the coal seam, the delineation of the ancient river scour zone, by using the method of BP artificial neural network inversion, seismic attributes sensitive independent by calculating the correlation coefficient of seismic attribute extraction, cross correlation analysis and validity analysis methods such as optimization, using optimized seismic attributes and drilling data, comprehensive training, forecast model of coal seam thickness was established, based on the forecast results and combined with seismic attribute, seismic time section successfully delineated the ancient river erosion zones, the result showed that BP artificial neural networks provided a new way for the prediction of coal seam thickness.