Prediction of coal seam thickness based on spatial autoregressive interpolation

LI Weilin; ZHAO Jialiang; RUAN Liutan; LI Zequan

doi:10.11799/ce202412050

Abstract：The accurate prediction of coal seam thickness is pivotal in enhancing the efficiency of coal mining operations,optimizing the layout of mines and facilitating the intelligent construction of coal mines.In the past,Kriging,Discrete Smooth Interpolation(DSI),and Inverse Distance Weighting(IDW)have been the most commonly used methods for calculating the interpolation of coal seam thickness or top and bottom plate contour.Because it is difficult to accurately capture the complex nonlinear relationship between spatial variables,these methods still have problems in predicting the thickness of coal seam.The spatial autoregressive method is employed to construct an interpolation model based on a fully connected neural network(FCNN),which is then validated on two distinct simulated datasets.The interpolation calculation was conducted using the model on the actual coal seam thickness data,resulting in the acquisition of continuity coal seam thickness data.The results demonstrate that FCNN exhibits a significant enhancement in model evaluation indices,including R?,RMSE,and MAE,in comparison to Kriging and IDW.It is evident that FCNN is capable of accurately modelling the intricate relationship between weights and spatial distances,thereby enhancing the precision and efficacy of coal seam thickness prediction.Moreover,FCNN offers a promising avenue for addressing similar spatial geologic prediction challenges.