Abstract: In order to study the trend of structural soft coal in Sijiazhuang coal mine, the PNN neural network inversion technique is used to invert the lithology of coal seam roof in the study area, so as to predict the structure of soft coal. Taking natural gamma curve as a constraint condition, by analyzing the response characteristics of the sandstone, mudstone density and gamma ray of the target area, the PNN inversion of the post stack 3D seismic data is carried out. The results show that the roof sandstone of No. 15 coal seam corresponds to low natural gamma value and positively correlated with the development of structural soft coal. By studying the distribution law of sandstone in coal seam roof, the spatial development characteristics of tectonic soft coal can be indirectly predicted.