Research On Prediction Of Surface Subsidence Based On The Adaboost Improved BP Neural Network

The prediction of surface subsidence in goaf is the key part of mine surface disaster research.For areas with rich coal resources,its research significance is particularly important.In order to reduce the adverse effects of surface subsidence in the goaf on the surface structures,highways,water bodies,railways and other infrastructures,it is of great practical significance to predict the surface subsidence of the goaf.Based on the movement law of surface subsidence deformation in goaf,the influencing factors of surface subsidence in goaf were analyzed from geological,mining and topographical aspects,and the correlation between each influencing factor and the maximum sinking amount of the surface were analyzed by grey correlation analysis.According to the degree of correlation,the mechanical properties of overlying strata,coal seam dip,mining depth,mining thickness,goaf size,Working face propulsion speed,mining method and roof management method,repeated mining wree selected as the main influencing factors of surface subsidence.And The weights of each major influencing factor were obtained by using Analytic Hierarchy Process.The paper used BP neural network to predict the surface subsidence of goaf.Because BP neural network algorithm has slow convergence speed and poor generalization ability,so Adaboost algorithm was introduced to improve BP neural network,and Adaboost improved BP neural network prediction model was constructed.It was proposed to improve the prediction efficiency and accuracy of surface subsidence prediction by this model,and provide a new way for surface subsidence prediction in goaf.In the calculation process,according to the sample data of the selected influencing factors,Principal Component Analysis(PCA)was used to further process each impact factor,simplify the relationship between various factors,and convert the impact factor into several integrated variables as input variables.Input variables and output variables were imported into Adaboost improved BP neural network prediction model to predict surface subsidence.Finally,it was compared and analyzed with the calculation results of GA-BP neural network and PSG-BP neural network prediction model respectively.The results showed that Adaboost improved BP neural network prediction model had better prediction effect and was more suitable for surface subsidence in goaf with strong engineering practicability.