Abstract: Coal seam15 is one of the main minable seams in Mine No.5 of Yangquan Coal Industry(Group) Co., Ltd., There is K₂ limestone developed in the roof of the seam. The limestone is not a good seismic reflection interface. It is difficult to trace continuously the limestone interface by conventional seismic profiles. The logging curves of K₂ limestone show high density and high apparent resistivity. We used integration of density and apparent resistivity logging curves to generate pseudo density curves, then to get data of formation lithology based on model inversion. So we can identify the geometrical shape and the thickness of the limestone. We selected 9 seismic attributes to constitute the training samples for neural network so as to conduct inversion of the neural network for the porosity and the apparent resistivity and finally to predict the aquosity of the limestone.