Abstract: In the prediction of tectonic coal thickness,the problem of low accuracy is often caused by various restrictive factors.Therefore,a method of optimizing the parameters of RBF neural networks by using adaptive genetic algorithm to optimize density clustering is used to predicte the thickness of tectonic coal.Firstly,the 3D seismic attribute data of the mining area are preprocessed,and the PCA algorithm is used to reduce the dimension and eliminate the linear correlation between variables.Then a RBF neural network model for predicting the thickness of tectonic coal is constructed,the genetic algorithm is used to optimize the density clustering to obtain the best core point,and the initial clustering center of k-means clustering is further calculated to optimize the k-means algorithm,so that the RBF neural network implicit layer basis function is obtained.An excellent center vector increases the accuracy and robustness of the model prediction.At the same time,aiming at the problem that genetic algorithm is easy to fall into local optimal problem,the global and local search ability of the genetic algorithm is improved by adaptively changing the crossover rate and the mutation rate with the increase of the number of evolutions,so that it can escape the local best advantage and obtain better evolutionary results.The L2 regularization term is added to effectively avoid the influence of noisy data for generalization performance of the model.Finally,the prosed model is applied to the 8# coal seam of the No.6 mining area of Luling Coal Mine.The predicted thickness of the model is highly consistent with the actual geological data.It is possible to promote the prediction of coal thickness in actual mining area.

Key words: Center vector, Density clustering, Genetic algorithm, Principal component analysis, Radial basis function neural network, Tectonic coal, Thickness prediction